Tilth context.

Co-authored-by: Mervap <megavaprold@gmail.com>
Co-authored-by: Sri Pasumarthi <sripas@google.com>

CONTRIBUTING.md

@@ -352,21 +352,6 @@ npm run lint
 - **Imports:** Pay special attention to import paths. The project uses ESLint to
   enforce restrictions on relative imports between packages.
 
-### Project structure
-
-- `packages/`: Contains the individual sub-packages of the project.
-  - `a2a-server`: A2A server implementation for the Gemini CLI. (Experimental)
-  - `cli/`: The command-line interface.
-  - `core/`: The core backend logic for the Gemini CLI.
-  - `test-utils` Utilities for creating and cleaning temporary file systems for
-    testing.
-  - `vscode-ide-companion/`: The Gemini CLI Companion extension pairs with
-    Gemini CLI.
-- `docs/`: Contains all project documentation.
-- `scripts/`: Utility scripts for building, testing, and development tasks.
-
-For more detailed architecture, see `docs/architecture.md`.
-
 ### Debugging
 
 #### VS Code

README.md

@@ -314,7 +314,6 @@ gemini
 
 - [**Headless Mode (Scripting)**](./docs/cli/headless.md) - Use Gemini CLI in
   automated workflows.
-- [**Architecture Overview**](./docs/architecture.md) - How Gemini CLI works.
 - [**IDE Integration**](./docs/ide-integration/index.md) - VS Code companion.
 - [**Sandboxing & Security**](./docs/cli/sandbox.md) - Safe execution
   environments.

docs/changelogs/preview.md

@@ -1,6 +1,6 @@
-# Preview release: v0.34.0-preview.4
+# Preview release: v0.35.0-preview.1
 
-Released: March 16, 2026
+Released: March 17, 2026
 
 Our preview release includes the latest, new, and experimental features. This
 release may not be as stable as our [latest weekly release](latest.md).
@@ -13,471 +13,364 @@ npm install -g @google/gemini-cli@preview
 
 ## Highlights
 
-- **Plan Mode Enabled by Default:** Plan Mode is now enabled out-of-the-box,
-  providing a structured planning workflow and keeping approved plans during
-  chat compression.
-- **Sandboxing Enhancements:** Added experimental LXC container sandbox support
-  and native gVisor (`runsc`) sandboxing for improved security and isolation.
-- **Tracker Visualization and Tools:** Introduced CRUD tools and visualization
-  for trackers, along with task tracker strategy improvements.
-- **Browser Agent Improvements:** Enhanced the browser agent with progress
-  emission, a new automation overlay, and additional integration tests.
-- **CLI and UI Updates:** Standardized semantic focus colors, polished shell
-  autocomplete rendering, unified keybinding infrastructure, and added custom
-  footer configuration options.
+- **Subagents & Architecture Enhancements**: Enabled subagents and laid the
+  foundation for subagent tool isolation. Added proxy routing support for remote
+  A2A subagents and integrated `SandboxManager` to sandbox all process-spawning
+  tools.
+- **CLI & UI Improvements**: Introduced customizable keyboard shortcuts and
+  support for literal character keybindings. Added missing vim mode motions and
+  CJK input support. Enabled code splitting and deferred UI loading for improved
+  performance.
+- **Context & Tools Optimization**: JIT context loading is now enabled by
+  default with deduplication for project memory. Introduced a model-driven
+  parallel tool scheduler and allowed safe tools to execute concurrently.
+- **Security & Extensions**: Implemented cryptographic integrity verification
+  for extension updates and added a `disableAlwaysAllow` setting to prevent
+  auto-approvals for enhanced security.
+- **Plan Mode & Web Fetch Updates**: Added an 'All the above' option for
+  multi-select AskUser questions in Plan Mode. Rolled out Stage 1 and Stage 2
+  security and consistency improvements for the `web_fetch` tool.
 
 ## What's Changed
 
-- fix(patch): cherry-pick 48130eb to release/v0.34.0-preview.3-pr-22665 to patch
-  version v0.34.0-preview.3 and create version 0.34.0-preview.4 by
+- feat(cli): customizable keyboard shortcuts by @scidomino in
+  [#21945](https://github.com/google-gemini/gemini-cli/pull/21945)
+- feat(core): Thread `AgentLoopContext` through core. by @joshualitt in
+  [#21944](https://github.com/google-gemini/gemini-cli/pull/21944)
+- chore(release): bump version to 0.35.0-nightly.20260311.657f19c1f by
   @gemini-cli-robot in
-  [#22719](https://github.com/google-gemini/gemini-cli/pull/22719)
-- fix(patch): cherry-pick 24adacd to release/v0.34.0-preview.2-pr-22332 to patch
-  version v0.34.0-preview.2 and create version 0.34.0-preview.3 by
-  @gemini-cli-robot in
-  [#22391](https://github.com/google-gemini/gemini-cli/pull/22391)
-- fix(patch): cherry-pick 8432bce to release/v0.34.0-preview.1-pr-22069 to patch
-  version v0.34.0-preview.1 and create version 0.34.0-preview.2 by
-  @gemini-cli-robot in
-  [#22205](https://github.com/google-gemini/gemini-cli/pull/22205)
-- fix(patch): cherry-pick 45faf4d to release/v0.34.0-preview.0-pr-22148
-  [CONFLICTS] by @gemini-cli-robot in
-  [#22174](https://github.com/google-gemini/gemini-cli/pull/22174)
-- feat(cli): add chat resume footer on session quit by @lordshashank in
-  [#20667](https://github.com/google-gemini/gemini-cli/pull/20667)
-- Support bold and other styles in svg snapshots by @jacob314 in
-  [#20937](https://github.com/google-gemini/gemini-cli/pull/20937)
-- fix(core): increase A2A agent timeout to 30 minutes by @adamfweidman in
-  [#21028](https://github.com/google-gemini/gemini-cli/pull/21028)
-- Cleanup old branches. by @jacob314 in
-  [#19354](https://github.com/google-gemini/gemini-cli/pull/19354)
-- chore(release): bump version to 0.34.0-nightly.20260303.34f0c1538 by
-  @gemini-cli-robot in
-  [#21034](https://github.com/google-gemini/gemini-cli/pull/21034)
-- feat(ui): standardize semantic focus colors and enhance history visibility by
-  @keithguerin in
-  [#20745](https://github.com/google-gemini/gemini-cli/pull/20745)
-- fix: merge duplicate imports in packages/core (3/4) by @Nixxx19 in
-  [#20928](https://github.com/google-gemini/gemini-cli/pull/20928)
-- Add extra safety checks for proto pollution by @jacob314 in
-  [#20396](https://github.com/google-gemini/gemini-cli/pull/20396)
-- feat(core): Add tracker CRUD tools & visualization by @anj-s in
-  [#19489](https://github.com/google-gemini/gemini-cli/pull/19489)
-- Revert "fix(ui): persist expansion in AskUser dialog when navigating options"
-  by @jacob314 in
-  [#21042](https://github.com/google-gemini/gemini-cli/pull/21042)
-- Changelog for v0.33.0-preview.0 by @gemini-cli-robot in
-  [#21030](https://github.com/google-gemini/gemini-cli/pull/21030)
-- fix: model persistence for all scenarios by @sripasg in
-  [#21051](https://github.com/google-gemini/gemini-cli/pull/21051)
-- chore/release: bump version to 0.34.0-nightly.20260304.28af4e127 by
-  @gemini-cli-robot in
-  [#21054](https://github.com/google-gemini/gemini-cli/pull/21054)
-- Consistently guard restarts against concurrent auto updates by @scidomino in
-  [#21016](https://github.com/google-gemini/gemini-cli/pull/21016)
-- Defensive coding to reduce the risk of Maximum update depth errors by
-  @jacob314 in [#20940](https://github.com/google-gemini/gemini-cli/pull/20940)
-- fix(cli): Polish shell autocomplete rendering to be a little more shell native
-  feeling. by @jacob314 in
-  [#20931](https://github.com/google-gemini/gemini-cli/pull/20931)
-- Docs: Update plan mode docs by @jkcinouye in
-  [#19682](https://github.com/google-gemini/gemini-cli/pull/19682)
-- fix(mcp): Notifications/tools/list_changed support not working by @jacob314 in
-  [#21050](https://github.com/google-gemini/gemini-cli/pull/21050)
-- fix(cli): register extension lifecycle events in DebugProfiler by
-  @fayerman-source in
-  [#20101](https://github.com/google-gemini/gemini-cli/pull/20101)
-- chore(dev): update vscode settings for typescriptreact by @rohit-4321 in
-  [#19907](https://github.com/google-gemini/gemini-cli/pull/19907)
-- fix(cli): enable multi-arch docker builds for sandbox by @ru-aish in
-  [#19821](https://github.com/google-gemini/gemini-cli/pull/19821)
-- Changelog for v0.32.0 by @gemini-cli-robot in
-  [#21033](https://github.com/google-gemini/gemini-cli/pull/21033)
-- Changelog for v0.33.0-preview.1 by @gemini-cli-robot in
-  [#21058](https://github.com/google-gemini/gemini-cli/pull/21058)
-- feat(core): improve @scripts/copy_files.js autocomplete to prioritize
-  filenames by @sehoon38 in
-  [#21064](https://github.com/google-gemini/gemini-cli/pull/21064)
-- feat(sandbox): add experimental LXC container sandbox support by @h30s in
-  [#20735](https://github.com/google-gemini/gemini-cli/pull/20735)
-- feat(evals): add overall pass rate row to eval nightly summary table by
-  @gundermanc in
-  [#20905](https://github.com/google-gemini/gemini-cli/pull/20905)
-- feat(telemetry): include language in telemetry and fix accepted lines
-  computation by @gundermanc in
-  [#21126](https://github.com/google-gemini/gemini-cli/pull/21126)
-- Changelog for v0.32.1 by @gemini-cli-robot in
-  [#21055](https://github.com/google-gemini/gemini-cli/pull/21055)
-- feat(core): add robustness tests, logging, and metrics for CodeAssistServer
-  SSE parsing by @yunaseoul in
-  [#21013](https://github.com/google-gemini/gemini-cli/pull/21013)
-- feat: add issue assignee workflow by @kartikangiras in
-  [#21003](https://github.com/google-gemini/gemini-cli/pull/21003)
-- fix: improve error message when OAuth succeeds but project ID is required by
-  @Nixxx19 in [#21070](https://github.com/google-gemini/gemini-cli/pull/21070)
-- feat(loop-reduction): implement iterative loop detection and model feedback by
-  @aishaneeshah in
-  [#20763](https://github.com/google-gemini/gemini-cli/pull/20763)
-- chore(github): require prompt approvers for agent prompt files by @gundermanc
-  in [#20896](https://github.com/google-gemini/gemini-cli/pull/20896)
-- Docs: Create tools reference by @jkcinouye in
-  [#19470](https://github.com/google-gemini/gemini-cli/pull/19470)
-- fix(core, a2a-server): prevent hang during OAuth in non-interactive sessions
-  by @spencer426 in
-  [#21045](https://github.com/google-gemini/gemini-cli/pull/21045)
-- chore(cli): enable deprecated settings removal by default by @yashodipmore in
-  [#20682](https://github.com/google-gemini/gemini-cli/pull/20682)
-- feat(core): Disable fast ack helper for hints. by @joshualitt in
-  [#21011](https://github.com/google-gemini/gemini-cli/pull/21011)
-- fix(ui): suppress redundant failure note when tool error note is shown by
-  @NTaylorMullen in
-  [#21078](https://github.com/google-gemini/gemini-cli/pull/21078)
-- docs: document planning workflows with Conductor example by @jerop in
-  [#21166](https://github.com/google-gemini/gemini-cli/pull/21166)
-- feat(release): ship esbuild bundle in npm package by @genneth in
-  [#19171](https://github.com/google-gemini/gemini-cli/pull/19171)
-- fix(extensions): preserve symlinks in extension source path while enforcing
-  folder trust by @galz10 in
-  [#20867](https://github.com/google-gemini/gemini-cli/pull/20867)
-- fix(cli): defer tool exclusions to policy engine in non-interactive mode by
-  @EricRahm in [#20639](https://github.com/google-gemini/gemini-cli/pull/20639)
-- fix(ui): removed double padding on rendered content by @devr0306 in
-  [#21029](https://github.com/google-gemini/gemini-cli/pull/21029)
-- fix(core): truncate excessively long lines in grep search output by
-  @gundermanc in
-  [#21147](https://github.com/google-gemini/gemini-cli/pull/21147)
-- feat: add custom footer configuration via `/footer` by @jackwotherspoon in
-  [#19001](https://github.com/google-gemini/gemini-cli/pull/19001)
-- perf(core): fix OOM crash in long-running sessions by @WizardsForgeGames in
-  [#19608](https://github.com/google-gemini/gemini-cli/pull/19608)
-- refactor(cli): categorize built-in themes into dark/ and light/ directories by
-  @JayadityaGit in
-  [#18634](https://github.com/google-gemini/gemini-cli/pull/18634)
-- fix(core): explicitly allow codebase_investigator and cli_help in read-only
-  mode by @Adib234 in
-  [#21157](https://github.com/google-gemini/gemini-cli/pull/21157)
-- test: add browser agent integration tests by @kunal-10-cloud in
-  [#21151](https://github.com/google-gemini/gemini-cli/pull/21151)
-- fix(cli): fix enabling kitty codes on Windows Terminal by @scidomino in
-  [#21136](https://github.com/google-gemini/gemini-cli/pull/21136)
-- refactor(core): extract shared OAuth flow primitives from MCPOAuthProvider by
-  @SandyTao520 in
-  [#20895](https://github.com/google-gemini/gemini-cli/pull/20895)
-- fix(ui): add partial output to cancelled shell UI by @devr0306 in
-  [#21178](https://github.com/google-gemini/gemini-cli/pull/21178)
-- fix(cli): replace hardcoded keybinding strings with dynamic formatters by
-  @scidomino in [#21159](https://github.com/google-gemini/gemini-cli/pull/21159)
-- DOCS: Update quota and pricing page by @g-samroberts in
-  [#21194](https://github.com/google-gemini/gemini-cli/pull/21194)
-- feat(telemetry): implement Clearcut logging for startup statistics by
-  @yunaseoul in [#21172](https://github.com/google-gemini/gemini-cli/pull/21172)
-- feat(triage): add area/documentation to issue triage by @g-samroberts in
-  [#21222](https://github.com/google-gemini/gemini-cli/pull/21222)
-- Fix so shell calls are formatted by @jacob314 in
-  [#21237](https://github.com/google-gemini/gemini-cli/pull/21237)
-- feat(cli): add native gVisor (runsc) sandboxing support by @Zheyuan-Lin in
-  [#21062](https://github.com/google-gemini/gemini-cli/pull/21062)
-- docs: use absolute paths for internal links in plan-mode.md by @jerop in
-  [#21299](https://github.com/google-gemini/gemini-cli/pull/21299)
-- fix(core): prevent unhandled AbortError crash during stream loop detection by
-  @7hokerz in [#21123](https://github.com/google-gemini/gemini-cli/pull/21123)
-- fix:reorder env var redaction checks to scan values first by @kartikangiras in
-  [#21059](https://github.com/google-gemini/gemini-cli/pull/21059)
-- fix(acp): rename --experimental-acp to --acp & remove Zed-specific refrences
-  by @skeshive in
-  [#21171](https://github.com/google-gemini/gemini-cli/pull/21171)
-- feat(core): fallback to 2.5 models with no access for toolcalls by @sehoon38
-  in [#21283](https://github.com/google-gemini/gemini-cli/pull/21283)
-- test(core): improve testing for API request/response parsing by @sehoon38 in
-  [#21227](https://github.com/google-gemini/gemini-cli/pull/21227)
-- docs(links): update docs-writer skill and fix broken link by @g-samroberts in
-  [#21314](https://github.com/google-gemini/gemini-cli/pull/21314)
-- Fix code colorizer ansi escape bug. by @jacob314 in
-  [#21321](https://github.com/google-gemini/gemini-cli/pull/21321)
-- remove wildcard behavior on keybindings by @scidomino in
-  [#21315](https://github.com/google-gemini/gemini-cli/pull/21315)
-- feat(acp): Add support for AI Gateway auth by @skeshive in
-  [#21305](https://github.com/google-gemini/gemini-cli/pull/21305)
-- fix(theme): improve theme color contrast for macOS Terminal.app by @clocky in
-  [#21175](https://github.com/google-gemini/gemini-cli/pull/21175)
-- feat (core): Implement tracker related SI changes by @anj-s in
-  [#19964](https://github.com/google-gemini/gemini-cli/pull/19964)
-- Changelog for v0.33.0-preview.2 by @gemini-cli-robot in
-  [#21333](https://github.com/google-gemini/gemini-cli/pull/21333)
-- Changelog for v0.33.0-preview.3 by @gemini-cli-robot in
-  [#21347](https://github.com/google-gemini/gemini-cli/pull/21347)
-- docs: format release times as HH:MM UTC by @pavan-sh in
-  [#20726](https://github.com/google-gemini/gemini-cli/pull/20726)
-- fix(cli): implement --all flag for extensions uninstall by @sehoon38 in
-  [#21319](https://github.com/google-gemini/gemini-cli/pull/21319)
-- docs: fix incorrect relative links to command reference by @kanywst in
-  [#20964](https://github.com/google-gemini/gemini-cli/pull/20964)
-- documentiong ensures ripgrep by @Jatin24062005 in
-  [#21298](https://github.com/google-gemini/gemini-cli/pull/21298)
-- fix(core): handle AbortError thrown during processTurn by @MumuTW in
-  [#21296](https://github.com/google-gemini/gemini-cli/pull/21296)
-- docs(cli): clarify ! command output visibility in shell commands tutorial by
-  @MohammedADev in
-  [#21041](https://github.com/google-gemini/gemini-cli/pull/21041)
-- fix: logic for task tracker strategy and remove tracker tools by @anj-s in
-  [#21355](https://github.com/google-gemini/gemini-cli/pull/21355)
-- fix(partUtils): display media type and size for inline data parts by @Aboudjem
-  in [#21358](https://github.com/google-gemini/gemini-cli/pull/21358)
-- Fix(accessibility): add screen reader support to RewindViewer by @Famous077 in
-  [#20750](https://github.com/google-gemini/gemini-cli/pull/20750)
-- fix(hooks): propagate stopHookActive in AfterAgent retry path (#20426) by
-  @Aarchi-07 in [#20439](https://github.com/google-gemini/gemini-cli/pull/20439)
-- fix(core): deduplicate GEMINI.md files by device/inode on case-insensitive
-  filesystems (#19904) by @Nixxx19 in
-  [#19915](https://github.com/google-gemini/gemini-cli/pull/19915)
-- feat(core): add concurrency safety guidance for subagent delegation (#17753)
-  by @abhipatel12 in
-  [#21278](https://github.com/google-gemini/gemini-cli/pull/21278)
-- feat(ui): dynamically generate all keybinding hints by @scidomino in
-  [#21346](https://github.com/google-gemini/gemini-cli/pull/21346)
-- feat(core): implement unified KeychainService and migrate token storage by
-  @ehedlund in [#21344](https://github.com/google-gemini/gemini-cli/pull/21344)
-- fix(cli): gracefully handle --resume when no sessions exist by @SandyTao520 in
-  [#21429](https://github.com/google-gemini/gemini-cli/pull/21429)
-- fix(plan): keep approved plan during chat compression by @ruomengz in
-  [#21284](https://github.com/google-gemini/gemini-cli/pull/21284)
-- feat(core): implement generic CacheService and optimize setupUser by @sehoon38
-  in [#21374](https://github.com/google-gemini/gemini-cli/pull/21374)
-- Update quota and pricing documentation with subscription tiers by @srithreepo
-  in [#21351](https://github.com/google-gemini/gemini-cli/pull/21351)
-- fix(core): append correct OTLP paths for HTTP exporters by
-  @sebastien-prudhomme in
-  [#16836](https://github.com/google-gemini/gemini-cli/pull/16836)
-- Changelog for v0.33.0-preview.4 by @gemini-cli-robot in
-  [#21354](https://github.com/google-gemini/gemini-cli/pull/21354)
-- feat(cli): implement dot-prefixing for slash command conflicts by @ehedlund in
-  [#20979](https://github.com/google-gemini/gemini-cli/pull/20979)
-- refactor(core): standardize MCP tool naming to mcp\_ FQN format by
-  @abhipatel12 in
-  [#21425](https://github.com/google-gemini/gemini-cli/pull/21425)
-- feat(cli): hide gemma settings from display and mark as experimental by
-  @abhipatel12 in
-  [#21471](https://github.com/google-gemini/gemini-cli/pull/21471)
-- feat(skills): refine string-reviewer guidelines and description by @clocky in
-  [#20368](https://github.com/google-gemini/gemini-cli/pull/20368)
-- fix(core): whitelist TERM and COLORTERM in environment sanitization by
-  @deadsmash07 in
-  [#20514](https://github.com/google-gemini/gemini-cli/pull/20514)
-- fix(billing): fix overage strategy lifecycle and settings integration by
-  @gsquared94 in
-  [#21236](https://github.com/google-gemini/gemini-cli/pull/21236)
-- fix: expand paste placeholders in TextInput on submit by @Jefftree in
-  [#19946](https://github.com/google-gemini/gemini-cli/pull/19946)
-- fix(core): add in-memory cache to ChatRecordingService to prevent OOM by
-  @SandyTao520 in
-  [#21502](https://github.com/google-gemini/gemini-cli/pull/21502)
-- feat(cli): overhaul thinking UI by @keithguerin in
-  [#18725](https://github.com/google-gemini/gemini-cli/pull/18725)
-- fix(ui): unify Ctrl+O expansion hint experience across buffer modes by
-  @jwhelangoog in
-  [#21474](https://github.com/google-gemini/gemini-cli/pull/21474)
-- fix(cli): correct shell height reporting by @jacob314 in
-  [#21492](https://github.com/google-gemini/gemini-cli/pull/21492)
-- Make test suite pass when the GEMINI_SYSTEM_MD env variable or
-  GEMINI_WRITE_SYSTEM_MD variable happens to be set locally/ by @jacob314 in
-  [#21480](https://github.com/google-gemini/gemini-cli/pull/21480)
-- Disallow underspecified types by @gundermanc in
-  [#21485](https://github.com/google-gemini/gemini-cli/pull/21485)
-- refactor(cli): standardize on 'reload' verb for all components by @keithguerin
-  in [#20654](https://github.com/google-gemini/gemini-cli/pull/20654)
-- feat(cli): Invert quota language to 'percent used' by @keithguerin in
-  [#20100](https://github.com/google-gemini/gemini-cli/pull/20100)
-- Docs: Add documentation for notifications (experimental)(macOS) by @jkcinouye
-  in [#21163](https://github.com/google-gemini/gemini-cli/pull/21163)
-- Code review comments as a pr by @jacob314 in
-  [#21209](https://github.com/google-gemini/gemini-cli/pull/21209)
-- feat(cli): unify /chat and /resume command UX by @LyalinDotCom in
-  [#20256](https://github.com/google-gemini/gemini-cli/pull/20256)
-- docs: fix typo 'allowslisted' -> 'allowlisted' in mcp-server.md by
+  [#21966](https://github.com/google-gemini/gemini-cli/pull/21966)
+- refactor(a2a): remove legacy CoreToolScheduler by @adamfweidman in
+  [#21955](https://github.com/google-gemini/gemini-cli/pull/21955)
+- feat(ui): add missing vim mode motions (X, ~, r, f/F/t/T, df/dt and friends)
+  by @aanari in [#21932](https://github.com/google-gemini/gemini-cli/pull/21932)
+- Feat/retry fetch notifications by @aishaneeshah in
+  [#21813](https://github.com/google-gemini/gemini-cli/pull/21813)
+- fix(core): remove OAuth check from handleFallback and clean up stray file by
+  @sehoon38 in [#21962](https://github.com/google-gemini/gemini-cli/pull/21962)
+- feat(cli): support literal character keybindings and extended Kitty protocol
+  keys by @scidomino in
+  [#21972](https://github.com/google-gemini/gemini-cli/pull/21972)
+- fix(ui): clamp cursor to last char after all NORMAL mode deletes by @aanari in
+  [#21973](https://github.com/google-gemini/gemini-cli/pull/21973)
+- test(core): add missing tests for prompts/utils.ts by @krrishverma1805-web in
+  [#19941](https://github.com/google-gemini/gemini-cli/pull/19941)
+- fix(cli): allow scrolling keys in copy mode (Ctrl+S selection mode) by
+  @nsalerni in [#19933](https://github.com/google-gemini/gemini-cli/pull/19933)
+- docs(cli): add custom keybinding documentation by @scidomino in
+  [#21980](https://github.com/google-gemini/gemini-cli/pull/21980)
+- docs: fix misleading YOLO mode description in defaultApprovalMode by
   @Gyanranjan-Priyam in
-  [#21665](https://github.com/google-gemini/gemini-cli/pull/21665)
-- fix(core): display actual graph output in tracker_visualize tool by @anj-s in
-  [#21455](https://github.com/google-gemini/gemini-cli/pull/21455)
-- fix(core): sanitize SSE-corrupted JSON and domain strings in error
-  classification by @gsquared94 in
-  [#21702](https://github.com/google-gemini/gemini-cli/pull/21702)
-- Docs: Make documentation links relative by @diodesign in
-  [#21490](https://github.com/google-gemini/gemini-cli/pull/21490)
-- feat(cli): expose /tools desc as explicit subcommand for discoverability by
-  @aworki in [#21241](https://github.com/google-gemini/gemini-cli/pull/21241)
-- feat(cli): add /compact alias for /compress command by @jackwotherspoon in
-  [#21711](https://github.com/google-gemini/gemini-cli/pull/21711)
-- feat(plan): enable Plan Mode by default by @jerop in
-  [#21713](https://github.com/google-gemini/gemini-cli/pull/21713)
-- feat(core): Introduce `AgentLoopContext`. by @joshualitt in
-  [#21198](https://github.com/google-gemini/gemini-cli/pull/21198)
-- fix(core): resolve symlinks for non-existent paths during validation by
-  @Adib234 in [#21487](https://github.com/google-gemini/gemini-cli/pull/21487)
-- docs: document tool exclusion from memory via deny policy by @Abhijit-2592 in
-  [#21428](https://github.com/google-gemini/gemini-cli/pull/21428)
-- perf(core): cache loadApiKey to reduce redundant keychain access by @sehoon38
-  in [#21520](https://github.com/google-gemini/gemini-cli/pull/21520)
-- feat(cli): implement /upgrade command by @sehoon38 in
-  [#21511](https://github.com/google-gemini/gemini-cli/pull/21511)
-- Feat/browser agent progress emission by @kunal-10-cloud in
-  [#21218](https://github.com/google-gemini/gemini-cli/pull/21218)
-- fix(settings): display objects as JSON instead of [object Object] by
-  @Zheyuan-Lin in
-  [#21458](https://github.com/google-gemini/gemini-cli/pull/21458)
-- Unmarshall update by @DavidAPierce in
-  [#21721](https://github.com/google-gemini/gemini-cli/pull/21721)
-- Update mcp's list function to check for disablement. by @DavidAPierce in
-  [#21148](https://github.com/google-gemini/gemini-cli/pull/21148)
-- robustness(core): static checks to validate history is immutable by @jacob314
-  in [#21228](https://github.com/google-gemini/gemini-cli/pull/21228)
-- refactor(cli): better react patterns for BaseSettingsDialog by @psinha40898 in
-  [#21206](https://github.com/google-gemini/gemini-cli/pull/21206)
-- feat(security): implement robust IP validation and safeFetch foundation by
-  @alisa-alisa in
-  [#21401](https://github.com/google-gemini/gemini-cli/pull/21401)
-- feat(core): improve subagent result display by @joshualitt in
-  [#20378](https://github.com/google-gemini/gemini-cli/pull/20378)
-- docs: fix broken markdown syntax and anchor links in /tools by @campox747 in
-  [#20902](https://github.com/google-gemini/gemini-cli/pull/20902)
-- feat(policy): support subagent-specific policies in TOML by @akh64bit in
-  [#21431](https://github.com/google-gemini/gemini-cli/pull/21431)
-- Add script to speed up reviewing PRs adding a worktree. by @jacob314 in
-  [#21748](https://github.com/google-gemini/gemini-cli/pull/21748)
-- fix(core): prevent infinite recursion in symlink resolution by @Adib234 in
-  [#21750](https://github.com/google-gemini/gemini-cli/pull/21750)
-- fix(docs): fix headless mode docs by @ame2en in
-  [#21287](https://github.com/google-gemini/gemini-cli/pull/21287)
-- feat/redesign header compact by @jacob314 in
-  [#20922](https://github.com/google-gemini/gemini-cli/pull/20922)
-- refactor: migrate to useKeyMatchers hook by @scidomino in
-  [#21753](https://github.com/google-gemini/gemini-cli/pull/21753)
-- perf(cli): cache loadSettings to reduce redundant disk I/O at startup by
-  @sehoon38 in [#21521](https://github.com/google-gemini/gemini-cli/pull/21521)
-- fix(core): resolve Windows line ending and path separation bugs across CLI by
-  @muhammadusman586 in
-  [#21068](https://github.com/google-gemini/gemini-cli/pull/21068)
-- docs: fix heading formatting in commands.md and phrasing in tools-api.md by
-  @campox747 in [#20679](https://github.com/google-gemini/gemini-cli/pull/20679)
-- refactor(ui): unify keybinding infrastructure and support string
-  initialization by @scidomino in
-  [#21776](https://github.com/google-gemini/gemini-cli/pull/21776)
-- Add support for updating extension sources and names by @chrstnb in
-  [#21715](https://github.com/google-gemini/gemini-cli/pull/21715)
-- fix(core): handle GUI editor non-zero exit codes gracefully by @reyyanxahmed
-  in [#20376](https://github.com/google-gemini/gemini-cli/pull/20376)
-- fix(core): destroy PTY on kill() and exception to prevent fd leak by @nbardy
-  in [#21693](https://github.com/google-gemini/gemini-cli/pull/21693)
-- fix(docs): update theme screenshots and add missing themes by @ashmod in
-  [#20689](https://github.com/google-gemini/gemini-cli/pull/20689)
-- refactor(cli): rename 'return' key to 'enter' internally by @scidomino in
-  [#21796](https://github.com/google-gemini/gemini-cli/pull/21796)
-- build(release): restrict npm bundling to non-stable tags by @sehoon38 in
-  [#21821](https://github.com/google-gemini/gemini-cli/pull/21821)
-- fix(core): override toolRegistry property for sub-agent schedulers by
-  @gsquared94 in
-  [#21766](https://github.com/google-gemini/gemini-cli/pull/21766)
-- fix(cli): make footer items equally spaced by @jacob314 in
-  [#21843](https://github.com/google-gemini/gemini-cli/pull/21843)
-- docs: clarify global policy rules application in plan mode by @jerop in
-  [#21864](https://github.com/google-gemini/gemini-cli/pull/21864)
-- fix(core): ensure correct flash model steering in plan mode implementation
-  phase by @jerop in
-  [#21871](https://github.com/google-gemini/gemini-cli/pull/21871)
-- fix(core): update @a2a-js/sdk to 0.3.11 by @adamfweidman in
-  [#21875](https://github.com/google-gemini/gemini-cli/pull/21875)
-- refactor(core): improve API response error logging when retry by @yunaseoul in
-  [#21784](https://github.com/google-gemini/gemini-cli/pull/21784)
-- fix(ui): handle headless execution in credits and upgrade dialogs by
-  @gsquared94 in
-  [#21850](https://github.com/google-gemini/gemini-cli/pull/21850)
-- fix(core): treat retryable errors with >5 min delay as terminal quota errors
-  by @gsquared94 in
-  [#21881](https://github.com/google-gemini/gemini-cli/pull/21881)
-- feat(telemetry): add specific PR, issue, and custom tracking IDs for GitHub
-  Actions by @cocosheng-g in
-  [#21129](https://github.com/google-gemini/gemini-cli/pull/21129)
-- feat(core): add OAuth2 Authorization Code auth provider for A2A agents by
-  @SandyTao520 in
-  [#21496](https://github.com/google-gemini/gemini-cli/pull/21496)
-- feat(cli): give visibility to /tools list command in the TUI and follow the
-  subcommand pattern of other commands by @JayadityaGit in
-  [#21213](https://github.com/google-gemini/gemini-cli/pull/21213)
-- Handle dirty worktrees better and warn about running scripts/review.sh on
-  untrusted code. by @jacob314 in
-  [#21791](https://github.com/google-gemini/gemini-cli/pull/21791)
-- feat(policy): support auto-add to policy by default and scoped persistence by
+  [#21878](https://github.com/google-gemini/gemini-cli/pull/21878)
+- fix: clean up /clear and /resume by @jackwotherspoon in
+  [#22007](https://github.com/google-gemini/gemini-cli/pull/22007)
+- fix(core)#20941: reap orphaned descendant processes on PTY abort by @manavmax
+  in [#21124](https://github.com/google-gemini/gemini-cli/pull/21124)
+- fix(core): update language detection to use LSP 3.18 identifiers by @yunaseoul
+  in [#21931](https://github.com/google-gemini/gemini-cli/pull/21931)
+- feat(cli): support removing keybindings via '-' prefix by @scidomino in
+  [#22042](https://github.com/google-gemini/gemini-cli/pull/22042)
+- feat(policy): add --admin-policy flag for supplemental admin policies by
+  @galz10 in [#20360](https://github.com/google-gemini/gemini-cli/pull/20360)
+- merge duplicate imports packages/cli/src subtask1 by @Nixxx19 in
+  [#22040](https://github.com/google-gemini/gemini-cli/pull/22040)
+- perf(core): parallelize user quota and experiments fetching in refreshAuth by
+  @sehoon38 in [#21648](https://github.com/google-gemini/gemini-cli/pull/21648)
+- Changelog for v0.34.0-preview.0 by @gemini-cli-robot in
+  [#21965](https://github.com/google-gemini/gemini-cli/pull/21965)
+- Changelog for v0.33.0 by @gemini-cli-robot in
+  [#21967](https://github.com/google-gemini/gemini-cli/pull/21967)
+- fix(core): handle EISDIR in robustRealpath on Windows by @sehoon38 in
+  [#21984](https://github.com/google-gemini/gemini-cli/pull/21984)
+- feat(core): include initiationMethod in conversation interaction telemetry by
+  @yunaseoul in [#22054](https://github.com/google-gemini/gemini-cli/pull/22054)
+- feat(ui): add vim yank/paste (y/p/P) with unnamed register by @aanari in
+  [#22026](https://github.com/google-gemini/gemini-cli/pull/22026)
+- fix(core): enable numerical routing for api key users by @sehoon38 in
+  [#21977](https://github.com/google-gemini/gemini-cli/pull/21977)
+- feat(telemetry): implement retry attempt telemetry for network related retries
+  by @aishaneeshah in
+  [#22027](https://github.com/google-gemini/gemini-cli/pull/22027)
+- fix(policy): remove unnecessary escapeRegex from pattern builders by
   @spencer426 in
-  [#20361](https://github.com/google-gemini/gemini-cli/pull/20361)
-- fix(core): handle AbortError when ESC cancels tool execution by @PrasannaPal21
-  in [#20863](https://github.com/google-gemini/gemini-cli/pull/20863)
-- fix(release): Improve Patch Release Workflow Comments: Clearer Approval
-  Guidance by @jerop in
-  [#21894](https://github.com/google-gemini/gemini-cli/pull/21894)
-- docs: clarify telemetry setup and comprehensive data map by @jerop in
-  [#21879](https://github.com/google-gemini/gemini-cli/pull/21879)
-- feat(core): add per-model token usage to stream-json output by @yongruilin in
-  [#21839](https://github.com/google-gemini/gemini-cli/pull/21839)
-- docs: remove experimental badge from plan mode in sidebar by @jerop in
-  [#21906](https://github.com/google-gemini/gemini-cli/pull/21906)
-- fix(cli): prevent race condition in loop detection retry by @skyvanguard in
-  [#17916](https://github.com/google-gemini/gemini-cli/pull/17916)
-- Add behavioral evals for tracker by @anj-s in
-  [#20069](https://github.com/google-gemini/gemini-cli/pull/20069)
-- fix(auth): update terminology to 'sign in' and 'sign out' by @clocky in
-  [#20892](https://github.com/google-gemini/gemini-cli/pull/20892)
-- docs(mcp): standardize mcp tool fqn documentation by @abhipatel12 in
-  [#21664](https://github.com/google-gemini/gemini-cli/pull/21664)
-- fix(ui): prevent empty tool-group border stubs after filtering by @Aaxhirrr in
-  [#21852](https://github.com/google-gemini/gemini-cli/pull/21852)
-- make command names consistent by @scidomino in
-  [#21907](https://github.com/google-gemini/gemini-cli/pull/21907)
-- refactor: remove agent_card_requires_auth config flag by @adamfweidman in
-  [#21914](https://github.com/google-gemini/gemini-cli/pull/21914)
-- feat(a2a): implement standardized normalization and streaming reassembly by
-  @alisa-alisa in
-  [#21402](https://github.com/google-gemini/gemini-cli/pull/21402)
-- feat(cli): enable skill activation via slash commands by @NTaylorMullen in
-  [#21758](https://github.com/google-gemini/gemini-cli/pull/21758)
-- docs(cli): mention per-model token usage in stream-json result event by
-  @yongruilin in
-  [#21908](https://github.com/google-gemini/gemini-cli/pull/21908)
-- fix(plan): prevent plan truncation in approval dialog by supporting
-  unconstrained heights by @Adib234 in
-  [#21037](https://github.com/google-gemini/gemini-cli/pull/21037)
-- feat(a2a): switch from callback-based to event-driven tool scheduler by
-  @cocosheng-g in
-  [#21467](https://github.com/google-gemini/gemini-cli/pull/21467)
-- feat(voice): implement speech-friendly response formatter by @Solventerritory
-  in [#20989](https://github.com/google-gemini/gemini-cli/pull/20989)
-- feat: add pulsating blue border automation overlay to browser agent by
-  @kunal-10-cloud in
-  [#21173](https://github.com/google-gemini/gemini-cli/pull/21173)
-- Add extensionRegistryURI setting to change where the registry is read from by
-  @kevinjwang1 in
-  [#20463](https://github.com/google-gemini/gemini-cli/pull/20463)
-- fix: patch gaxios v7 Array.toString() stream corruption by @gsquared94 in
-  [#21884](https://github.com/google-gemini/gemini-cli/pull/21884)
-- fix: prevent hangs in non-interactive mode and improve agent guidance by
-  @cocosheng-g in
-  [#20893](https://github.com/google-gemini/gemini-cli/pull/20893)
-- Add ExtensionDetails dialog and support install by @chrstnb in
-  [#20845](https://github.com/google-gemini/gemini-cli/pull/20845)
-- chore/release: bump version to 0.34.0-nightly.20260310.4653b126f by
-  @gemini-cli-robot in
-  [#21816](https://github.com/google-gemini/gemini-cli/pull/21816)
-- Changelog for v0.33.0-preview.13 by @gemini-cli-robot in
-  [#21927](https://github.com/google-gemini/gemini-cli/pull/21927)
-- fix(cli): stabilize prompt layout to prevent jumping when typing by
+  [#21921](https://github.com/google-gemini/gemini-cli/pull/21921)
+- fix(core): preserve dynamic tool descriptions on session resume by @sehoon38
+  in [#18835](https://github.com/google-gemini/gemini-cli/pull/18835)
+- chore: allow 'gemini-3.1' in sensitive keyword linter by @scidomino in
+  [#22065](https://github.com/google-gemini/gemini-cli/pull/22065)
+- feat(core): support custom base URL via env vars by @junaiddshaukat in
+  [#21561](https://github.com/google-gemini/gemini-cli/pull/21561)
+- merge duplicate imports packages/cli/src subtask2 by @Nixxx19 in
+  [#22051](https://github.com/google-gemini/gemini-cli/pull/22051)
+- fix(core): silently retry API errors up to 3 times before halting session by
+  @spencer426 in
+  [#21989](https://github.com/google-gemini/gemini-cli/pull/21989)
+- feat(core): simplify subagent success UI and improve early termination display
+  by @abhipatel12 in
+  [#21917](https://github.com/google-gemini/gemini-cli/pull/21917)
+- merge duplicate imports packages/cli/src subtask3 by @Nixxx19 in
+  [#22056](https://github.com/google-gemini/gemini-cli/pull/22056)
+- fix(hooks): fix BeforeAgent/AfterAgent inconsistencies (#18514) by @krishdef7
+  in [#21383](https://github.com/google-gemini/gemini-cli/pull/21383)
+- feat(core): implement SandboxManager interface and config schema by @galz10 in
+  [#21774](https://github.com/google-gemini/gemini-cli/pull/21774)
+- docs: document npm deprecation warnings as safe to ignore by @h30s in
+  [#20692](https://github.com/google-gemini/gemini-cli/pull/20692)
+- fix: remove status/need-triage from maintainer-only issues by @SandyTao520 in
+  [#22044](https://github.com/google-gemini/gemini-cli/pull/22044)
+- fix(core): propagate subagent context to policy engine by @NTaylorMullen in
+  [#22086](https://github.com/google-gemini/gemini-cli/pull/22086)
+- fix(cli): resolve skill uninstall failure when skill name is updated by
   @NTaylorMullen in
-  [#21081](https://github.com/google-gemini/gemini-cli/pull/21081)
-- fix: preserve prompt text when cancelling streaming by @Nixxx19 in
-  [#21103](https://github.com/google-gemini/gemini-cli/pull/21103)
-- fix: robust UX for remote agent errors by @Shyam-Raghuwanshi in
-  [#20307](https://github.com/google-gemini/gemini-cli/pull/20307)
-- feat: implement background process logging and cleanup by @galz10 in
-  [#21189](https://github.com/google-gemini/gemini-cli/pull/21189)
-- Changelog for v0.33.0-preview.14 by @gemini-cli-robot in
-  [#21938](https://github.com/google-gemini/gemini-cli/pull/21938)
+  [#22085](https://github.com/google-gemini/gemini-cli/pull/22085)
+- docs(plan): clarify interactive plan editing with Ctrl+X by @Adib234 in
+  [#22076](https://github.com/google-gemini/gemini-cli/pull/22076)
+- fix(policy): ensure user policies are loaded when policyPaths is empty by
+  @NTaylorMullen in
+  [#22090](https://github.com/google-gemini/gemini-cli/pull/22090)
+- Docs: Add documentation for model steering (experimental). by @jkcinouye in
+  [#21154](https://github.com/google-gemini/gemini-cli/pull/21154)
+- Add issue for automated changelogs by @g-samroberts in
+  [#21912](https://github.com/google-gemini/gemini-cli/pull/21912)
+- fix(core): secure argsPattern and revert WEB_FETCH_TOOL_NAME escalation by
+  @spencer426 in
+  [#22104](https://github.com/google-gemini/gemini-cli/pull/22104)
+- feat(core): differentiate User-Agent for a2a-server and ACP clients by
+  @bdmorgan in [#22059](https://github.com/google-gemini/gemini-cli/pull/22059)
+- refactor(core): extract ExecutionLifecycleService for tool backgrounding by
+  @adamfweidman in
+  [#21717](https://github.com/google-gemini/gemini-cli/pull/21717)
+- feat: Display pending and confirming tool calls by @sripasg in
+  [#22106](https://github.com/google-gemini/gemini-cli/pull/22106)
+- feat(browser): implement input blocker overlay during automation by
+  @kunal-10-cloud in
+  [#21132](https://github.com/google-gemini/gemini-cli/pull/21132)
+- fix: register themes on extension load not start by @jackwotherspoon in
+  [#22148](https://github.com/google-gemini/gemini-cli/pull/22148)
+- feat(ui): Do not show Ultra users /upgrade hint (#22154) by @sehoon38 in
+  [#22156](https://github.com/google-gemini/gemini-cli/pull/22156)
+- chore: remove unnecessary log for themes by @jackwotherspoon in
+  [#22165](https://github.com/google-gemini/gemini-cli/pull/22165)
+- fix(core): resolve MCP tool FQN validation, schema export, and wildcards in
+  subagents by @abhipatel12 in
+  [#22069](https://github.com/google-gemini/gemini-cli/pull/22069)
+- fix(cli): validate --model argument at startup by @JaisalJain in
+  [#21393](https://github.com/google-gemini/gemini-cli/pull/21393)
+- fix(core): handle policy ALLOW for exit_plan_mode by @backnotprop in
+  [#21802](https://github.com/google-gemini/gemini-cli/pull/21802)
+- feat(telemetry): add Clearcut instrumentation for AI credits billing events by
+  @gsquared94 in
+  [#22153](https://github.com/google-gemini/gemini-cli/pull/22153)
+- feat(core): add google credentials provider for remote agents by @adamfweidman
+  in [#21024](https://github.com/google-gemini/gemini-cli/pull/21024)
+- test(cli): add integration test for node deprecation warnings by @Nixxx19 in
+  [#20215](https://github.com/google-gemini/gemini-cli/pull/20215)
+- feat(cli): allow safe tools to execute concurrently while agent is busy by
+  @spencer426 in
+  [#21988](https://github.com/google-gemini/gemini-cli/pull/21988)
+- feat(core): implement model-driven parallel tool scheduler by @abhipatel12 in
+  [#21933](https://github.com/google-gemini/gemini-cli/pull/21933)
+- update vulnerable deps by @scidomino in
+  [#22180](https://github.com/google-gemini/gemini-cli/pull/22180)
+- fix(core): fix startup stats to use int values for timestamps and durations by
+  @yunaseoul in [#22201](https://github.com/google-gemini/gemini-cli/pull/22201)
+- fix(core): prevent duplicate tool schemas for instantiated tools by
+  @abhipatel12 in
+  [#22204](https://github.com/google-gemini/gemini-cli/pull/22204)
+- fix(core): add proxy routing support for remote A2A subagents by @adamfweidman
+  in [#22199](https://github.com/google-gemini/gemini-cli/pull/22199)
+- fix(core/ide): add Antigravity CLI fallbacks by @apfine in
+  [#22030](https://github.com/google-gemini/gemini-cli/pull/22030)
+- fix(browser): fix duplicate function declaration error in browser agent by
+  @gsquared94 in
+  [#22207](https://github.com/google-gemini/gemini-cli/pull/22207)
+- feat(core): implement Stage 1 improvements for webfetch tool by @aishaneeshah
+  in [#21313](https://github.com/google-gemini/gemini-cli/pull/21313)
+- Changelog for v0.34.0-preview.1 by @gemini-cli-robot in
+  [#22194](https://github.com/google-gemini/gemini-cli/pull/22194)
+- perf(cli): enable code splitting and deferred UI loading by @sehoon38 in
+  [#22117](https://github.com/google-gemini/gemini-cli/pull/22117)
+- fix: remove unused img.png from project root by @SandyTao520 in
+  [#22222](https://github.com/google-gemini/gemini-cli/pull/22222)
+- docs(local model routing): add docs on how to use Gemma for local model
+  routing by @douglas-reid in
+  [#21365](https://github.com/google-gemini/gemini-cli/pull/21365)
+- feat(a2a): enable native gRPC support and protocol routing by @alisa-alisa in
+  [#21403](https://github.com/google-gemini/gemini-cli/pull/21403)
+- fix(cli): escape @ symbols on paste to prevent unintended file expansion by
+  @krishdef7 in [#21239](https://github.com/google-gemini/gemini-cli/pull/21239)
+- feat(core): add trajectoryId to ConversationOffered telemetry by @yunaseoul in
+  [#22214](https://github.com/google-gemini/gemini-cli/pull/22214)
+- docs: clarify that tools.core is an allowlist for ALL built-in tools by
+  @hobostay in [#18813](https://github.com/google-gemini/gemini-cli/pull/18813)
+- docs(plan): document hooks with plan mode by @ruomengz in
+  [#22197](https://github.com/google-gemini/gemini-cli/pull/22197)
+- Changelog for v0.33.1 by @gemini-cli-robot in
+  [#22235](https://github.com/google-gemini/gemini-cli/pull/22235)
+- build(ci): fix false positive evals trigger on merge commits by @gundermanc in
+  [#22237](https://github.com/google-gemini/gemini-cli/pull/22237)
+- fix(core): explicitly pass messageBus to policy engine for MCP tool saves by
+  @abhipatel12 in
+  [#22255](https://github.com/google-gemini/gemini-cli/pull/22255)
+- feat(core): Fully migrate packages/core to AgentLoopContext. by @joshualitt in
+  [#22115](https://github.com/google-gemini/gemini-cli/pull/22115)
+- feat(core): increase sub-agent turn and time limits by @bdmorgan in
+  [#22196](https://github.com/google-gemini/gemini-cli/pull/22196)
+- feat(core): instrument file system tools for JIT context discovery by
+  @SandyTao520 in
+  [#22082](https://github.com/google-gemini/gemini-cli/pull/22082)
+- refactor(ui): extract pure session browser utilities by @abhipatel12 in
+  [#22256](https://github.com/google-gemini/gemini-cli/pull/22256)
+- fix(plan): Fix AskUser evals by @Adib234 in
+  [#22074](https://github.com/google-gemini/gemini-cli/pull/22074)
+- fix(settings): prevent j/k navigation keys from intercepting edit buffer input
+  by @student-ankitpandit in
+  [#21865](https://github.com/google-gemini/gemini-cli/pull/21865)
+- feat(skills): improve async-pr-review workflow and logging by @mattKorwel in
+  [#21790](https://github.com/google-gemini/gemini-cli/pull/21790)
+- refactor(cli): consolidate getErrorMessage utility to core by @scidomino in
+  [#22190](https://github.com/google-gemini/gemini-cli/pull/22190)
+- fix(core): show descriptive error messages when saving settings fails by
+  @afarber in [#18095](https://github.com/google-gemini/gemini-cli/pull/18095)
+- docs(core): add authentication guide for remote subagents by @adamfweidman in
+  [#22178](https://github.com/google-gemini/gemini-cli/pull/22178)
+- docs: overhaul subagents documentation and add /agents command by @abhipatel12
+  in [#22345](https://github.com/google-gemini/gemini-cli/pull/22345)
+- refactor(ui): extract SessionBrowser static ui components by @abhipatel12 in
+  [#22348](https://github.com/google-gemini/gemini-cli/pull/22348)
+- test: add Object.create context regression test and tool confirmation
+  integration test by @gsquared94 in
+  [#22356](https://github.com/google-gemini/gemini-cli/pull/22356)
+- feat(tracker): return TodoList display for tracker tools by @anj-s in
+  [#22060](https://github.com/google-gemini/gemini-cli/pull/22060)
+- feat(agent): add allowed domain restrictions for browser agent by
+  @cynthialong0-0 in
+  [#21775](https://github.com/google-gemini/gemini-cli/pull/21775)
+- chore/release: bump version to 0.35.0-nightly.20260313.bb060d7a9 by
+  @gemini-cli-robot in
+  [#22251](https://github.com/google-gemini/gemini-cli/pull/22251)
+- Move keychain fallback to keychain service by @chrstnb in
+  [#22332](https://github.com/google-gemini/gemini-cli/pull/22332)
+- feat(core): integrate SandboxManager to sandbox all process-spawning tools by
+  @galz10 in [#22231](https://github.com/google-gemini/gemini-cli/pull/22231)
+- fix(cli): support CJK input and full Unicode scalar values in terminal
+  protocols by @scidomino in
+  [#22353](https://github.com/google-gemini/gemini-cli/pull/22353)
+- Promote stable tests. by @gundermanc in
+  [#22253](https://github.com/google-gemini/gemini-cli/pull/22253)
+- feat(tracker): add tracker policy by @anj-s in
+  [#22379](https://github.com/google-gemini/gemini-cli/pull/22379)
+- feat(security): add disableAlwaysAllow setting to disable auto-approvals by
+  @galz10 in [#21941](https://github.com/google-gemini/gemini-cli/pull/21941)
+- Revert "fix(cli): validate --model argument at startup" by @sehoon38 in
+  [#22378](https://github.com/google-gemini/gemini-cli/pull/22378)
+- fix(mcp): handle equivalent root resource URLs in OAuth validation by @galz10
+  in [#20231](https://github.com/google-gemini/gemini-cli/pull/20231)
+- fix(core): use session-specific temp directory for task tracker by @anj-s in
+  [#22382](https://github.com/google-gemini/gemini-cli/pull/22382)
+- Fix issue where config was undefined. by @gundermanc in
+  [#22397](https://github.com/google-gemini/gemini-cli/pull/22397)
+- fix(core): deduplicate project memory when JIT context is enabled by
+  @SandyTao520 in
+  [#22234](https://github.com/google-gemini/gemini-cli/pull/22234)
+- feat(prompts): implement Topic-Action-Summary model for verbosity reduction by
+  @Abhijit-2592 in
+  [#21503](https://github.com/google-gemini/gemini-cli/pull/21503)
+- fix(core): fix manual deletion of subagent histories by @abhipatel12 in
+  [#22407](https://github.com/google-gemini/gemini-cli/pull/22407)
+- Add registry var by @kevinjwang1 in
+  [#22224](https://github.com/google-gemini/gemini-cli/pull/22224)
+- Add ModelDefinitions to ModelConfigService by @kevinjwang1 in
+  [#22302](https://github.com/google-gemini/gemini-cli/pull/22302)
+- fix(cli): improve command conflict handling for skills by @NTaylorMullen in
+  [#21942](https://github.com/google-gemini/gemini-cli/pull/21942)
+- fix(core): merge user settings with extension-provided MCP servers by
+  @abhipatel12 in
+  [#22484](https://github.com/google-gemini/gemini-cli/pull/22484)
+- fix(core): skip discovery for incomplete MCP configs and resolve merge race
+  condition by @abhipatel12 in
+  [#22494](https://github.com/google-gemini/gemini-cli/pull/22494)
+- fix(automation): harden stale PR closer permissions and maintainer detection
+  by @bdmorgan in
+  [#22558](https://github.com/google-gemini/gemini-cli/pull/22558)
+- fix(automation): evaluate staleness before checking protected labels by
+  @bdmorgan in [#22561](https://github.com/google-gemini/gemini-cli/pull/22561)
+- feat(agent): replace the runtime npx for browser agent chrome devtool mcp with
+  pre-built bundle by @cynthialong0-0 in
+  [#22213](https://github.com/google-gemini/gemini-cli/pull/22213)
+- perf: optimize TrackerService dependency checks by @anj-s in
+  [#22384](https://github.com/google-gemini/gemini-cli/pull/22384)
+- docs(policy): remove trailing space from commandPrefix examples by @kawasin73
+  in [#22264](https://github.com/google-gemini/gemini-cli/pull/22264)
+- fix(a2a-server): resolve unsafe assignment lint errors by @ehedlund in
+  [#22661](https://github.com/google-gemini/gemini-cli/pull/22661)
+- fix: Adjust ToolGroupMessage filtering to hide Confirming and show Canceled
+  tool calls. by @sripasg in
+  [#22230](https://github.com/google-gemini/gemini-cli/pull/22230)
+- Disallow Object.create() and reflect. by @gundermanc in
+  [#22408](https://github.com/google-gemini/gemini-cli/pull/22408)
+- Guard pro model usage by @sehoon38 in
+  [#22665](https://github.com/google-gemini/gemini-cli/pull/22665)
+- refactor(core): Creates AgentSession abstraction for consolidated agent
+  interface. by @mbleigh in
+  [#22270](https://github.com/google-gemini/gemini-cli/pull/22270)
+- docs(changelog): remove internal commands from release notes by
+  @jackwotherspoon in
+  [#22529](https://github.com/google-gemini/gemini-cli/pull/22529)
+- feat: enable subagents by @abhipatel12 in
+  [#22386](https://github.com/google-gemini/gemini-cli/pull/22386)
+- feat(extensions): implement cryptographic integrity verification for extension
+  updates by @ehedlund in
+  [#21772](https://github.com/google-gemini/gemini-cli/pull/21772)
+- feat(tracker): polish UI sorting and formatting by @anj-s in
+  [#22437](https://github.com/google-gemini/gemini-cli/pull/22437)
+- Changelog for v0.34.0-preview.2 by @gemini-cli-robot in
+  [#22220](https://github.com/google-gemini/gemini-cli/pull/22220)
+- fix(core): fix three JIT context bugs in read_file, read_many_files, and
+  memoryDiscovery by @SandyTao520 in
+  [#22679](https://github.com/google-gemini/gemini-cli/pull/22679)
+- refactor(core): introduce InjectionService with source-aware injection and
+  backend-native background completions by @adamfweidman in
+  [#22544](https://github.com/google-gemini/gemini-cli/pull/22544)
+- Linux sandbox bubblewrap by @DavidAPierce in
+  [#22680](https://github.com/google-gemini/gemini-cli/pull/22680)
+- feat(core): increase thought signature retry resilience by @bdmorgan in
+  [#22202](https://github.com/google-gemini/gemini-cli/pull/22202)
+- feat(core): implement Stage 2 security and consistency improvements for
+  web_fetch by @aishaneeshah in
+  [#22217](https://github.com/google-gemini/gemini-cli/pull/22217)
+- refactor(core): replace positional execute params with ExecuteOptions bag by
+  @adamfweidman in
+  [#22674](https://github.com/google-gemini/gemini-cli/pull/22674)
+- feat(config): enable JIT context loading by default by @SandyTao520 in
+  [#22736](https://github.com/google-gemini/gemini-cli/pull/22736)
+- fix(config): ensure discoveryMaxDirs is passed to global config during
+  initialization by @kevin-ramdass in
+  [#22744](https://github.com/google-gemini/gemini-cli/pull/22744)
+- fix(plan): allowlist get_internal_docs in Plan Mode by @Adib234 in
+  [#22668](https://github.com/google-gemini/gemini-cli/pull/22668)
+- Changelog for v0.34.0-preview.3 by @gemini-cli-robot in
+  [#22393](https://github.com/google-gemini/gemini-cli/pull/22393)
+- feat(core): add foundation for subagent tool isolation by @akh64bit in
+  [#22708](https://github.com/google-gemini/gemini-cli/pull/22708)
+- fix(core): handle surrogate pairs in truncateString by @sehoon38 in
+  [#22754](https://github.com/google-gemini/gemini-cli/pull/22754)
+- fix(cli): override j/k navigation in settings dialog to fix search input
+  conflict by @sehoon38 in
+  [#22800](https://github.com/google-gemini/gemini-cli/pull/22800)
+- feat(plan): add 'All the above' option to multi-select AskUser questions by
+  @Adib234 in [#22365](https://github.com/google-gemini/gemini-cli/pull/22365)
+- docs: distribute package-specific GEMINI.md context to each package by
+  @SandyTao520 in
+  [#22734](https://github.com/google-gemini/gemini-cli/pull/22734)
+- fix(cli): clean up stale pasted placeholder metadata after word/line deletions
+  by @Jomak-x in
+  [#20375](https://github.com/google-gemini/gemini-cli/pull/20375)
+- refactor(core): align JIT memory placement with tiered context model by
+  @SandyTao520 in
+  [#22766](https://github.com/google-gemini/gemini-cli/pull/22766)
+- Linux sandbox seccomp by @DavidAPierce in
+  [#22815](https://github.com/google-gemini/gemini-cli/pull/22815)
 
 **Full Changelog**:
-https://github.com/google-gemini/gemini-cli/compare/v0.33.0-preview.15...v0.34.0-preview.4
+https://github.com/google-gemini/gemini-cli/compare/v0.34.0-preview.4...v0.35.0-preview.1

docs/cli/plan-mode.md

@@ -460,6 +460,26 @@ Manual deletion also removes all associated artifacts:
 If you use a [custom plans directory](#custom-plan-directory-and-policies),
 those files are not automatically deleted and must be managed manually.
 
+## Non-interactive execution
+
+When running Gemini CLI in non-interactive environments (such as headless
+scripts or CI/CD pipelines), Plan Mode optimizes for automated workflows:
+
+- **Automatic transitions:** The policy engine automatically approves the
+  `enter_plan_mode` and `exit_plan_mode` tools without prompting for user
+  confirmation.
+- **Automated implementation:** When exiting Plan Mode to execute the plan,
+  Gemini CLI automatically switches to
+  [YOLO mode](../reference/policy-engine.md#approval-modes) instead of the
+  standard Default mode. This allows the CLI to execute the implementation steps
+  automatically without hanging on interactive tool approvals.
+
+**Example:**
+
+```bash
+gemini --approval-mode plan -p "Analyze telemetry and suggest improvements"
+```
+
 [`plan.toml`]:
   https://github.com/google-gemini/gemini-cli/blob/main/packages/core/src/policy/policies/plan.toml
 [Conductor]: https://github.com/gemini-cli-extensions/conductor

docs/reference/policy-engine.md

@@ -90,6 +90,17 @@ If `argsPattern` is specified, the tool's arguments are converted to a stable
 JSON string, which is then tested against the provided regular expression. If
 the arguments don't match the pattern, the rule does not apply.
 
+#### Execution environment
+
+If `interactive` is specified, the rule will only apply if the CLI's execution
+environment matches the specified boolean value:
+
+- `true`: The rule applies only in interactive mode.
+- `false`: The rule applies only in non-interactive (headless) mode.
+
+If omitted, the rule applies to both interactive and non-interactive
+environments.
+
 ### Decisions
 
 There are three possible decisions a rule can enforce:
@@ -286,6 +297,10 @@ deny_message = "Deletion is permanent"
 
 # (Optional) An array of approval modes where this rule is active.
 modes = ["autoEdit"]
+
+# (Optional) A boolean to restrict the rule to interactive (true) or non-interactive (false) environments.
+# If omitted, the rule applies to both.
+interactive = true
 ```
 
 ### Using arrays (lists)
@@ -360,6 +375,8 @@ priority = 200
 Specify only the `mcpName` to apply a rule to every tool provided by that
 server.
 
+**Note:** This applies to all decision types (`allow`, `deny`, `ask_user`).
+
 ```toml
 # Denies all tools from the `untrusted-server` MCP
 [[rule]]

docs/resources/faq.md

@@ -58,6 +58,19 @@ your total token usage using the `/stats` command in Gemini CLI.
 
 ## Installation and updates
 
+### How do I check which version of Gemini CLI I'm currently running?
+
+You can check your current Gemini CLI version using one of these methods:
+
+- Run `gemini --version` or `gemini -v` from your terminal
+- Check the globally installed version using your package manager:
+  - npm: `npm list -g @google/gemini-cli`
+  - pnpm: `pnpm list -g @google/gemini-cli`
+  - yarn: `yarn global list @google/gemini-cli`
+  - bun: `bun pm ls -g @google/gemini-cli`
+  - homebrew: `brew list --versions gemini-cli`
+- Inside an active Gemini CLI session, use the `/about` command
+
 ### How do I update Gemini CLI to the latest version?
 
 If you installed it globally via `npm`, update it using the command

evals/plan_mode.eval.ts

@@ -18,6 +18,18 @@ describe('plan_mode', () => {
     experimental: { plan: true },
   };
 
+  const getWriteTargets = (logs: any[]) =>
+    logs
+      .filter((log) => ['write_file', 'replace'].includes(log.toolRequest.name))
+      .map((log) => {
+        try {
+          return JSON.parse(log.toolRequest.args).file_path as string;
+        } catch {
+          return '';
+        }
+      })
+      .filter(Boolean);
+
   evalTest('ALWAYS_PASSES', {
     name: 'should refuse file modification when in plan mode',
     approvalMode: ApprovalMode.PLAN,
@@ -32,27 +44,23 @@ describe('plan_mode', () => {
       await rig.waitForTelemetryReady();
       const toolLogs = rig.readToolLogs();
 
-      const writeTargets = toolLogs
-        .filter((log) =>
-          ['write_file', 'replace'].includes(log.toolRequest.name),
-        )
-        .map((log) => {
-          try {
-            return JSON.parse(log.toolRequest.args).file_path;
-          } catch {
-            return null;
-          }
-        });
+      const exitPlanIndex = toolLogs.findIndex(
+        (log) => log.toolRequest.name === 'exit_plan_mode',
+      );
+
+      const writeTargetsBeforeExitPlan = getWriteTargets(
+        toolLogs.slice(0, exitPlanIndex !== -1 ? exitPlanIndex : undefined),
+      );
 
       expect(
-        writeTargets,
+        writeTargetsBeforeExitPlan,
         'Should not attempt to modify README.md in plan mode',
       ).not.toContain('README.md');
 
       assertModelHasOutput(result);
       checkModelOutputContent(result, {
         expectedContent: [/plan mode|read-only|cannot modify|refuse|exiting/i],
-        testName: `${TEST_PREFIX}should refuse file modification`,
+        testName: `${TEST_PREFIX}should refuse file modification in plan mode`,
       });
     },
   });
@@ -69,24 +77,20 @@ describe('plan_mode', () => {
       await rig.waitForTelemetryReady();
       const toolLogs = rig.readToolLogs();
 
-      const writeTargets = toolLogs
-        .filter((log) =>
-          ['write_file', 'replace'].includes(log.toolRequest.name),
-        )
-        .map((log) => {
-          try {
-            return JSON.parse(log.toolRequest.args).file_path;
-          } catch {
-            return null;
-          }
-        });
+      const exitPlanIndex = toolLogs.findIndex(
+        (log) => log.toolRequest.name === 'exit_plan_mode',
+      );
+
+      const writeTargetsBeforeExit = getWriteTargets(
+        toolLogs.slice(0, exitPlanIndex !== -1 ? exitPlanIndex : undefined),
+      );
 
       // It should NOT write to the docs folder or any other repo path
-      const hasRepoWrite = writeTargets.some(
+      const hasRepoWriteBeforeExit = writeTargetsBeforeExit.some(
         (path) => path && !path.includes('/plans/'),
       );
       expect(
-        hasRepoWrite,
+        hasRepoWriteBeforeExit,
         'Should not attempt to create files in the repository while in plan mode',
       ).toBe(false);
 
@@ -166,4 +170,65 @@ describe('plan_mode', () => {
       assertModelHasOutput(result);
     },
   });
+
+  evalTest('USUALLY_PASSES', {
+    name: 'should create a plan in plan mode and implement it for a refactoring task',
+    params: {
+      settings,
+    },
+    files: {
+      'src/mathUtils.ts':
+        'export const sum = (a: number, b: number) => a + b;\nexport const multiply = (a: number, b: number) => a * b;',
+      'src/main.ts':
+        'import { sum } from "./mathUtils";\nconsole.log(sum(1, 2));',
+    },
+    prompt:
+      'I want to refactor our math utilities. Move the `sum` function from `src/mathUtils.ts` to a new file `src/basicMath.ts` and update `src/main.ts` to use the new file. Please create a detailed implementation plan first, then execute it.',
+    assert: async (rig, result) => {
+      const enterPlanCalled = await rig.waitForToolCall('enter_plan_mode');
+      expect(
+        enterPlanCalled,
+        'Expected enter_plan_mode tool to be called',
+      ).toBe(true);
+
+      const exitPlanCalled = await rig.waitForToolCall('exit_plan_mode');
+      expect(exitPlanCalled, 'Expected exit_plan_mode tool to be called').toBe(
+        true,
+      );
+
+      await rig.waitForTelemetryReady();
+      const toolLogs = rig.readToolLogs();
+
+      // Check if plan was written
+      const planWrite = toolLogs.find(
+        (log) =>
+          log.toolRequest.name === 'write_file' &&
+          log.toolRequest.args.includes('/plans/'),
+      );
+      expect(
+        planWrite,
+        'Expected a plan file to be written in the plans directory',
+      ).toBeDefined();
+
+      // Check for implementation files
+      const newFileWrite = toolLogs.find(
+        (log) =>
+          log.toolRequest.name === 'write_file' &&
+          log.toolRequest.args.includes('src/basicMath.ts'),
+      );
+      expect(
+        newFileWrite,
+        'Expected src/basicMath.ts to be created',
+      ).toBeDefined();
+
+      const mainUpdate = toolLogs.find(
+        (log) =>
+          ['write_file', 'replace'].includes(log.toolRequest.name) &&
+          log.toolRequest.args.includes('src/main.ts'),
+      );
+      expect(mainUpdate, 'Expected src/main.ts to be updated').toBeDefined();
+
+      assertModelHasOutput(result);
+    },
+  });
 });

packages/cli/src/acp/acpClient.test.ts

@@ -894,6 +894,9 @@ describe('Session', () => {
         update: expect.objectContaining({
           sessionUpdate: 'tool_call_update',
           status: 'completed',
+          title: 'Test Tool',
+          locations: [],
+          kind: 'read',
         }),
       }),
     );
@@ -1306,6 +1309,18 @@ describe('Session', () => {
     expect(path.resolve).toHaveBeenCalled();
     expect(fs.stat).toHaveBeenCalled();
 
+    expect(mockConnection.sessionUpdate).toHaveBeenCalledWith(
+      expect.objectContaining({
+        update: expect.objectContaining({
+          sessionUpdate: 'tool_call_update',
+          status: 'completed',
+          title: 'Read files',
+          locations: [],
+          kind: 'read',
+        }),
+      }),
+    );
+
     // Verify ReadManyFilesTool was used (implicitly by checking if sendMessageStream was called with resolved content)
     // Since we mocked ReadManyFilesTool to return specific content, we can check the args passed to sendMessageStream
     expect(mockChat.sendMessageStream).toHaveBeenCalledWith(
@@ -1321,6 +1336,65 @@ describe('Session', () => {
     );
   });
 
+  it('should handle @path resolution error', async () => {
+    (path.resolve as unknown as Mock).mockReturnValue('/tmp/error.txt');
+    (fs.stat as unknown as Mock).mockResolvedValue({
+      isDirectory: () => false,
+    });
+    (isWithinRoot as unknown as Mock).mockReturnValue(true);
+
+    const MockReadManyFilesTool = ReadManyFilesTool as unknown as Mock;
+    MockReadManyFilesTool.mockImplementationOnce(() => ({
+      name: 'read_many_files',
+      kind: 'read',
+      build: vi.fn().mockReturnValue({
+        getDescription: () => 'Read files',
+        toolLocations: () => [],
+        execute: vi.fn().mockRejectedValue(new Error('File read failed')),
+      }),
+    }));
+
+    const stream = createMockStream([
+      {
+        type: StreamEventType.CHUNK,
+        value: { candidates: [] },
+      },
+    ]);
+    mockChat.sendMessageStream.mockResolvedValue(stream);
+
+    await expect(
+      session.prompt({
+        sessionId: 'session-1',
+        prompt: [
+          { type: 'text', text: 'Read' },
+          {
+            type: 'resource_link',
+            uri: 'file://error.txt',
+            mimeType: 'text/plain',
+            name: 'error.txt',
+          },
+        ],
+      }),
+    ).rejects.toThrow('File read failed');
+
+    expect(mockConnection.sessionUpdate).toHaveBeenCalledWith(
+      expect.objectContaining({
+        update: expect.objectContaining({
+          sessionUpdate: 'tool_call_update',
+          status: 'failed',
+          content: expect.arrayContaining([
+            expect.objectContaining({
+              content: expect.objectContaining({
+                text: expect.stringMatching(/File read failed/),
+              }),
+            }),
+          ]),
+          kind: 'read',
+        }),
+      }),
+    );
+  });
+
   it('should handle cancellation during prompt', async () => {
     let streamController: ReadableStreamDefaultController<unknown>;
     const stream = new ReadableStream({
@@ -1434,6 +1508,7 @@ describe('Session', () => {
               content: expect.objectContaining({ text: 'Tool failed' }),
             }),
           ]),
+          kind: 'read',
         }),
       }),
     );

packages/cli/src/acp/acpClient.ts

@@ -360,10 +360,7 @@ export class GeminiAgent {
       config.setFileSystemService(acpFileSystemService);
     }
 
-    const clientHistory = convertSessionToClientHistory(
-      sessionData.messages,
-      sessionData.lastCompressionIndex,
-    );
+    const clientHistory = convertSessionToClientHistory(sessionData.messages);
 
     const geminiClient = config.getGeminiClient();
     await geminiClient.initialize();
@@ -969,7 +966,10 @@ export class Session {
         sessionUpdate: 'tool_call_update',
         toolCallId: callId,
         status: 'completed',
+        title: invocation.getDescription(),
         content: content ? [content] : [],
+        locations: invocation.toolLocations(),
+        kind: toAcpToolKind(tool.kind),
       });
 
       const durationMs = Date.now() - startTime;
@@ -1033,6 +1033,7 @@ export class Session {
         content: [
           { type: 'content', content: { type: 'text', text: error.message } },
         ],
+        kind: toAcpToolKind(tool.kind),
       });
 
       this.chat.recordCompletedToolCalls(this.config.getActiveModel(), [
@@ -1327,7 +1328,10 @@ export class Session {
           sessionUpdate: 'tool_call_update',
           toolCallId: callId,
           status: 'completed',
+          title: invocation.getDescription(),
           content: content ? [content] : [],
+          locations: invocation.toolLocations(),
+          kind: toAcpToolKind(readManyFilesTool.kind),
         });
         if (Array.isArray(result.llmContent)) {
           const fileContentRegex = /^--- (.*?) ---\n\n([\s\S]*?)\n\n$/;
@@ -1371,6 +1375,7 @@ export class Session {
               },
             },
           ],
+          kind: toAcpToolKind(readManyFilesTool.kind),
         });
 
         throw error;

packages/cli/src/config/config.ts

@@ -97,7 +97,6 @@ export interface CliArgs {
   rawOutput: boolean | undefined;
   acceptRawOutputRisk: boolean | undefined;
   isCommand: boolean | undefined;
-  forever: boolean | undefined;
 }
 
 /**
@@ -299,12 +298,6 @@ export async function parseArguments(
         .option('accept-raw-output-risk', {
           type: 'boolean',
           description: 'Suppress the security warning when using --raw-output.',
-        })
-        .option('forever', {
-          type: 'boolean',
-          description:
-            'Run as a long-running autonomous agent with auto-resume and schedule_work support.',
-          default: false,
         }),
     )
     // Register MCP subcommands
@@ -900,7 +893,6 @@ export async function loadCliConfig(
       };
     },
     enableConseca: settings.security?.enableConseca,
-    isForeverMode: !!argv.forever,
   });
 }
 

packages/cli/src/config/settingsSchema.ts

@@ -2312,18 +2312,6 @@ const SETTINGS_SCHEMA = {
         ref: 'HookDefinitionArray',
         mergeStrategy: MergeStrategy.CONCAT,
       },
-      Idle: {
-        type: 'array',
-        label: 'Idle Hooks',
-        category: 'Advanced',
-        requiresRestart: false,
-        default: [],
-        description:
-          'Hooks that execute after a period of inactivity. Can trigger maintenance tasks like memory consolidation.',
-        showInDialog: false,
-        ref: 'HookDefinitionArray',
-        mergeStrategy: MergeStrategy.CONCAT,
-      },
     },
     additionalProperties: {
       type: 'array',

packages/cli/src/external-listener.ts

@@ -1,453 +0,0 @@
-/**
- * @license
- * Copyright 2026 Google LLC
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import http from 'node:http';
-import { writeFileSync, mkdirSync, unlinkSync } from 'node:fs';
-import { join } from 'node:path';
-import os from 'node:os';
-import crypto from 'node:crypto';
-import { appEvents, AppEvent } from './utils/events.js';
-
-// --- A2A Task management ---
-
-interface A2AResponseMessage {
-  kind: 'message';
-  role: 'agent';
-  parts: Array<{ kind: 'text'; text: string }>;
-  messageId: string;
-}
-
-interface A2ATask {
-  id: string;
-  contextId: string;
-  status: {
-    state: 'submitted' | 'working' | 'completed' | 'failed';
-    timestamp: string;
-    message?: A2AResponseMessage;
-  };
-}
-
-const tasks = new Map<string, A2ATask>();
-
-const TASK_CLEANUP_DELAY_MS = 10 * 60 * 1000; // 10 minutes
-const DEFAULT_BLOCKING_TIMEOUT_MS = 5 * 60 * 1000; // 5 minutes
-
-interface ResponseWaiter {
-  taskId: string;
-  resolve: (text: string) => void;
-}
-
-const responseWaiters: ResponseWaiter[] = [];
-
-// Queue for unsolicited responses (e.g. forever mode auto-resume output)
-const unsolicitedResponses: string[] = [];
-
-/**
- * Called by AppContainer when streaming transitions from non-Idle to Idle.
- * If there's a pending A2A task, resolves it. Otherwise queues as unsolicited.
- */
-export function notifyResponse(responseText: string): void {
-  if (!responseText) return;
-
-  const waiter = responseWaiters.shift();
-  if (!waiter) {
-    // No A2A task waiting — queue as unsolicited (forever mode, etc.)
-    unsolicitedResponses.push(responseText);
-    return;
-  }
-
-  const task = tasks.get(waiter.taskId);
-  if (task) {
-    task.status = {
-      state: 'completed',
-      timestamp: new Date().toISOString(),
-      message: {
-        kind: 'message',
-        role: 'agent',
-        parts: [{ kind: 'text', text: responseText }],
-        messageId: crypto.randomUUID(),
-      },
-    };
-    scheduleTaskCleanup(task.id);
-  }
-
-  waiter.resolve(responseText);
-}
-
-/**
- * Drain all unsolicited responses (from forever mode auto-resume, etc.).
- */
-export function drainUnsolicitedResponses(): string[] {
-  return unsolicitedResponses.splice(0, unsolicitedResponses.length);
-}
-
-/**
- * Returns true if there are any in-flight tasks waiting for a response.
- */
-export function hasPendingTasks(): boolean {
-  return responseWaiters.length > 0;
-}
-
-/**
- * Called when streaming starts (Idle -> non-Idle) to mark the oldest
- * submitted task as "working".
- */
-export function markTasksWorking(): void {
-  const waiter = responseWaiters[0];
-  if (!waiter) return;
-  const task = tasks.get(waiter.taskId);
-  if (task && task.status.state === 'submitted') {
-    task.status = {
-      state: 'working',
-      timestamp: new Date().toISOString(),
-    };
-  }
-}
-
-function scheduleTaskCleanup(taskId: string): void {
-  setTimeout(() => {
-    tasks.delete(taskId);
-  }, TASK_CLEANUP_DELAY_MS);
-}
-
-function createTask(): A2ATask {
-  const task: A2ATask = {
-    id: crypto.randomUUID(),
-    contextId: `session-${process.pid}`,
-    status: {
-      state: 'submitted',
-      timestamp: new Date().toISOString(),
-    },
-  };
-  tasks.set(task.id, task);
-  return task;
-}
-
-function formatTaskResult(task: A2ATask): object {
-  return {
-    kind: 'task',
-    id: task.id,
-    contextId: task.contextId,
-    status: task.status,
-  };
-}
-
-// --- JSON-RPC helpers ---
-
-interface JsonRpcRequest {
-  jsonrpc?: string;
-  id?: string | number | null;
-  method?: string;
-  params?: Record<string, unknown>;
-}
-
-function jsonRpcSuccess(id: string | number | null, result: object): object {
-  return { jsonrpc: '2.0', id, result };
-}
-
-function jsonRpcError(
-  id: string | number | null,
-  code: number,
-  message: string,
-): object {
-  return { jsonrpc: '2.0', id, error: { code, message } };
-}
-
-// --- HTTP utilities ---
-
-function getSessionsDir(): string {
-  return join(os.homedir(), '.gemini', 'sessions');
-}
-
-function getPortFilePath(): string {
-  return join(getSessionsDir(), `interactive-${process.pid}.port`);
-}
-
-function buildAgentCard(port: number): object {
-  return {
-    name: 'Gemini CLI Interactive Session',
-    url: `http://localhost:${port}/`,
-    protocolVersion: '0.3.0',
-    provider: { organization: 'Google', url: 'https://google.com' },
-    capabilities: { streaming: false, pushNotifications: false },
-    defaultInputModes: ['text'],
-    defaultOutputModes: ['text'],
-    skills: [
-      {
-        id: 'interactive_session',
-        name: 'Interactive Session',
-        description: 'Send messages to the live interactive Gemini CLI session',
-      },
-    ],
-  };
-}
-
-interface A2AMessagePart {
-  kind?: string;
-  text?: string;
-}
-
-function extractTextFromParts(
-  parts: A2AMessagePart[] | undefined,
-): string | null {
-  if (!Array.isArray(parts)) {
-    return null;
-  }
-  const texts: string[] = [];
-  for (const part of parts) {
-    if (part.kind === 'text' && typeof part.text === 'string') {
-      texts.push(part.text);
-    }
-  }
-  return texts.length > 0 ? texts.join('\n') : null;
-}
-
-function sendJson(
-  res: http.ServerResponse,
-  statusCode: number,
-  data: object,
-): void {
-  const body = JSON.stringify(data);
-  res.writeHead(statusCode, {
-    'Content-Type': 'application/json',
-    'Content-Length': Buffer.byteLength(body),
-  });
-  res.end(body);
-}
-
-function readBody(req: http.IncomingMessage): Promise<string> {
-  return new Promise((resolve, reject) => {
-    const chunks: Buffer[] = [];
-    let size = 0;
-    const maxSize = 1024 * 1024; // 1MB limit
-    req.on('data', (chunk: Buffer) => {
-      size += chunk.length;
-      if (size > maxSize) {
-        req.destroy();
-        reject(new Error('Request body too large'));
-        return;
-      }
-      chunks.push(chunk);
-    });
-    req.on('end', () => resolve(Buffer.concat(chunks).toString('utf-8')));
-    req.on('error', reject);
-  });
-}
-
-// --- JSON-RPC request handlers ---
-
-function handleMessageSend(
-  rpcId: string | number | null,
-  params: Record<string, unknown>,
-  res: http.ServerResponse,
-): void {
-  const messageVal = params['message'];
-  const message =
-    messageVal && typeof messageVal === 'object'
-      ? (messageVal as { role?: string; parts?: A2AMessagePart[] })
-      : undefined;
-  const text = extractTextFromParts(message?.parts);
-  if (!text) {
-    sendJson(
-      res,
-      200,
-      jsonRpcError(
-        rpcId,
-        -32602,
-        'Missing or empty text. Expected: params.message.parts with kind "text".',
-      ),
-    );
-    return;
-  }
-
-  const task = createTask();
-
-  // Inject message into the session
-  appEvents.emit(AppEvent.ExternalMessage, text);
-
-  // Block until response (standard A2A message/send semantics)
-  const timer = setTimeout(() => {
-    const idx = responseWaiters.findIndex((w) => w.taskId === task.id);
-    if (idx !== -1) {
-      responseWaiters.splice(idx, 1);
-    }
-    task.status = {
-      state: 'failed',
-      timestamp: new Date().toISOString(),
-    };
-    scheduleTaskCleanup(task.id);
-    sendJson(res, 200, jsonRpcError(rpcId, -32000, 'Request timed out'));
-  }, DEFAULT_BLOCKING_TIMEOUT_MS);
-
-  responseWaiters.push({
-    taskId: task.id,
-    resolve: () => {
-      clearTimeout(timer);
-      // Task is already updated in notifyResponse
-      const updatedTask = tasks.get(task.id);
-      sendJson(
-        res,
-        200,
-        jsonRpcSuccess(rpcId, formatTaskResult(updatedTask ?? task)),
-      );
-    },
-  });
-}
-
-function handleResponsesPoll(
-  rpcId: string | number | null,
-  res: http.ServerResponse,
-): void {
-  const responses = drainUnsolicitedResponses();
-  sendJson(res, 200, jsonRpcSuccess(rpcId, { responses }));
-}
-
-function handleTasksGet(
-  rpcId: string | number | null,
-  params: Record<string, unknown>,
-  res: http.ServerResponse,
-): void {
-  const taskId = params['id'];
-  if (typeof taskId !== 'string') {
-    sendJson(
-      res,
-      200,
-      jsonRpcError(rpcId, -32602, 'Missing or invalid params.id'),
-    );
-    return;
-  }
-
-  const task = tasks.get(taskId);
-  if (!task) {
-    sendJson(res, 200, jsonRpcError(rpcId, -32001, 'Task not found'));
-    return;
-  }
-
-  sendJson(res, 200, jsonRpcSuccess(rpcId, formatTaskResult(task)));
-}
-
-// --- Server ---
-
-export interface ExternalListenerResult {
-  port: number;
-  cleanup: () => void;
-}
-
-/**
- * Start an embedded HTTP server that accepts A2A-format JSON-RPC messages
- * and bridges them into the interactive session's message queue.
- */
-export function startExternalListener(options?: {
-  port?: number;
-}): Promise<ExternalListenerResult> {
-  const port = options?.port ?? 0;
-
-  return new Promise((resolve, reject) => {
-    const server = http.createServer(
-      (req: http.IncomingMessage, res: http.ServerResponse) => {
-        const url = new URL(req.url ?? '/', `http://localhost`);
-
-        // GET /.well-known/agent-card.json
-        if (
-          req.method === 'GET' &&
-          url.pathname === '/.well-known/agent-card.json'
-        ) {
-          const address = server.address();
-          const actualPort =
-            typeof address === 'object' && address ? address.port : port;
-          sendJson(res, 200, buildAgentCard(actualPort));
-          return;
-        }
-
-        // POST / — JSON-RPC 2.0 routing
-        if (req.method === 'POST' && url.pathname === '/') {
-          readBody(req)
-            .then((rawBody) => {
-              let parsed: JsonRpcRequest;
-              try {
-                // eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
-                parsed = JSON.parse(rawBody) as JsonRpcRequest;
-              } catch {
-                sendJson(
-                  res,
-                  200,
-                  jsonRpcError(null, -32700, 'Parse error: invalid JSON'),
-                );
-                return;
-              }
-
-              const rpcId = parsed.id ?? null;
-              const method = parsed.method;
-              const params = parsed.params ?? {};
-
-              switch (method) {
-                case 'message/send':
-                  handleMessageSend(rpcId, params, res);
-                  break;
-                case 'tasks/get':
-                  handleTasksGet(rpcId, params, res);
-                  break;
-                case 'responses/poll':
-                  handleResponsesPoll(rpcId, res);
-                  break;
-                default:
-                  sendJson(
-                    res,
-                    200,
-                    jsonRpcError(
-                      rpcId,
-                      -32601,
-                      `Method not found: ${method ?? '(none)'}`,
-                    ),
-                  );
-              }
-            })
-            .catch(() => {
-              sendJson(
-                res,
-                200,
-                jsonRpcError(null, -32603, 'Failed to read request body'),
-              );
-            });
-          return;
-        }
-
-        // 404 for everything else
-        sendJson(res, 404, { error: 'Not found' });
-      },
-    );
-
-    server.listen(port, '127.0.0.1', () => {
-      const address = server.address();
-      const actualPort =
-        typeof address === 'object' && address ? address.port : port;
-
-      // Write port file
-      try {
-        const sessionsDir = getSessionsDir();
-        mkdirSync(sessionsDir, { recursive: true });
-        writeFileSync(getPortFilePath(), String(actualPort), 'utf-8');
-      } catch {
-        // Non-fatal: port file is a convenience, not a requirement
-      }
-
-      const cleanup = () => {
-        server.close();
-        try {
-          unlinkSync(getPortFilePath());
-        } catch {
-          // Ignore: file may already be deleted
-        }
-      };
-
-      resolve({ port: actualPort, cleanup });
-    });
-
-    server.on('error', (err) => {
-      reject(err);
-    });
-  });
-}

packages/cli/src/gemini.test.tsx

@@ -513,7 +513,6 @@ describe('gemini.tsx main function kitty protocol', () => {
       rawOutput: undefined,
       acceptRawOutputRisk: undefined,
       isCommand: undefined,
-      forever: undefined,
     });
 
     await act(async () => {

packages/cli/src/interactiveCli.tsx

@@ -10,8 +10,6 @@ import { basename } from 'node:path';
 import { AppContainer } from './ui/AppContainer.js';
 import { ConsolePatcher } from './ui/utils/ConsolePatcher.js';
 import { registerCleanup, setupTtyCheck } from './utils/cleanup.js';
-import { startExternalListener } from './external-listener.js';
-import { appEvents, AppEvent } from './utils/events.js';
 import {
   type StartupWarning,
   type Config,
@@ -183,24 +181,6 @@ export async function startInteractiveUI(
 
   registerCleanup(() => instance.unmount());
 
-  // Auto-start A2A HTTP listener in Forever Mode
-  if (config.getIsForeverMode()) {
-    try {
-      const listener = await startExternalListener({ port: 0 });
-      registerCleanup(listener.cleanup);
-      appEvents.emit(AppEvent.A2AListenerStarted, listener.port);
-      coreEvents.emitFeedback(
-        'info',
-        `A2A endpoint listening on port ${listener.port}`,
-      );
-    } catch (err) {
-      coreEvents.emitFeedback(
-        'warning',
-        `Failed to start A2A listener: ${err instanceof Error ? err.message : String(err)}`,
-      );
-    }
-  }
-
   registerCleanup(setupTtyCheck());
 }
 

packages/cli/src/nonInteractiveCli.ts

@@ -222,7 +222,6 @@ export async function runNonInteractive({
         await geminiClient.resumeChat(
           convertSessionToClientHistory(
             resumedSessionData.conversation.messages,
-            resumedSessionData.conversation.lastCompressionIndex,
           ),
           resumedSessionData,
         );

packages/cli/src/test-utils/mockConfig.ts

@@ -173,14 +173,6 @@ export const createMockConfig = (overrides: Partial<Config> = {}): Config =>
     getHasAccessToPreviewModel: vi.fn().mockReturnValue(false),
     validatePathAccess: vi.fn().mockReturnValue(null),
     getUseAlternateBuffer: vi.fn().mockReturnValue(false),
-    getIsForeverMode: vi.fn().mockReturnValue(false),
-    getWorkScheduler: vi.fn().mockReturnValue({
-      on: vi.fn(),
-      off: vi.fn(),
-      serialize: vi.fn().mockReturnValue([]),
-      restore: vi.fn(),
-      dispose: vi.fn(),
-    }),
     ...overrides,
   }) as unknown as Config;
 

packages/cli/src/ui/App.test.tsx

@@ -82,7 +82,6 @@ describe('App', () => {
       updateItem: vi.fn(),
       clearItems: vi.fn(),
       loadHistory: vi.fn(),
-      pruneItems: vi.fn(),
     },
     history: [],
     pendingHistoryItems: [],

packages/cli/src/ui/AppContainer.tsx

@@ -127,7 +127,6 @@ import { useFolderTrust } from './hooks/useFolderTrust.js';
 import { useIdeTrustListener } from './hooks/useIdeTrustListener.js';
 import { type IdeIntegrationNudgeResult } from './IdeIntegrationNudge.js';
 import { appEvents, AppEvent, TransientMessageType } from '../utils/events.js';
-import { notifyResponse, markTasksWorking } from '../external-listener.js';
 import { type UpdateObject } from './utils/updateCheck.js';
 import { setUpdateHandler } from '../utils/handleAutoUpdate.js';
 import { registerCleanup, runExitCleanup } from '../utils/cleanup.js';
@@ -1143,7 +1142,6 @@ Logging in with Google... Restarting Gemini CLI to continue.
     terminalHeight,
     embeddedShellFocused,
     consumePendingHints,
-    historyManager.pruneItems,
   );
 
   toggleBackgroundShellRef.current = toggleBackgroundShell;
@@ -1215,89 +1213,6 @@ Logging in with Google... Restarting Gemini CLI to continue.
     isMcpReady,
   });
 
-  // --- A2A listener integration ---
-  const [a2aListenerPort, setA2aListenerPort] = useState<number | null>(null);
-
-  useEffect(() => {
-    const handler = (port: number) => {
-      setA2aListenerPort(port);
-    };
-    appEvents.on(AppEvent.A2AListenerStarted, handler);
-    return () => {
-      appEvents.off(AppEvent.A2AListenerStarted, handler);
-    };
-  }, []);
-
-  // Bridge external messages from A2A HTTP listener to message queue
-  useEffect(() => {
-    const handler = (text: string) => {
-      addMessage(text);
-    };
-    appEvents.on(AppEvent.ExternalMessage, handler);
-    return () => {
-      appEvents.off(AppEvent.ExternalMessage, handler);
-    };
-  }, [addMessage]);
-
-  // Wire WorkScheduler: inject fired prompts and persist schedule changes
-  useEffect(() => {
-    const scheduler = config.getWorkScheduler();
-
-    const onFire = (prompt: string) => {
-      appEvents.emit(AppEvent.ExternalMessage, prompt);
-    };
-
-    const onChanged = () => {
-      // Persist pending items to the session file
-      const recordingService = config
-        .getGeminiClient()
-        ?.getChatRecordingService();
-      if (recordingService) {
-        recordingService.recordScheduledWork(scheduler.serialize());
-      }
-    };
-
-    scheduler.on('fire', onFire);
-    scheduler.on('changed', onChanged);
-    return () => {
-      scheduler.off('fire', onFire);
-      scheduler.off('changed', onChanged);
-    };
-  }, [config]);
-
-  // Track streaming state transitions for A2A response capture
-  const prevStreamingStateRef = useRef(streamingState);
-
-  useEffect(() => {
-    const prev = prevStreamingStateRef.current;
-    prevStreamingStateRef.current = streamingState;
-
-    // Mark tasks as "working" when streaming starts
-    if (
-      prev === StreamingState.Idle &&
-      streamingState !== StreamingState.Idle
-    ) {
-      markTasksWorking();
-    }
-
-    // Capture response when streaming ends (for A2A tasks or unsolicited output)
-    if (
-      prev !== StreamingState.Idle &&
-      streamingState === StreamingState.Idle
-    ) {
-      const history = historyManager.history;
-      const parts: string[] = [];
-      for (let i = history.length - 1; i >= 0; i--) {
-        const item = history[i];
-        if (item.type !== 'gemini' && item.type !== 'gemini_content') break;
-        if (typeof item.text === 'string' && item.text) {
-          parts.unshift(item.text);
-        }
-      }
-      notifyResponse(parts.join('\n'));
-    }
-  }, [streamingState, historyManager.history]);
-
   cancelHandlerRef.current = useCallback(
     (shouldRestorePrompt: boolean = true) => {
       const pendingHistoryItems = [
@@ -2385,7 +2300,6 @@ Logging in with Google... Restarting Gemini CLI to continue.
       adminSettingsChanged,
       newAgents,
       showIsExpandableHint,
-      a2aListenerPort,
       hintMode:
         config.isModelSteeringEnabled() &&
         isToolExecuting([
@@ -2514,7 +2428,6 @@ Logging in with Google... Restarting Gemini CLI to continue.
       adminSettingsChanged,
       newAgents,
       showIsExpandableHint,
-      a2aListenerPort,
     ],
   );
 

packages/cli/src/ui/components/AnsiOutput.tsx

@@ -35,7 +35,11 @@ export const AnsiOutputText: React.FC<AnsiOutputProps> = ({
       ? Math.min(availableHeightLimit, maxLines)
       : (availableHeightLimit ?? maxLines ?? DEFAULT_HEIGHT);
 
-  const lastLines = disableTruncation ? data : data.slice(-numLinesRetained);
+  const lastLines = disableTruncation
+    ? data
+    : numLinesRetained === 0
+      ? []
+      : data.slice(-numLinesRetained);
   return (
     <Box flexDirection="column" width={width} flexShrink={0} overflow="hidden">
       {lastLines.map((line: AnsiLine, lineIndex: number) => (

packages/cli/src/ui/components/Banner.test.tsx

@@ -4,7 +4,7 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 
-import { render } from '../../test-utils/render.js';
+import { renderWithProviders } from '../../test-utils/render.js';
 import { Banner } from './Banner.js';
 import { describe, it, expect } from 'vitest';
 
@@ -12,22 +12,23 @@ describe('Banner', () => {
   it.each([
     ['warning mode', true, 'Warning Message'],
     ['info mode', false, 'Info Message'],
+    ['multi-line warning', true, 'Title Line\\nBody Line 1\\nBody Line 2'],
   ])('renders in %s', async (_, isWarning, text) => {
-    const { lastFrame, waitUntilReady, unmount } = render(
+    const renderResult = renderWithProviders(
       <Banner bannerText={text} isWarning={isWarning} width={80} />,
     );
-    await waitUntilReady();
-    expect(lastFrame()).toMatchSnapshot();
-    unmount();
+    await renderResult.waitUntilReady();
+    await expect(renderResult).toMatchSvgSnapshot();
+    renderResult.unmount();
   });
 
   it('handles newlines in text', async () => {
     const text = 'Line 1\\nLine 2';
-    const { lastFrame, waitUntilReady, unmount } = render(
+    const renderResult = renderWithProviders(
       <Banner bannerText={text} isWarning={false} width={80} />,
     );
-    await waitUntilReady();
-    expect(lastFrame()).toMatchSnapshot();
-    unmount();
+    await renderResult.waitUntilReady();
+    await expect(renderResult).toMatchSvgSnapshot();
+    renderResult.unmount();
   });
 });

packages/cli/src/ui/components/Banner.tsx

@@ -14,20 +14,21 @@ export function getFormattedBannerContent(
   isWarning: boolean,
   subsequentLineColor: string,
 ): ReactNode {
-  if (isWarning) {
-    return (
-      <Text color={theme.status.warning}>{rawText.replace(/\\n/g, '\n')}</Text>
-    );
-  }
-
   const text = rawText.replace(/\\n/g, '\n');
   const lines = text.split('\n');
 
   return lines.map((line, index) => {
     if (index === 0) {
+      if (isWarning) {
+        return (
+          <Text key={index} bold color={theme.status.warning}>
+            {line}
+          </Text>
+        );
+      }
       return (
         <ThemedGradient key={index}>
-          <Text>{line}</Text>
+          <Text bold>{line}</Text>
         </ThemedGradient>
       );
     }

packages/cli/src/ui/components/Composer.test.tsx

@@ -61,10 +61,6 @@ vi.mock('./StatusDisplay.js', () => ({
   StatusDisplay: () => <Text>StatusDisplay</Text>,
 }));
 
-vi.mock('./ScheduledWorkDisplay.js', () => ({
-  ScheduledWorkDisplay: () => null,
-}));
-
 vi.mock('./ToastDisplay.js', () => ({
   ToastDisplay: () => <Text>ToastDisplay</Text>,
   shouldShowToast: (uiState: UIState) =>
@@ -206,7 +202,6 @@ const createMockUIState = (overrides: Partial<UIState> = {}): UIState =>
     activeHooks: [],
     isBackgroundShellVisible: false,
     embeddedShellFocused: false,
-    a2aListenerPort: null,
     quota: {
       userTier: undefined,
       stats: undefined,

packages/cli/src/ui/components/Composer.tsx

@@ -13,7 +13,6 @@ import {
 } from '@google/gemini-cli-core';
 import { LoadingIndicator } from './LoadingIndicator.js';
 import { StatusDisplay } from './StatusDisplay.js';
-import { ScheduledWorkDisplay } from './ScheduledWorkDisplay.js';
 import { ToastDisplay, shouldShowToast } from './ToastDisplay.js';
 import { ApprovalModeIndicator } from './ApprovalModeIndicator.js';
 import { ShellModeIndicator } from './ShellModeIndicator.js';
@@ -337,7 +336,6 @@ export const Composer = ({ isFocused = true }: { isFocused?: boolean }) => {
         )}
         {showShortcutsHelp && <ShortcutsHelp />}
         {showUiDetails && <HorizontalLine />}
-        <ScheduledWorkDisplay />
         {showUiDetails && (
           <Box
             justifyContent={

packages/cli/src/ui/components/MainContent.tsx

@@ -48,6 +48,7 @@ export const MainContent = () => {
     pendingHistoryItems,
     mainAreaWidth,
     staticAreaMaxItemHeight,
+    availableTerminalHeight,
     cleanUiDetailsVisible,
   } = uiState;
   const showHeaderDetails = cleanUiDetailsVisible;
@@ -141,7 +142,7 @@ export const MainContent = () => {
             <HistoryItemDisplay
               key={i}
               availableTerminalHeight={
-                uiState.constrainHeight ? staticAreaMaxItemHeight : undefined
+                uiState.constrainHeight ? availableTerminalHeight : undefined
               }
               terminalWidth={mainAreaWidth}
               item={{ ...item, id: 0 }}
@@ -160,7 +161,7 @@ export const MainContent = () => {
     [
       pendingHistoryItems,
       uiState.constrainHeight,
-      staticAreaMaxItemHeight,
+      availableTerminalHeight,
       mainAreaWidth,
       showConfirmationQueue,
       confirmingTool,

packages/cli/src/ui/components/ScheduledWorkDisplay.tsx

@@ -1,66 +0,0 @@
-/**
- * @license
- * Copyright 2026 Google LLC
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import type React from 'react';
-import { useState, useEffect } from 'react';
-import { Box, Text } from 'ink';
-import { theme } from '../semantic-colors.js';
-import { useConfig } from '../contexts/ConfigContext.js';
-import type { ScheduledItem } from '@google/gemini-cli-core';
-
-/**
- * Displays all pending scheduled work items above the context summary.
- * Only renders when there are pending items.
- */
-export const ScheduledWorkDisplay: React.FC = () => {
-  const config = useConfig();
-  const [items, setItems] = useState<readonly ScheduledItem[]>([]);
-
-  useEffect(() => {
-    const scheduler = config.getWorkScheduler();
-
-    const update = () => {
-      setItems(scheduler.getPendingItems());
-    };
-
-    scheduler.on('changed', update);
-    update();
-
-    return () => {
-      scheduler.off('changed', update);
-    };
-  }, [config]);
-
-  if (items.length === 0) {
-    return null;
-  }
-
-  return (
-    <Box flexDirection="column" paddingX={1}>
-      <Text color={theme.text.secondary}>
-        ⏰ Scheduled work ({items.length}):
-      </Text>
-      {items.map((item) => {
-        const timeStr = item.fireAt.toLocaleTimeString([], {
-          hour: '2-digit',
-          minute: '2-digit',
-        });
-        const diffMs = item.fireAt.getTime() - Date.now();
-        const diffMins = Math.max(0, Math.ceil(diffMs / 60000));
-        const truncatedPrompt =
-          item.prompt.length > 60
-            ? item.prompt.slice(0, 57) + '...'
-            : item.prompt;
-        return (
-          <Text key={item.id} color={theme.text.secondary}>
-            {'  '}
-            {timeStr} (in {diffMins}m) — {truncatedPrompt}
-          </Text>
-        );
-      })}
-    </Box>
-  );
-};

packages/cli/src/ui/components/StatusDisplay.test.tsx

@@ -52,7 +52,6 @@ const createMockUIState = (overrides: UIStateOverrides = {}): UIState =>
     geminiMdFileCount: 0,
     contextFileNames: [],
     backgroundShellCount: 0,
-    a2aListenerPort: null,
     buffer: { text: '' },
     history: [{ id: 1, type: 'user', text: 'test' }],
     ...overrides,
@@ -172,16 +171,4 @@ describe('StatusDisplay', () => {
     expect(lastFrame()).toContain('Shells: 3');
     unmount();
   });
-
-  it('renders A2A listener port when a2aListenerPort is set', async () => {
-    const uiState = createMockUIState({
-      a2aListenerPort: 8080,
-    });
-    const { lastFrame, unmount } = await renderStatusDisplay(
-      { hideContextSummary: false },
-      uiState,
-    );
-    expect(lastFrame()).toContain('A2A :8080');
-    unmount();
-  });
 });

packages/cli/src/ui/components/StatusDisplay.tsx

@@ -28,12 +28,6 @@ export const StatusDisplay: React.FC<StatusDisplayProps> = ({
     return <Text color={theme.status.error}>|⌐■_■|</Text>;
   }
 
-  if (uiState.a2aListenerPort !== null) {
-    return (
-      <Text color={theme.text.accent}>⚡ A2A :{uiState.a2aListenerPort}</Text>
-    );
-  }
-
   if (
     uiState.activeHooks.length > 0 &&
     settings.merged.hooksConfig.notifications

packages/cli/src/ui/components/__snapshots__/Banner-Banner-handles-newlines-in-text.snap.svg

@@ -0,0 +1,20 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="920" height="88" viewBox="0 0 920 88">
+  <style>
+    text { font-family: Consolas, "Courier New", monospace; font-size: 14px; dominant-baseline: text-before-edge; white-space: pre; }
+  </style>
+  <rect width="920" height="88" fill="#000000" />
+  <g transform="translate(10, 10)">
+    <text x="0" y="2" fill="#333333" textLength="720" lengthAdjust="spacingAndGlyphs">╭──────────────────────────────────────────────────────────────────────────────╮</text>
+    <text x="0" y="19" fill="#333333" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="19" fill="#4796e4" textLength="9" lengthAdjust="spacingAndGlyphs">L</text>
+    <text x="27" y="19" fill="#6688d9" textLength="9" lengthAdjust="spacingAndGlyphs">i</text>
+    <text x="36" y="19" fill="#847ace" textLength="9" lengthAdjust="spacingAndGlyphs">n</text>
+    <text x="45" y="19" fill="#9974b4" textLength="9" lengthAdjust="spacingAndGlyphs">e</text>
+    <text x="63" y="19" fill="#c3677f" textLength="9" lengthAdjust="spacingAndGlyphs">1</text>
+    <text x="711" y="19" fill="#333333" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="36" fill="#333333" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="36" fill="#ffffff" textLength="54" lengthAdjust="spacingAndGlyphs">Line 2</text>
+    <text x="711" y="36" fill="#333333" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="53" fill="#333333" textLength="720" lengthAdjust="spacingAndGlyphs">╰──────────────────────────────────────────────────────────────────────────────╯</text>
+  </g>
+</svg>

packages/cli/src/ui/components/__snapshots__/Banner-Banner-renders-in-info-mode.snap.svg

@@ -0,0 +1,23 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="920" height="71" viewBox="0 0 920 71">
+  <style>
+    text { font-family: Consolas, "Courier New", monospace; font-size: 14px; dominant-baseline: text-before-edge; white-space: pre; }
+  </style>
+  <rect width="920" height="71" fill="#000000" />
+  <g transform="translate(10, 10)">
+    <text x="0" y="2" fill="#333333" textLength="720" lengthAdjust="spacingAndGlyphs">╭──────────────────────────────────────────────────────────────────────────────╮</text>
+    <text x="0" y="19" fill="#333333" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="19" fill="#4796e4" textLength="9" lengthAdjust="spacingAndGlyphs">I</text>
+    <text x="27" y="19" fill="#5390e0" textLength="9" lengthAdjust="spacingAndGlyphs">n</text>
+    <text x="36" y="19" fill="#5f8bdb" textLength="9" lengthAdjust="spacingAndGlyphs">f</text>
+    <text x="45" y="19" fill="#6c85d7" textLength="9" lengthAdjust="spacingAndGlyphs">o</text>
+    <text x="63" y="19" fill="#847ace" textLength="9" lengthAdjust="spacingAndGlyphs">M</text>
+    <text x="72" y="19" fill="#8f77c1" textLength="9" lengthAdjust="spacingAndGlyphs">e</text>
+    <text x="81" y="19" fill="#9974b4" textLength="9" lengthAdjust="spacingAndGlyphs">s</text>
+    <text x="90" y="19" fill="#a471a7" textLength="9" lengthAdjust="spacingAndGlyphs">s</text>
+    <text x="99" y="19" fill="#ae6d99" textLength="9" lengthAdjust="spacingAndGlyphs">a</text>
+    <text x="108" y="19" fill="#b96a8c" textLength="9" lengthAdjust="spacingAndGlyphs">g</text>
+    <text x="117" y="19" fill="#c3677f" textLength="9" lengthAdjust="spacingAndGlyphs">e</text>
+    <text x="711" y="19" fill="#333333" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="36" fill="#333333" textLength="720" lengthAdjust="spacingAndGlyphs">╰──────────────────────────────────────────────────────────────────────────────╯</text>
+  </g>
+</svg>

packages/cli/src/ui/components/__snapshots__/Banner-Banner-renders-in-multi-line-warning.snap.svg

@@ -0,0 +1,19 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="920" height="105" viewBox="0 0 920 105">
+  <style>
+    text { font-family: Consolas, "Courier New", monospace; font-size: 14px; dominant-baseline: text-before-edge; white-space: pre; }
+  </style>
+  <rect width="920" height="105" fill="#000000" />
+  <g transform="translate(10, 10)">
+    <text x="0" y="2" fill="#ffffaf" textLength="720" lengthAdjust="spacingAndGlyphs">╭──────────────────────────────────────────────────────────────────────────────╮</text>
+    <text x="0" y="19" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="19" fill="#ffffaf" textLength="90" lengthAdjust="spacingAndGlyphs" font-weight="bold">Title Line</text>
+    <text x="711" y="19" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="36" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="36" fill="#ffffff" textLength="99" lengthAdjust="spacingAndGlyphs">Body Line 1</text>
+    <text x="711" y="36" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="53" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="53" fill="#ffffff" textLength="99" lengthAdjust="spacingAndGlyphs">Body Line 2</text>
+    <text x="711" y="53" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="70" fill="#ffffaf" textLength="720" lengthAdjust="spacingAndGlyphs">╰──────────────────────────────────────────────────────────────────────────────╯</text>
+  </g>
+</svg>

packages/cli/src/ui/components/__snapshots__/Banner-Banner-renders-in-warning-mode.snap.svg

@@ -0,0 +1,13 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="920" height="71" viewBox="0 0 920 71">
+  <style>
+    text { font-family: Consolas, "Courier New", monospace; font-size: 14px; dominant-baseline: text-before-edge; white-space: pre; }
+  </style>
+  <rect width="920" height="71" fill="#000000" />
+  <g transform="translate(10, 10)">
+    <text x="0" y="2" fill="#ffffaf" textLength="720" lengthAdjust="spacingAndGlyphs">╭──────────────────────────────────────────────────────────────────────────────╮</text>
+    <text x="0" y="19" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="18" y="19" fill="#ffffaf" textLength="135" lengthAdjust="spacingAndGlyphs" font-weight="bold">Warning Message</text>
+    <text x="711" y="19" fill="#ffffaf" textLength="9" lengthAdjust="spacingAndGlyphs">│</text>
+    <text x="0" y="36" fill="#ffffaf" textLength="720" lengthAdjust="spacingAndGlyphs">╰──────────────────────────────────────────────────────────────────────────────╯</text>
+  </g>
+</svg>

packages/cli/src/ui/components/__snapshots__/Banner.test.tsx.snap

@@ -4,20 +4,25 @@ exports[`Banner > handles newlines in text 1`] = `
 "╭──────────────────────────────────────────────────────────────────────────────╮
 │ Line 1                                                                       │
 │ Line 2                                                                       │
-╰──────────────────────────────────────────────────────────────────────────────╯
-"
+╰──────────────────────────────────────────────────────────────────────────────╯"
 `;
 
 exports[`Banner > renders in info mode 1`] = `
 "╭──────────────────────────────────────────────────────────────────────────────╮
 │ Info Message                                                                 │
-╰──────────────────────────────────────────────────────────────────────────────╯
-"
+╰──────────────────────────────────────────────────────────────────────────────╯"
+`;
+
+exports[`Banner > renders in multi-line warning 1`] = `
+"╭──────────────────────────────────────────────────────────────────────────────╮
+│ Title Line                                                                   │
+│ Body Line 1                                                                  │
+│ Body Line 2                                                                  │
+╰──────────────────────────────────────────────────────────────────────────────╯"
 `;
 
 exports[`Banner > renders in warning mode 1`] = `
 "╭──────────────────────────────────────────────────────────────────────────────╮
 │ Warning Message                                                              │
-╰──────────────────────────────────────────────────────────────────────────────╯
-"
+╰──────────────────────────────────────────────────────────────────────────────╯"
 `;

packages/cli/src/ui/components/__snapshots__/MainContent.test.tsx.snap

@@ -6,11 +6,12 @@ AppHeader(full)
 ╭──────────────────────────────────────────────────────────────────────────────────────────────╮
 │ ⊶  Shell Command Running a long command...                                                   │
 │                                                                                              │
+│ Line 9                                                                                       │
 │ Line 10                                                                                      │
 │ Line 11                                                                                      │
 │ Line 12                                                                                      │
 │ Line 13                                                                                      │
-│ Line 14                                                                                      │
+│ Line 14                                                                                    █ │
 │ Line 15                                                                                    █ │
 │ Line 16                                                                                    █ │
 │ Line 17                                                                                    █ │
@@ -27,11 +28,12 @@ AppHeader(full)
 ╭──────────────────────────────────────────────────────────────────────────────────────────────╮
 │ ⊶  Shell Command Running a long command...                                                   │
 │                                                                                              │
+│ Line 9                                                                                       │
 │ Line 10                                                                                      │
 │ Line 11                                                                                      │
 │ Line 12                                                                                      │
 │ Line 13                                                                                      │
-│ Line 14                                                                                      │
+│ Line 14                                                                                    █ │
 │ Line 15                                                                                    █ │
 │ Line 16                                                                                    █ │
 │ Line 17                                                                                    █ │
@@ -47,7 +49,9 @@ exports[`MainContent > MainContent Tool Output Height Logic > 'Normal mode - Con
 ╭──────────────────────────────────────────────────────────────────────────────────────────────╮
 │ ⊶  Shell Command Running a long command...                                                   │
 │                                                                                              │
-│ ... first 11 lines hidden (Ctrl+O to show) ...                                               │
+│ ... first 9 lines hidden (Ctrl+O to show) ...                                                │
+│ Line 10                                                                                      │
+│ Line 11                                                                                      │
 │ Line 12                                                                                      │
 │ Line 13                                                                                      │
 │ Line 14                                                                                      │

packages/cli/src/ui/components/messages/ShellToolMessage.test.tsx

@@ -199,7 +199,7 @@ describe('<ShellToolMessage />', () => {
       [
         'uses full availableTerminalHeight when focused in alternate buffer mode',
         100,
-        98, // 100 - 2
+        98,
         true,
         false,
       ],

packages/cli/src/ui/components/messages/ToolConfirmationMessage.test.tsx

@@ -4,16 +4,18 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 
-import { describe, it, expect, vi } from 'vitest';
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
 import { ToolConfirmationMessage } from './ToolConfirmationMessage.js';
-import type {
-  SerializableConfirmationDetails,
-  ToolCallConfirmationDetails,
-  Config,
+import {
+  type SerializableConfirmationDetails,
+  type ToolCallConfirmationDetails,
+  type Config,
+  ToolConfirmationOutcome,
 } from '@google/gemini-cli-core';
 import { renderWithProviders } from '../../../test-utils/render.js';
 import { createMockSettings } from '../../../test-utils/settings.js';
 import { useToolActions } from '../../contexts/ToolActionsContext.js';
+import { act } from 'react';
 
 vi.mock('../../contexts/ToolActionsContext.js', async (importOriginal) => {
   const actual =
@@ -646,4 +648,63 @@ describe('ToolConfirmationMessage', () => {
     expect(output).not.toContain('Invocation Arguments:');
     unmount();
   });
+
+  describe('ESCAPE key behavior', () => {
+    beforeEach(() => {
+      vi.useFakeTimers();
+    });
+
+    afterEach(() => {
+      vi.useRealTimers();
+      vi.restoreAllMocks();
+    });
+
+    it('should call confirm(Cancel) asynchronously via useEffect when ESC is pressed', async () => {
+      const mockConfirm = vi.fn().mockResolvedValue(undefined);
+
+      vi.mocked(useToolActions).mockReturnValue({
+        confirm: mockConfirm,
+        cancel: vi.fn(),
+        isDiffingEnabled: false,
+      });
+
+      const confirmationDetails: SerializableConfirmationDetails = {
+        type: 'info',
+        title: 'Confirm Web Fetch',
+        prompt: 'https://example.com',
+        urls: ['https://example.com'],
+      };
+
+      const { stdin, waitUntilReady, unmount } = renderWithProviders(
+        <ToolConfirmationMessage
+          callId="test-call-id"
+          confirmationDetails={confirmationDetails}
+          config={mockConfig}
+          getPreferredEditor={vi.fn()}
+          availableTerminalHeight={30}
+          terminalWidth={80}
+        />,
+      );
+      await waitUntilReady();
+
+      stdin.write('\x1b');
+
+      // To assert that the confirmation happens asynchronously (via useEffect) rather than
+      // synchronously (directly inside the keystroke handler), we must run our assertion
+      // *inside* the act() block.
+      await act(async () => {
+        await vi.runAllTimersAsync();
+        expect(mockConfirm).not.toHaveBeenCalled();
+      });
+
+      // Now that the act() block has returned, React flushes the useEffect, calling handleConfirm.
+      expect(mockConfirm).toHaveBeenCalledWith(
+        'test-call-id',
+        ToolConfirmationOutcome.Cancel,
+        undefined,
+      );
+
+      unmount();
+    });
+  });
 });

packages/cli/src/ui/components/messages/ToolConfirmationMessage.tsx

@@ -5,7 +5,7 @@
  */
 
 import type React from 'react';
-import { useMemo, useCallback, useState } from 'react';
+import { useEffect, useMemo, useCallback, useState } from 'react';
 import { Box, Text } from 'ink';
 import { DiffRenderer } from './DiffRenderer.js';
 import { RenderInline } from '../../utils/InlineMarkdownRenderer.js';
@@ -79,6 +79,7 @@ export const ToolConfirmationMessage: React.FC<
     callId,
     expanded: false,
   });
+  const [isCancelling, setIsCancelling] = useState(false);
   const isMcpToolDetailsExpanded =
     mcpDetailsExpansionState.callId === callId
       ? mcpDetailsExpansionState.expanded
@@ -183,7 +184,7 @@ export const ToolConfirmationMessage: React.FC<
         return true;
       }
       if (keyMatchers[Command.ESCAPE](key)) {
-        handleConfirm(ToolConfirmationOutcome.Cancel);
+        setIsCancelling(true);
         return true;
       }
       if (keyMatchers[Command.QUIT](key)) {
@@ -196,6 +197,20 @@ export const ToolConfirmationMessage: React.FC<
     { isActive: isFocused, priority: true },
   );
 
+  // TODO(#23009): Remove this hack once we migrate to the new renderer.
+  // Why useEffect is used here instead of calling handleConfirm directly:
+  // There is a race condition where calling handleConfirm immediately upon
+  // keypress removes the tool UI component while the UI is in an expanded state.
+  // This simultaneously triggers setConstrainHeight, causing render two footers.
+  // By bridging the cancel action through state (isCancelling) and this useEffect,
+  // we delay handleConfirm until the next render cycle, ensuring setConstrainHeight
+  // resolves properly first.
+  useEffect(() => {
+    if (isCancelling) {
+      handleConfirm(ToolConfirmationOutcome.Cancel);
+    }
+  }, [isCancelling, handleConfirm]);
+
   const handleSelect = useCallback(
     (item: ToolConfirmationOutcome) => handleConfirm(item),
     [handleConfirm],

packages/cli/src/ui/components/shared/SlicingMaxSizedBox.tsx

@@ -46,7 +46,7 @@ export function SlicingMaxSizedBox<T>({
           text = '...' + text.slice(-MAXIMUM_RESULT_DISPLAY_CHARACTERS);
         }
       }
-      if (maxLines) {
+      if (maxLines !== undefined) {
         const hasTrailingNewline = text.endsWith('\n');
         const contentText = hasTrailingNewline ? text.slice(0, -1) : text;
         const lines = contentText.split('\n');
@@ -71,7 +71,7 @@ export function SlicingMaxSizedBox<T>({
       };
     }
 
-    if (Array.isArray(data) && !isAlternateBuffer && maxLines) {
+    if (Array.isArray(data) && !isAlternateBuffer && maxLines !== undefined) {
       if (data.length > maxLines) {
         // We will have a label from MaxSizedBox. Reserve space for it.
         const targetLines = Math.max(1, maxLines - 1);

packages/cli/src/ui/constants/tips.ts

@@ -75,7 +75,6 @@ export const INFORMATIVE_TIPS = [
   'Set the character threshold for truncating tool outputs (/settings)…',
   'Set the number of lines to keep when truncating outputs (/settings)…',
   'Enable policy-based tool confirmation via message bus (/settings)…',
-  'Enable write_todos_list tool to generate task lists (/settings)…',
   'Enable experimental subagents for task delegation (/settings)…',
   'Enable extension management features (settings.json)…',
   'Enable extension reloading within the CLI session (settings.json)…',

packages/cli/src/ui/contexts/UIStateContext.tsx

@@ -223,7 +223,6 @@ export interface UIState {
   showIsExpandableHint: boolean;
   hintMode: boolean;
   hintBuffer: string;
-  a2aListenerPort: number | null;
   transientMessage: {
     text: string;
     type: TransientMessageType;

packages/cli/src/ui/hooks/creditsFlowHandler.test.ts

@@ -57,7 +57,6 @@ describe('handleCreditsFlow', () => {
       updateItem: vi.fn(),
       clearItems: vi.fn(),
       loadHistory: vi.fn(),
-      pruneItems: vi.fn(),
     };
     isDialogPending = { current: false };
     mockSetOverageMenuRequest = vi.fn();

packages/cli/src/ui/hooks/useGeminiStream.ts

@@ -220,7 +220,6 @@ export const useGeminiStream = (
   terminalHeight: number,
   isShellFocused?: boolean,
   consumeUserHint?: () => string | null,
-  pruneItems?: () => void,
 ) => {
   const [initError, setInitError] = useState<string | null>(null);
   const [retryStatus, setRetryStatus] = useState<RetryAttemptPayload | null>(
@@ -257,7 +256,6 @@ export const useGeminiStream = (
   const [_isFirstToolInGroup, isFirstToolInGroupRef, setIsFirstToolInGroup] =
     useStateAndRef<boolean>(true);
   const processedMemoryToolsRef = useRef<Set<string>>(new Set());
-
   const { startNewPrompt, getPromptCount } = useSessionStats();
   const storage = config.storage;
   const logger = useLogger(storage);
@@ -1165,12 +1163,8 @@ export const useGeminiStream = (
         } as HistoryItemInfo,
         userMessageTimestamp,
       );
-
-      // Prune old UI history items to prevent unbounded memory growth
-      // in long-running sessions.
-      pruneItems?.();
     },
-    [addItem, pendingHistoryItemRef, setPendingHistoryItem, config, pruneItems],
+    [addItem, pendingHistoryItemRef, setPendingHistoryItem, config],
   );
 
   const handleMaxSessionTurnsEvent = useCallback(
@@ -1512,20 +1506,9 @@ export const useGeminiStream = (
             lastQueryRef.current = queryToSend;
             lastPromptIdRef.current = prompt_id!;
 
-            // Prepend current time (and schedule if items exist) so the
-            // model can reason about time and scheduling in forever mode.
-            let queryWithContext = queryToSend;
-            if (
-              config.getIsForeverMode() &&
-              typeof queryWithContext === 'string'
-            ) {
-              const scheduler = config.getWorkScheduler();
-              queryWithContext = `[${scheduler.formatScheduleSummary()}]\n\n${queryWithContext}`;
-            }
-
             try {
               const stream = geminiClient.sendMessageStream(
-                queryWithContext,
+                queryToSend,
                 abortSignal,
                 prompt_id!,
                 undefined,
@@ -1963,66 +1946,6 @@ export const useGeminiStream = (
     storage,
   ]);
 
-  // Idle hook timer: fires after idleTimeout seconds of no activity.
-  // The timeout is read from the Idle hook definitions themselves.
-  // If hooks exist but don't declare a timeout, fall back to 300 seconds.
-  const DEFAULT_IDLE_TIMEOUT = 300;
-  const idleTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-  useEffect(() => {
-    // Clear any existing timer
-    if (idleTimerRef.current) {
-      clearTimeout(idleTimerRef.current);
-      idleTimerRef.current = null;
-    }
-
-    if (streamingState !== StreamingState.Idle || !config.getEnableHooks()) {
-      return;
-    }
-
-    // Derive timeout from registered Idle hook definitions.
-    const hookSystem = config.getHookSystem();
-    const idleHooks = hookSystem
-      ?.getAllHooks()
-      .filter((h) => h.eventName === 'Idle' && h.enabled);
-
-    if (!idleHooks || idleHooks.length === 0) {
-      return;
-    }
-
-    // Use the max idleTimeout declared by any Idle hook, or the default.
-    const declaredTimeouts = idleHooks
-      .map((h) => h.idleTimeout)
-      .filter((t): t is number => typeof t === 'number' && t > 0);
-    const idleTimeoutSeconds =
-      declaredTimeouts.length > 0
-        ? Math.max(...declaredTimeouts)
-        : DEFAULT_IDLE_TIMEOUT;
-
-    const startTime = Date.now();
-    idleTimerRef.current = setTimeout(async () => {
-      if (!hookSystem) return;
-
-      const elapsed = Math.round((Date.now() - startTime) / 1000);
-      try {
-        const result = await hookSystem.fireIdleEvent(elapsed);
-        const prompt = result?.finalOutput?.hookSpecificOutput?.['prompt'];
-        if (typeof prompt === 'string' && prompt.trim()) {
-          // Auto-submit the prompt returned by the hook
-          void submitQuery(prompt);
-        }
-      } catch {
-        // Idle hook failures are non-fatal
-      }
-    }, idleTimeoutSeconds * 1000);
-
-    return () => {
-      if (idleTimerRef.current) {
-        clearTimeout(idleTimerRef.current);
-        idleTimerRef.current = null;
-      }
-    };
-  }, [streamingState, config, submitQuery]);
-
   const lastOutputTime = Math.max(
     lastToolOutputTime,
     lastShellOutputTime,

packages/cli/src/ui/hooks/useHistoryManager.test.ts

@@ -7,7 +7,7 @@
 import { describe, it, expect } from 'vitest';
 import { act } from 'react';
 import { renderHook } from '../../test-utils/render.js';
-import { useHistory, PRUNE_KEEP_COUNT } from './useHistoryManager.js';
+import { useHistory } from './useHistoryManager.js';
 import type { HistoryItem } from '../types.js';
 
 describe('useHistoryManager', () => {
@@ -255,82 +255,4 @@ describe('useHistoryManager', () => {
       expect(result.current.history[0].type).toBe('info');
     });
   });
-
-  describe('pruneItems', () => {
-    it('should prune history to PRUNE_KEEP_COUNT + 1 (marker) when over limit', () => {
-      const { result } = renderHook(() => useHistory());
-      const itemCount = PRUNE_KEEP_COUNT + 20;
-
-      act(() => {
-        for (let i = 0; i < itemCount; i++) {
-          result.current.addItem({
-            type: 'user',
-            text: `Message ${i}`,
-          });
-        }
-      });
-
-      expect(result.current.history).toHaveLength(itemCount);
-
-      act(() => {
-        result.current.pruneItems();
-      });
-
-      // PRUNE_KEEP_COUNT items + 1 prune marker
-      expect(result.current.history).toHaveLength(PRUNE_KEEP_COUNT + 1);
-      // First item should be the prune marker
-      expect(result.current.history[0].type).toBe('info');
-      expect(result.current.history[0].text).toContain('pruned');
-      // Last item should be the most recent message
-      expect(
-        result.current.history[result.current.history.length - 1].text,
-      ).toBe(`Message ${itemCount - 1}`);
-    });
-
-    it('should be a no-op when history is under the threshold', () => {
-      const { result } = renderHook(() => useHistory());
-      const itemCount = 10;
-
-      act(() => {
-        for (let i = 0; i < itemCount; i++) {
-          result.current.addItem({
-            type: 'user',
-            text: `Message ${i}`,
-          });
-        }
-      });
-
-      const historyBefore = result.current.history;
-
-      act(() => {
-        result.current.pruneItems();
-      });
-
-      // Should be unchanged (same reference)
-      expect(result.current.history).toBe(historyBefore);
-      expect(result.current.history).toHaveLength(itemCount);
-    });
-
-    it('should be a no-op when history is exactly at the threshold', () => {
-      const { result } = renderHook(() => useHistory());
-
-      act(() => {
-        for (let i = 0; i < PRUNE_KEEP_COUNT; i++) {
-          result.current.addItem({
-            type: 'user',
-            text: `Message ${i}`,
-          });
-        }
-      });
-
-      const historyBefore = result.current.history;
-
-      act(() => {
-        result.current.pruneItems();
-      });
-
-      expect(result.current.history).toBe(historyBefore);
-      expect(result.current.history).toHaveLength(PRUNE_KEEP_COUNT);
-    });
-  });
 });

packages/cli/src/ui/hooks/useHistoryManager.ts

@@ -8,12 +8,6 @@ import { useState, useRef, useCallback, useMemo } from 'react';
 import type { HistoryItem } from '../types.js';
 import type { ChatRecordingService } from '@google/gemini-cli-core/src/services/chatRecordingService.js';
 
-/**
- * Number of history items to keep when pruning after context compression.
- * Exported for testing purposes.
- */
-export const PRUNE_KEEP_COUNT = 50;
-
 // Type for the updater function passed to updateHistoryItem
 type HistoryItemUpdater = (
   prevItem: HistoryItem,
@@ -32,7 +26,6 @@ export interface UseHistoryManagerReturn {
   ) => void;
   clearItems: () => void;
   loadHistory: (newHistory: HistoryItem[]) => void;
-  pruneItems: () => void;
 }
 
 /**
@@ -163,25 +156,6 @@ export function useHistory({
     messageIdCounterRef.current = 0;
   }, []);
 
-  // Prunes old history items, keeping only the most recent PRUNE_KEEP_COUNT.
-  // Intended to be called after context compression to free memory in
-  // long-running sessions.
-  const pruneItems = useCallback(() => {
-    setHistory((prevHistory) => {
-      if (prevHistory.length <= PRUNE_KEEP_COUNT) {
-        return prevHistory;
-      }
-      const kept = prevHistory.slice(-PRUNE_KEEP_COUNT);
-
-      const marker = {
-        id: getNextMessageId(Date.now()),
-        type: 'info',
-        text: `ℹ️  Earlier history was pruned after context compression.`,
-      } as HistoryItem;
-      return [marker, ...kept];
-    });
-  }, [getNextMessageId]);
-
   return useMemo(
     () => ({
       history,
@@ -189,8 +163,7 @@ export function useHistory({
       updateItem,
       clearItems,
       loadHistory,
-      pruneItems,
     }),
-    [history, addItem, updateItem, clearItems, loadHistory, pruneItems],
+    [history, addItem, updateItem, clearItems, loadHistory],
   );
 }

packages/cli/src/ui/hooks/useIncludeDirsTrust.test.tsx

@@ -91,7 +91,6 @@ describe('useIncludeDirsTrust', () => {
       updateItem: vi.fn(),
       clearItems: vi.fn(),
       loadHistory: vi.fn(),
-      pruneItems: vi.fn(),
     };
     mockSetCustomDialog = vi.fn();
   });

packages/cli/src/ui/hooks/useQuotaAndFallback.test.ts

@@ -85,7 +85,6 @@ describe('useQuotaAndFallback', () => {
       updateItem: vi.fn(),
       clearItems: vi.fn(),
       loadHistory: vi.fn(),
-      pruneItems: vi.fn(),
     };
     mockSetModelSwitchedFromQuotaError = vi.fn();
     mockOnShowAuthSelection = vi.fn();

packages/cli/src/ui/hooks/useSessionBrowser.ts

@@ -78,17 +78,12 @@ export const useSessionBrowser = (
 
           // We've loaded it; tell the UI about it.
           setIsSessionBrowserOpen(false);
-          const compressionIndex = conversation.lastCompressionIndex;
           const historyData = convertSessionToHistoryFormats(
             conversation.messages,
-            compressionIndex,
           );
           await onLoadHistory(
             historyData.uiHistory,
-            convertSessionToClientHistory(
-              conversation.messages,
-              compressionIndex,
-            ),
+            convertSessionToClientHistory(conversation.messages),
             resumedSessionData,
           );
         } catch (error) {

packages/cli/src/ui/hooks/useSessionResume.test.ts

@@ -33,7 +33,6 @@ describe('useSessionResume', () => {
     updateItem: vi.fn(),
     clearItems: vi.fn(),
     loadHistory: vi.fn(),
-    pruneItems: vi.fn(),
   });
 
   let mockHistoryManager: UseHistoryManagerReturn;
@@ -530,60 +529,5 @@ describe('useSessionResume', () => {
       // But UI history should have both
       expect(mockHistoryManager.addItem).toHaveBeenCalledTimes(2);
     });
-
-    it('should restore scheduled work from resumed session data', async () => {
-      const mockRestore = vi.fn();
-      const configWithScheduler = {
-        ...mockConfig,
-        getWorkScheduler: vi.fn().mockReturnValue({
-          restore: mockRestore,
-        }),
-      };
-
-      const scheduledWork = [
-        {
-          id: 'test-1',
-          prompt: 'check status',
-          fireAt: new Date(Date.now() + 60_000).toISOString(),
-          createdAt: new Date().toISOString(),
-        },
-      ];
-
-      const conversation: ConversationRecord = {
-        sessionId: 'auto-resume-scheduled',
-        projectHash: 'project-123',
-        startTime: '2025-01-01T00:00:00Z',
-        lastUpdated: '2025-01-01T01:00:00Z',
-        messages: [
-          {
-            id: 'msg-1',
-            timestamp: '2025-01-01T00:01:00Z',
-            content: 'Hello',
-            type: 'user',
-          },
-        ] as MessageRecord[],
-        scheduledWork,
-      };
-
-      await act(async () => {
-        renderHook(() =>
-          useSessionResume({
-            ...getDefaultProps(),
-            config: configWithScheduler as unknown as Config,
-            resumedSessionData: {
-              conversation,
-              filePath: '/path/to/session.json',
-            },
-          }),
-        );
-      });
-
-      await waitFor(() => {
-        expect(mockHistoryManager.clearItems).toHaveBeenCalled();
-      });
-
-      expect(configWithScheduler.getWorkScheduler).toHaveBeenCalled();
-      expect(mockRestore).toHaveBeenCalledWith(scheduledWork);
-    });
   });
 });

packages/cli/src/ui/hooks/useSessionResume.ts

@@ -83,16 +83,6 @@ export function useSessionResume({
           workspaceContext.addDirectories(resumedData.conversation.directories);
         }
 
-        // Restore scheduled work items from the resumed session.
-        // Past-due items fire immediately; future items get timers re-armed.
-        if (
-          resumedData.conversation.scheduledWork &&
-          resumedData.conversation.scheduledWork.length > 0
-        ) {
-          const scheduler = config.getWorkScheduler();
-          scheduler.restore(resumedData.conversation.scheduledWork);
-        }
-
         // Give the history to the Gemini client.
         await config.getGeminiClient()?.resumeChat(clientHistory, resumedData);
       } catch (error) {
@@ -119,18 +109,12 @@ export function useSessionResume({
       !hasLoadedResumedSession.current
     ) {
       hasLoadedResumedSession.current = true;
-      const compressionIndex =
-        resumedSessionData.conversation.lastCompressionIndex;
       const historyData = convertSessionToHistoryFormats(
         resumedSessionData.conversation.messages,
-        compressionIndex,
       );
       void loadHistoryForResume(
         historyData.uiHistory,
-        convertSessionToClientHistory(
-          resumedSessionData.conversation.messages,
-          compressionIndex,
-        ),
+        convertSessionToClientHistory(resumedSessionData.conversation.messages),
         resumedSessionData,
       );
     }

packages/cli/src/ui/utils/toolLayoutUtils.test.ts

@@ -0,0 +1,208 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import { describe, it, expect } from 'vitest';
+import {
+  calculateToolContentMaxLines,
+  calculateShellMaxLines,
+  SHELL_CONTENT_OVERHEAD,
+} from './toolLayoutUtils.js';
+import { CoreToolCallStatus } from '@google/gemini-cli-core';
+import {
+  ACTIVE_SHELL_MAX_LINES,
+  COMPLETED_SHELL_MAX_LINES,
+} from '../constants.js';
+
+describe('toolLayoutUtils', () => {
+  describe('calculateToolContentMaxLines', () => {
+    interface CalculateToolContentMaxLinesTestCase {
+      desc: string;
+      options: Parameters<typeof calculateToolContentMaxLines>[0];
+      expected: number | undefined;
+    }
+
+    const testCases: CalculateToolContentMaxLinesTestCase[] = [
+      {
+        desc: 'returns undefined if availableTerminalHeight is undefined',
+        options: {
+          availableTerminalHeight: undefined,
+          isAlternateBuffer: false,
+        },
+        expected: undefined,
+      },
+      {
+        desc: 'returns maxLinesLimit if maxLinesLimit applies but availableTerminalHeight is undefined',
+        options: {
+          availableTerminalHeight: undefined,
+          isAlternateBuffer: false,
+          maxLinesLimit: 10,
+        },
+        expected: 10,
+      },
+      {
+        desc: 'returns available space directly in constrained terminal (Standard mode)',
+        options: {
+          availableTerminalHeight: 2,
+          isAlternateBuffer: false,
+        },
+        expected: 3,
+      },
+      {
+        desc: 'returns available space directly in constrained terminal (ASB mode)',
+        options: {
+          availableTerminalHeight: 4,
+          isAlternateBuffer: true,
+        },
+        expected: 3,
+      },
+      {
+        desc: 'returns remaining space if sufficient space exists (Standard mode)',
+        options: {
+          availableTerminalHeight: 20,
+          isAlternateBuffer: false,
+        },
+        expected: 17,
+      },
+      {
+        desc: 'returns remaining space if sufficient space exists (ASB mode)',
+        options: {
+          availableTerminalHeight: 20,
+          isAlternateBuffer: true,
+        },
+        expected: 13,
+      },
+    ];
+
+    it.each(testCases)('$desc', ({ options, expected }) => {
+      const result = calculateToolContentMaxLines(options);
+      expect(result).toBe(expected);
+    });
+  });
+
+  describe('calculateShellMaxLines', () => {
+    interface CalculateShellMaxLinesTestCase {
+      desc: string;
+      options: Parameters<typeof calculateShellMaxLines>[0];
+      expected: number | undefined;
+    }
+
+    const testCases: CalculateShellMaxLinesTestCase[] = [
+      {
+        desc: 'returns undefined when not constrained and is expandable',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: false,
+          isThisShellFocused: false,
+          availableTerminalHeight: 20,
+          constrainHeight: false,
+          isExpandable: true,
+        },
+        expected: undefined,
+      },
+      {
+        desc: 'returns ACTIVE_SHELL_MAX_LINES - SHELL_CONTENT_OVERHEAD for ASB mode when availableTerminalHeight is undefined',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: true,
+          isThisShellFocused: false,
+          availableTerminalHeight: undefined,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: ACTIVE_SHELL_MAX_LINES - SHELL_CONTENT_OVERHEAD,
+      },
+      {
+        desc: 'returns undefined for Standard mode when availableTerminalHeight is undefined',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: false,
+          isThisShellFocused: false,
+          availableTerminalHeight: undefined,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: undefined,
+      },
+      {
+        desc: 'handles small availableTerminalHeight gracefully without overflow in Standard mode',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: false,
+          isThisShellFocused: false,
+          availableTerminalHeight: 2,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: 1,
+      },
+      {
+        desc: 'handles small availableTerminalHeight gracefully without overflow in ASB mode',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: true,
+          isThisShellFocused: false,
+          availableTerminalHeight: 6,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: 4,
+      },
+      {
+        desc: 'handles negative availableTerminalHeight gracefully',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: false,
+          isThisShellFocused: false,
+          availableTerminalHeight: -5,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: 1,
+      },
+      {
+        desc: 'returns maxLinesBasedOnHeight for focused ASB shells',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: true,
+          isThisShellFocused: true,
+          availableTerminalHeight: 30,
+          constrainHeight: false,
+          isExpandable: false,
+        },
+        expected: 28,
+      },
+      {
+        desc: 'falls back to COMPLETED_SHELL_MAX_LINES - SHELL_CONTENT_OVERHEAD for completed shells if space allows',
+        options: {
+          status: CoreToolCallStatus.Success,
+          isAlternateBuffer: false,
+          isThisShellFocused: false,
+          availableTerminalHeight: 100,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: COMPLETED_SHELL_MAX_LINES - SHELL_CONTENT_OVERHEAD,
+      },
+      {
+        desc: 'falls back to ACTIVE_SHELL_MAX_LINES - SHELL_CONTENT_OVERHEAD for executing shells if space allows',
+        options: {
+          status: CoreToolCallStatus.Executing,
+          isAlternateBuffer: false,
+          isThisShellFocused: false,
+          availableTerminalHeight: 100,
+          constrainHeight: true,
+          isExpandable: false,
+        },
+        expected: ACTIVE_SHELL_MAX_LINES - SHELL_CONTENT_OVERHEAD,
+      },
+    ];
+
+    it.each(testCases)('$desc', ({ options, expected }) => {
+      const result = calculateShellMaxLines(options);
+      expect(result).toBe(expected);
+    });
+  });
+});

packages/cli/src/ui/utils/toolLayoutUtils.ts

@@ -46,12 +46,13 @@ export function calculateToolContentMaxLines(options: {
     ? TOOL_RESULT_ASB_RESERVED_LINE_COUNT
     : TOOL_RESULT_STANDARD_RESERVED_LINE_COUNT;
 
-  let contentHeight = availableTerminalHeight
-    ? Math.max(
-        availableTerminalHeight - TOOL_RESULT_STATIC_HEIGHT - reservedLines,
-        TOOL_RESULT_MIN_LINES_SHOWN + 1,
-      )
-    : undefined;
+  let contentHeight =
+    availableTerminalHeight !== undefined
+      ? Math.max(
+          availableTerminalHeight - TOOL_RESULT_STATIC_HEIGHT - reservedLines,
+          TOOL_RESULT_MIN_LINES_SHOWN + 1,
+        )
+      : undefined;
 
   if (maxLinesLimit !== undefined) {
     contentHeight =
@@ -100,7 +101,10 @@ export function calculateShellMaxLines(options: {
       : undefined;
   }
 
-  const maxLinesBasedOnHeight = Math.max(1, availableTerminalHeight - 2);
+  const maxLinesBasedOnHeight = Math.max(
+    1,
+    availableTerminalHeight - TOOL_RESULT_STANDARD_RESERVED_LINE_COUNT,
+  );
 
   // 3. Handle ASB mode focus expansion.
   // We allow a focused shell in ASB mode to take up the full available height,

packages/cli/src/utils/events.ts

@@ -23,8 +23,6 @@ export enum AppEvent {
   PasteTimeout = 'paste-timeout',
   TerminalBackground = 'terminal-background',
   TransientMessage = 'transient-message',
-  ExternalMessage = 'external-message',
-  A2AListenerStarted = 'a2a-listener-started',
 }
 
 export interface AppEvents {
@@ -34,8 +32,6 @@ export interface AppEvents {
   [AppEvent.PasteTimeout]: never[];
   [AppEvent.TerminalBackground]: [string];
   [AppEvent.TransientMessage]: [TransientMessagePayload];
-  [AppEvent.ExternalMessage]: [string];
-  [AppEvent.A2AListenerStarted]: [number];
 }
 
 export const appEvents = new EventEmitter<AppEvents>();

packages/cli/src/utils/handleAutoUpdate.test.ts

@@ -202,7 +202,12 @@ describe('handleAutoUpdate', () => {
     expect(mockSpawn).not.toHaveBeenCalled();
   });
 
-  it.each([PackageManager.NPX, PackageManager.PNPX, PackageManager.BUNX])(
+  it.each([
+    PackageManager.NPX,
+    PackageManager.PNPX,
+    PackageManager.BUNX,
+    PackageManager.BINARY,
+  ])(
     'should suppress update notifications when running via %s',
     (packageManager) => {
       mockGetInstallationInfo.mockReturnValue({

packages/cli/src/utils/handleAutoUpdate.ts

@@ -87,9 +87,12 @@ export function handleAutoUpdate(
   );
 
   if (
-    [PackageManager.NPX, PackageManager.PNPX, PackageManager.BUNX].includes(
-      installationInfo.packageManager,
-    )
+    [
+      PackageManager.NPX,
+      PackageManager.PNPX,
+      PackageManager.BUNX,
+      PackageManager.BINARY,
+    ].includes(installationInfo.packageManager)
   ) {
     return;
   }

packages/cli/src/utils/installationInfo.test.ts

@@ -58,6 +58,19 @@ describe('getInstallationInfo', () => {
     process.argv = originalArgv;
   });
 
+  it('should detect running as a standalone binary', () => {
+    vi.stubEnv('IS_BINARY', 'true');
+    process.argv[1] = '/path/to/binary';
+    const info = getInstallationInfo(projectRoot, true);
+    expect(info.packageManager).toBe(PackageManager.BINARY);
+    expect(info.isGlobal).toBe(true);
+    expect(info.updateMessage).toBe(
+      'Running as a standalone binary. Please update by downloading the latest version from GitHub.',
+    );
+    expect(info.updateCommand).toBeUndefined();
+    vi.unstubAllEnvs();
+  });
+
   it('should return UNKNOWN when cliPath is not available', () => {
     process.argv[1] = '';
     const info = getInstallationInfo(projectRoot, true);

packages/cli/src/utils/installationInfo.ts

@@ -21,6 +21,7 @@ export enum PackageManager {
   BUNX = 'bunx',
   HOMEBREW = 'homebrew',
   NPX = 'npx',
+  BINARY = 'binary',
   UNKNOWN = 'unknown',
 }
 
@@ -41,6 +42,16 @@ export function getInstallationInfo(
   }
 
   try {
+    // Check for standalone binary first
+    if (process.env['IS_BINARY'] === 'true') {
+      return {
+        packageManager: PackageManager.BINARY,
+        isGlobal: true,
+        updateMessage:
+          'Running as a standalone binary. Please update by downloading the latest version from GitHub.',
+      };
+    }
+
     // Normalize path separators to forward slashes for consistent matching.
     const realPath = fs.realpathSync(cliPath).replace(/\\/g, '/');
     const normalizedProjectRoot = projectRoot?.replace(/\\/g, '/');

packages/cli/src/utils/sessionUtils.ts

@@ -539,24 +539,12 @@ export class SessionSelector {
  */
 export function convertSessionToHistoryFormats(
   messages: ConversationRecord['messages'],
-  startIndex?: number,
 ): {
   uiHistory: HistoryItemWithoutId[];
 } {
   const uiHistory: HistoryItemWithoutId[] = [];
 
-  const hasCompressedHistory =
-    startIndex != null && startIndex > 0 && startIndex < messages.length;
-  const slice = hasCompressedHistory ? messages.slice(startIndex) : messages;
-
-  if (hasCompressedHistory) {
-    uiHistory.push({
-      type: MessageType.INFO,
-      text: `ℹ️  Earlier history (${startIndex} messages) was compressed. Showing post-compression messages only.`,
-    });
-  }
-
-  for (const msg of slice) {
+  for (const msg of messages) {
     // Add thoughts if present
     if (msg.type === 'gemini' && msg.thoughts && msg.thoughts.length > 0) {
       for (const thought of msg.thoughts) {

packages/core/src/agents/browser/browserManager.test.ts

@@ -44,6 +44,11 @@ vi.mock('../../utils/debugLogger.js', () => ({
   },
 }));
 
+// Mock browser consent to always grant consent by default
+vi.mock('../../utils/browserConsent.js', () => ({
+  getBrowserConsentIfNeeded: vi.fn().mockResolvedValue(true),
+}));
+
 vi.mock('./automationOverlay.js', () => ({
   injectAutomationOverlay: vi.fn().mockResolvedValue(undefined),
 }));
@@ -64,6 +69,7 @@ vi.mock('node:fs', async (importOriginal) => {
 import * as fs from 'node:fs';
 import { Client } from '@modelcontextprotocol/sdk/client/index.js';
 import { StdioClientTransport } from '@modelcontextprotocol/sdk/client/stdio.js';
+import { getBrowserConsentIfNeeded } from '../../utils/browserConsent.js';
 
 describe('BrowserManager', () => {
   let mockConfig: Config;
@@ -72,6 +78,9 @@ describe('BrowserManager', () => {
     vi.resetAllMocks();
     vi.mocked(injectAutomationOverlay).mockClear();
 
+    // Re-establish consent mock after resetAllMocks
+    vi.mocked(getBrowserConsentIfNeeded).mockResolvedValue(true);
+
     // Setup mock config
     mockConfig = makeFakeConfig({
       agents: {
@@ -527,6 +536,41 @@ describe('BrowserManager', () => {
         /sessionMode: persistent/,
       );
     });
+
+    it('should pass --no-usage-statistics and --no-performance-crux when privacy is disabled', async () => {
+      const privacyDisabledConfig = makeFakeConfig({
+        agents: {
+          overrides: {
+            browser_agent: {
+              enabled: true,
+            },
+          },
+          browser: {
+            headless: false,
+          },
+        },
+        usageStatisticsEnabled: false,
+      });
+
+      const manager = new BrowserManager(privacyDisabledConfig);
+      await manager.ensureConnection();
+
+      const args = vi.mocked(StdioClientTransport).mock.calls[0]?.[0]
+        ?.args as string[];
+      expect(args).toContain('--no-usage-statistics');
+      expect(args).toContain('--no-performance-crux');
+    });
+
+    it('should NOT pass privacy flags when usage statistics are enabled', async () => {
+      // Default config has usageStatisticsEnabled: true (or undefined)
+      const manager = new BrowserManager(mockConfig);
+      await manager.ensureConnection();
+
+      const args = vi.mocked(StdioClientTransport).mock.calls[0]?.[0]
+        ?.args as string[];
+      expect(args).not.toContain('--no-usage-statistics');
+      expect(args).not.toContain('--no-performance-crux');
+    });
   });
 
   describe('MCP isolation', () => {

packages/core/src/agents/browser/browserManager.ts

@@ -23,6 +23,7 @@ import type { Tool as McpTool } from '@modelcontextprotocol/sdk/types.js';
 import { debugLogger } from '../../utils/debugLogger.js';
 import type { Config } from '../../config/config.js';
 import { Storage } from '../../config/storage.js';
+import { getBrowserConsentIfNeeded } from '../../utils/browserConsent.js';
 import { injectInputBlocker } from './inputBlocker.js';
 import * as path from 'node:path';
 import * as fs from 'node:fs';
@@ -260,6 +261,16 @@ export class BrowserManager {
     if (this.rawMcpClient) {
       return;
     }
+
+    // Request browser consent if needed (first-run privacy notice)
+    const consentGranted = await getBrowserConsentIfNeeded();
+    if (!consentGranted) {
+      throw new Error(
+        'Browser agent requires user consent to proceed. ' +
+          'Please re-run and accept the privacy notice.',
+      );
+    }
+
     await this.connectMcp();
   }
 
@@ -352,6 +363,11 @@ export class BrowserManager {
       mcpArgs.push('--userDataDir', defaultProfilePath);
     }
 
+    // Respect the user's privacy.usageStatisticsEnabled setting
+    if (!this.config.getUsageStatisticsEnabled()) {
+      mcpArgs.push('--no-usage-statistics', '--no-performance-crux');
+    }
+
     if (
       browserConfig.customConfig.allowedDomains &&
       browserConfig.customConfig.allowedDomains.length > 0

packages/core/src/agents/local-executor.test.ts

@@ -91,9 +91,18 @@ import {
   type LocalAgentDefinition,
   type SubagentActivityEvent,
   type OutputConfig,
+  SubagentActivityErrorType,
 } from './types.js';
-import type { AnyDeclarativeTool, AnyToolInvocation } from '../tools/tools.js';
-import type { ToolCallRequestInfo } from '../scheduler/types.js';
+import {
+  ToolConfirmationOutcome,
+  type AnyDeclarativeTool,
+  type AnyToolInvocation,
+} from '../tools/tools.js';
+import {
+  type ToolCallRequestInfo,
+  CoreToolCallStatus,
+} from '../scheduler/types.js';
+
 import { CompressionStatus } from '../core/turn.js';
 import { ChatCompressionService } from '../services/chatCompressionService.js';
 import type {
@@ -365,6 +374,76 @@ describe('LocalAgentExecutor', () => {
   });
 
   describe('create (Initialization and Validation)', () => {
+    it('should explicitly map execution context properties to prevent unintended propagation', async () => {
+      const definition = createTestDefinition([LS_TOOL_NAME]);
+      const mockGeminiClient =
+        {} as unknown as import('../core/client.js').GeminiClient;
+      const mockSandboxManager =
+        {} as unknown as import('../services/sandboxManager.js').SandboxManager;
+      const extendedContext = {
+        config: mockConfig,
+        promptId: mockConfig.promptId,
+        toolRegistry: parentToolRegistry,
+        promptRegistry: mockConfig.promptRegistry,
+        resourceRegistry: mockConfig.resourceRegistry,
+        messageBus: mockConfig.messageBus,
+        geminiClient: mockGeminiClient,
+        sandboxManager: mockSandboxManager,
+        unintendedProperty: 'should not be here',
+      } as unknown as import('../config/agent-loop-context.js').AgentLoopContext;
+
+      const executor = await LocalAgentExecutor.create(
+        definition,
+        extendedContext,
+        onActivity,
+      );
+
+      mockModelResponse([
+        {
+          name: TASK_COMPLETE_TOOL_NAME,
+          args: { finalResult: 'done' },
+          id: 'call1',
+        },
+      ]);
+
+      await executor.run({ goal: 'test' }, signal);
+
+      const chatConstructorArgs = MockedGeminiChat.mock.calls[0];
+      const executionContext = chatConstructorArgs[0];
+
+      expect(executionContext).toBeDefined();
+      expect(executionContext.config).toBe(extendedContext.config);
+      expect(executionContext.promptId).toBe(extendedContext.promptId);
+      expect(executionContext.geminiClient).toBe(extendedContext.geminiClient);
+      expect(executionContext.sandboxManager).toBe(
+        extendedContext.sandboxManager,
+      );
+
+      const agentToolRegistry = executor['toolRegistry'];
+      const agentPromptRegistry = executor['promptRegistry'];
+      const agentResourceRegistry = executor['resourceRegistry'];
+
+      expect(executionContext.toolRegistry).toBe(agentToolRegistry);
+      expect(executionContext.promptRegistry).toBe(agentPromptRegistry);
+      expect(executionContext.resourceRegistry).toBe(agentResourceRegistry);
+
+      expect(executionContext.messageBus).toBe(
+        agentToolRegistry.getMessageBus(),
+      );
+
+      // Ensure the unintended property was not spread
+      expect(
+        (executionContext as unknown as { unintendedProperty?: string })
+          .unintendedProperty,
+      ).toBeUndefined();
+
+      // Ensure registries and message bus are not the parent's
+      expect(executionContext.toolRegistry).not.toBe(
+        extendedContext.toolRegistry,
+      );
+      expect(executionContext.messageBus).not.toBe(extendedContext.messageBus);
+    });
+
     it('should create successfully with allowed tools', async () => {
       const definition = createTestDefinition([LS_TOOL_NAME]);
       const executor = await LocalAgentExecutor.create(
@@ -943,6 +1022,7 @@ describe('LocalAgentExecutor', () => {
           data: expect.objectContaining({
             context: 'protocol_violation',
             error: expectedError,
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -988,6 +1068,7 @@ describe('LocalAgentExecutor', () => {
             context: 'tool_call',
             name: TASK_COMPLETE_TOOL_NAME,
             error: expectedError,
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1091,7 +1172,7 @@ describe('LocalAgentExecutor', () => {
               if (callsStarted === 2) resolveCalls();
               await vi.advanceTimersByTimeAsync(100);
               return {
-                status: 'success',
+                status: CoreToolCallStatus.Success,
                 request: reqInfo,
                 tool: {} as AnyDeclarativeTool,
                 invocation: {} as AnyToolInvocation,
@@ -1109,7 +1190,7 @@ describe('LocalAgentExecutor', () => {
                   ],
                   error: undefined,
                   errorType: undefined,
-                  contentLength: undefined,
+                  contentLength: 0,
                 },
               };
             }),
@@ -1147,10 +1228,10 @@ describe('LocalAgentExecutor', () => {
       expect(parts).toEqual(
         expect.arrayContaining([
           expect.objectContaining({
-            functionResponse: expect.objectContaining({ id: 'c1' }),
+            functionResponse: expect.objectContaining({ name: LS_TOOL_NAME }),
           }),
           expect.objectContaining({
-            functionResponse: expect.objectContaining({ id: 'c2' }),
+            functionResponse: expect.objectContaining({ name: LS_TOOL_NAME }),
           }),
         ]),
       );
@@ -1221,6 +1302,7 @@ describe('LocalAgentExecutor', () => {
           data: expect.objectContaining({
             context: 'tool_call_unauthorized',
             name: READ_FILE_TOOL_NAME,
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1274,6 +1356,7 @@ describe('LocalAgentExecutor', () => {
             context: 'tool_call',
             name: TASK_COMPLETE_TOOL_NAME,
             error: expect.stringContaining('Output validation failed'),
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1320,6 +1403,7 @@ describe('LocalAgentExecutor', () => {
           type: 'ERROR',
           data: expect.objectContaining({
             error: `Error: Failed to create chat object: ${getErrorMessage(initError)}`,
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1348,7 +1432,7 @@ describe('LocalAgentExecutor', () => {
       ]);
       mockScheduleAgentTools.mockResolvedValueOnce([
         {
-          status: 'error',
+          status: CoreToolCallStatus.Error,
           request: {
             callId: 'call1',
             name: LS_TOOL_NAME,
@@ -1399,6 +1483,7 @@ describe('LocalAgentExecutor', () => {
             context: 'tool_call',
             name: LS_TOOL_NAME,
             error: toolErrorMessage,
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1421,6 +1506,157 @@ describe('LocalAgentExecutor', () => {
       expect(output.terminate_reason).toBe(AgentTerminateMode.GOAL);
       expect(output.result).toBe('Aborted due to tool failure.');
     });
+
+    it('should handle a soft tool rejection (outcome: Cancel) and provide direct instructions to the model', async () => {
+      const definition = createTestDefinition([LS_TOOL_NAME]);
+      const executor = await LocalAgentExecutor.create(
+        definition,
+        mockConfig,
+        onActivity,
+      );
+
+      // Turn 1: Model calls a tool that will be rejected
+      mockModelResponse([
+        { name: LS_TOOL_NAME, args: { path: '/secret' }, id: 'call1' },
+      ]);
+      mockScheduleAgentTools.mockResolvedValueOnce([
+        {
+          status: 'cancelled',
+          request: {
+            callId: 'call1',
+            name: LS_TOOL_NAME,
+            args: { path: '/secret' },
+            isClientInitiated: false,
+            prompt_id: 'test-prompt',
+          },
+          tool: {} as AnyDeclarativeTool,
+          invocation: {} as AnyToolInvocation,
+          outcome: ToolConfirmationOutcome.Cancel, // Soft rejection
+          response: {
+            callId: 'call1',
+            resultDisplay: '',
+            responseParts: [
+              {
+                functionResponse: {
+                  name: LS_TOOL_NAME,
+                  response: {
+                    error:
+                      '[Operation Cancelled] Reason: User denied execution.',
+                  },
+                  id: 'call1',
+                },
+              },
+            ],
+            error: undefined,
+            errorType: undefined,
+            contentLength: 0,
+          },
+        },
+      ]);
+
+      // Turn 2: Model sees the rejection + consolidated instructions and completes
+      mockModelResponse([
+        {
+          name: TASK_COMPLETE_TOOL_NAME,
+          args: { finalResult: 'User rejected access to /secret.' },
+          id: 'call2',
+        },
+      ]);
+
+      const output = await executor.run(
+        { goal: 'Soft rejection test' },
+        signal,
+      );
+
+      // Verify the activity stream reported the consolidated instruction
+      expect(activities).toContainEqual(
+        expect.objectContaining({
+          type: 'ERROR',
+          data: expect.objectContaining({
+            context: 'tool_call',
+            name: LS_TOOL_NAME,
+            error: expect.stringContaining('User rejected this operation'),
+            errorType: SubagentActivityErrorType.REJECTED,
+          }),
+        }),
+      );
+
+      // Verify the instruction was sent back to the model as the tool error
+      const turn2Params = getMockMessageParams(1);
+      const parts = turn2Params.message as Part[];
+      const errorMsg = parts[0].functionResponse?.response?.['error'];
+      expect(typeof errorMsg).toBe('string');
+      if (typeof errorMsg === 'string') {
+        expect(errorMsg).toContain('User rejected this operation');
+        expect(errorMsg).toContain('acknowledge this, rethink your strategy');
+      }
+
+      expect(output.terminate_reason).toBe(AgentTerminateMode.GOAL);
+      expect(output.result).toBe('User rejected access to /secret.');
+    });
+
+    it('should handle a hard tool abort (cancelled with no outcome) and terminate the agent', async () => {
+      const definition = createTestDefinition([LS_TOOL_NAME]);
+      const executor = await LocalAgentExecutor.create(
+        definition,
+        mockConfig,
+        onActivity,
+      );
+
+      // Turn 1: Model calls a tool that will be aborted (e.g. Ctrl+C)
+      mockModelResponse([
+        { name: LS_TOOL_NAME, args: { path: '/secret' }, id: 'call1' },
+      ]);
+      mockScheduleAgentTools.mockResolvedValueOnce([
+        {
+          status: 'cancelled',
+          request: {
+            callId: 'call1',
+            name: LS_TOOL_NAME,
+            args: { path: '/secret' },
+            isClientInitiated: false,
+            prompt_id: 'test-prompt',
+          },
+          tool: {} as AnyDeclarativeTool,
+          invocation: {} as AnyToolInvocation,
+          outcome: undefined, // Hard abort
+          response: {
+            callId: 'call1',
+            resultDisplay: '',
+            responseParts: [
+              {
+                functionResponse: {
+                  name: LS_TOOL_NAME,
+                  response: { error: 'Request cancelled.' },
+                  id: 'call1',
+                },
+              },
+            ],
+            error: undefined,
+            errorType: undefined,
+            contentLength: 0,
+          },
+        },
+      ]);
+
+      const output = await executor.run({ goal: 'Hard abort test' }, signal);
+
+      // Verify the activity stream reported the cancellation
+      expect(activities).toContainEqual(
+        expect.objectContaining({
+          type: 'ERROR',
+          data: expect.objectContaining({
+            context: 'tool_call',
+            name: LS_TOOL_NAME,
+            error: 'Request cancelled.',
+            errorType: SubagentActivityErrorType.CANCELLED,
+          }),
+        }),
+      );
+
+      // Agent should terminate with ABORTED status
+      expect(output.terminate_reason).toBe(AgentTerminateMode.ABORTED);
+    });
   });
 
   describe('Model Routing', () => {
@@ -1615,6 +1851,7 @@ describe('LocalAgentExecutor', () => {
           data: expect.objectContaining({
             context: 'timeout',
             error: 'Agent timed out after 0.5 minutes.',
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1803,6 +2040,7 @@ describe('LocalAgentExecutor', () => {
           data: expect.objectContaining({
             context: 'recovery_turn',
             error: 'Graceful recovery attempt failed. Reason: stop',
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -1886,6 +2124,7 @@ describe('LocalAgentExecutor', () => {
           data: expect.objectContaining({
             context: 'recovery_turn',
             error: 'Graceful recovery attempt failed. Reason: stop',
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );
@@ -2007,6 +2246,7 @@ describe('LocalAgentExecutor', () => {
           data: expect.objectContaining({
             context: 'recovery_turn',
             error: 'Graceful recovery attempt failed. Reason: stop',
+            errorType: SubagentActivityErrorType.GENERIC,
           }),
         }),
       );

packages/core/src/agents/local-executor.ts

@@ -18,7 +18,10 @@ import {
 import { ToolRegistry } from '../tools/tool-registry.js';
 import { PromptRegistry } from '../prompts/prompt-registry.js';
 import { ResourceRegistry } from '../resources/resource-registry.js';
-import { type AnyDeclarativeTool } from '../tools/tools.js';
+import {
+  type AnyDeclarativeTool,
+  ToolConfirmationOutcome,
+} from '../tools/tools.js';
 import {
   DiscoveredMCPTool,
   isMcpToolName,
@@ -46,6 +49,9 @@ import {
   DEFAULT_QUERY_STRING,
   DEFAULT_MAX_TURNS,
   DEFAULT_MAX_TIME_MINUTES,
+  SubagentActivityErrorType,
+  SUBAGENT_REJECTED_ERROR_PREFIX,
+  SUBAGENT_CANCELLED_ERROR_MESSAGE,
   type LocalAgentDefinition,
   type AgentInputs,
   type OutputObject,
@@ -113,7 +119,10 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
 
   private get executionContext(): AgentLoopContext {
     return {
-      ...this.context,
+      config: this.context.config,
+      promptId: this.context.promptId,
+      geminiClient: this.context.geminiClient,
+      sandboxManager: this.context.sandboxManager,
       toolRegistry: this.toolRegistry,
       promptRegistry: this.promptRegistry,
       resourceRegistry: this.resourceRegistry,
@@ -335,6 +344,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
       this.emitActivity('ERROR', {
         error: `Agent stopped calling tools but did not call '${TASK_COMPLETE_TOOL_NAME}' to finalize the session.`,
         context: 'protocol_violation',
+        errorType: SubagentActivityErrorType.GENERIC,
       });
       return {
         status: 'stop',
@@ -468,6 +478,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
       this.emitActivity('ERROR', {
         error: `Graceful recovery attempt failed. Reason: ${turnResult.status}`,
         context: 'recovery_turn',
+        errorType: SubagentActivityErrorType.GENERIC,
       });
       return null;
     } catch (error) {
@@ -475,6 +486,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
       this.emitActivity('ERROR', {
         error: `Graceful recovery attempt failed: ${String(error)}`,
         context: 'recovery_turn',
+        errorType: SubagentActivityErrorType.GENERIC,
       });
       return null;
     } finally {
@@ -680,12 +692,14 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
             this.emitActivity('ERROR', {
               error: finalResult,
               context: 'timeout',
+              errorType: SubagentActivityErrorType.GENERIC,
             });
           } else if (terminateReason === AgentTerminateMode.MAX_TURNS) {
             finalResult = `Agent reached max turns limit (${maxTurns}).`;
             this.emitActivity('ERROR', {
               error: finalResult,
               context: 'max_turns',
+              errorType: SubagentActivityErrorType.GENERIC,
             });
           } else if (
             terminateReason === AgentTerminateMode.ERROR_NO_COMPLETE_TASK_CALL
@@ -697,6 +711,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
             this.emitActivity('ERROR', {
               error: finalResult,
               context: 'protocol_violation',
+              errorType: SubagentActivityErrorType.GENERIC,
             });
           }
         }
@@ -751,6 +766,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
         this.emitActivity('ERROR', {
           error: finalResult,
           context: 'timeout',
+          errorType: SubagentActivityErrorType.GENERIC,
         });
         return {
           result: finalResult,
@@ -758,7 +774,10 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
         };
       }
 
-      this.emitActivity('ERROR', { error: String(error) });
+      this.emitActivity('ERROR', {
+        error: String(error),
+        errorType: SubagentActivityErrorType.GENERIC,
+      });
       throw error; // Re-throw other errors or external aborts.
     } finally {
       deadlineTimer.abort();
@@ -1027,6 +1046,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
             context: 'tool_call',
             name: toolName,
             error,
+            errorType: SubagentActivityErrorType.GENERIC,
           });
           continue;
         }
@@ -1054,6 +1074,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
                 context: 'tool_call',
                 name: toolName,
                 error,
+                errorType: SubagentActivityErrorType.GENERIC,
               });
               continue;
             }
@@ -1096,6 +1117,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
               name: toolName,
               callId,
               error,
+              errorType: SubagentActivityErrorType.GENERIC,
             });
           }
         } else {
@@ -1139,6 +1161,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
               name: toolName,
               callId,
               error,
+              errorType: SubagentActivityErrorType.GENERIC,
             });
           }
         }
@@ -1163,6 +1186,7 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
           name: toolName,
           callId,
           error,
+          errorType: SubagentActivityErrorType.GENERIC,
         });
 
         continue;
@@ -1210,18 +1234,46 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
             name: toolName,
             callId: call.request.callId,
             error: call.response.error?.message || 'Unknown error',
+            errorType: SubagentActivityErrorType.GENERIC,
           });
         } else if (call.status === 'cancelled') {
-          this.emitActivity('ERROR', {
-            context: 'tool_call',
-            name: toolName,
-            callId: call.request.callId,
-            error: 'Request cancelled.',
-          });
-          aborted = true;
+          const isSoftRejection =
+            call.outcome === ToolConfirmationOutcome.Cancel;
+
+          if (isSoftRejection) {
+            const error = `${SUBAGENT_REJECTED_ERROR_PREFIX} Please acknowledge this, rethink your strategy, and try a different approach. If you cannot proceed without the rejected operation, summarize the issue and use \`${TASK_COMPLETE_TOOL_NAME}\` to report your findings and the blocker.`;
+            this.emitActivity('ERROR', {
+              context: 'tool_call',
+              name: toolName,
+              callId: call.request.callId,
+              error,
+              errorType: SubagentActivityErrorType.REJECTED,
+            });
+            // Soft rejection: we do NOT set aborted=true, allowing the agent to rethink.
+
+            // Provide the direct instruction to the model as the tool error response.
+            syncResults.set(call.request.callId, {
+              functionResponse: {
+                name: toolName,
+                id: call.request.callId,
+                response: { error },
+              },
+            });
+            continue; // Skip the generic syncResults.set below
+          } else {
+            // Hard abort (Ctrl+C)
+            this.emitActivity('ERROR', {
+              context: 'tool_call',
+              name: toolName,
+              callId: call.request.callId,
+              error: SUBAGENT_CANCELLED_ERROR_MESSAGE,
+              errorType: SubagentActivityErrorType.CANCELLED,
+            });
+            aborted = true;
+          }
         }
 
-        // Add result to syncResults to preserve order later
+        // Add result to syncResults for other statuses (success, error, hard abort)
         syncResults.set(call.request.callId, call.response.responseParts[0]);
       }
     }
@@ -1332,7 +1384,8 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
 Important Rules:
 * You are running in a non-interactive mode. You CANNOT ask the user for input or clarification.
 * Work systematically using available tools to complete your task.
-* Always use absolute paths for file operations. Construct them using the provided "Environment Context".`;
+* Always use absolute paths for file operations. Construct them using the provided "Environment Context".
+* If a tool call is rejected by the user, acknowledge the rejection, rethink your strategy, and try a different approach. Do not repeatedly attempt the same rejected operation.`;
 
     if (this.definition.outputConfig) {
       finalPrompt += `

packages/core/src/agents/local-invocation.test.ts

@@ -19,6 +19,8 @@ import {
   type SubagentActivityEvent,
   type AgentInputs,
   type SubagentProgress,
+  SubagentActivityErrorType,
+  SUBAGENT_REJECTED_ERROR_PREFIX,
 } from './types.js';
 import { LocalSubagentInvocation } from './local-invocation.js';
 import { LocalAgentExecutor } from './local-executor.js';
@@ -230,7 +232,7 @@ describe('LocalSubagentInvocation', () => {
       expect(display.terminateReason).toBe(AgentTerminateMode.TIMEOUT);
     });
 
-    it('should stream THOUGHT_CHUNK activities from the executor', async () => {
+    it('should stream THOUGHT_CHUNK activities from the executor, replacing the last running thought', async () => {
       mockExecutorInstance.run.mockImplementation(async () => {
         const onActivity = MockLocalAgentExecutor.create.mock.calls[0][2];
 
@@ -245,7 +247,7 @@ describe('LocalSubagentInvocation', () => {
             isSubagentActivityEvent: true,
             agentName: 'MockAgent',
             type: 'THOUGHT_CHUNK',
-            data: { text: ' Still thinking.' },
+            data: { text: 'Thinking about next steps.' },
           } as SubagentActivityEvent);
         }
         return { result: 'Done', terminate_reason: AgentTerminateMode.GOAL };
@@ -258,7 +260,13 @@ describe('LocalSubagentInvocation', () => {
       expect(lastCall.recentActivity).toContainEqual(
         expect.objectContaining({
           type: 'thought',
-          content: 'Analyzing... Still thinking.',
+          content: 'Thinking about next steps.',
+        }),
+      );
+      expect(lastCall.recentActivity).not.toContainEqual(
+        expect.objectContaining({
+          type: 'thought',
+          content: 'Analyzing...',
         }),
       );
     });
@@ -297,6 +305,48 @@ describe('LocalSubagentInvocation', () => {
       );
     });
 
+    it('should reflect tool rejections in the activity stream as cancelled but not abort the agent', async () => {
+      mockExecutorInstance.run.mockImplementation(async () => {
+        const onActivity = MockLocalAgentExecutor.create.mock.calls[0][2];
+
+        if (onActivity) {
+          onActivity({
+            isSubagentActivityEvent: true,
+            agentName: 'MockAgent',
+            type: 'TOOL_CALL_START',
+            data: { name: 'ls', args: {}, callId: 'call1' },
+          } as SubagentActivityEvent);
+          onActivity({
+            isSubagentActivityEvent: true,
+            agentName: 'MockAgent',
+            type: 'ERROR',
+            data: {
+              name: 'ls',
+              callId: 'call1',
+              error: `${SUBAGENT_REJECTED_ERROR_PREFIX} Please acknowledge this, rethink your strategy, and try a different approach. If you cannot proceed without the rejected operation, summarize the issue and use \`complete_task\` to report your findings and the blocker.`,
+              errorType: SubagentActivityErrorType.REJECTED,
+            },
+          } as SubagentActivityEvent);
+        }
+        return {
+          result: 'Rethinking...',
+          terminate_reason: AgentTerminateMode.GOAL,
+        };
+      });
+
+      await invocation.execute(signal, updateOutput);
+
+      expect(updateOutput).toHaveBeenCalledTimes(4);
+      const lastCall = updateOutput.mock.calls[3][0] as SubagentProgress;
+      expect(lastCall.recentActivity).toContainEqual(
+        expect.objectContaining({
+          type: 'tool_call',
+          content: 'ls',
+          status: 'cancelled',
+        }),
+      );
+    });
+
     it('should run successfully without an updateOutput callback', async () => {
       mockExecutorInstance.run.mockImplementation(async () => {
         const onActivity = MockLocalAgentExecutor.create.mock.calls[0][2];

packages/core/src/agents/local-invocation.ts

@@ -18,6 +18,9 @@ import {
   type SubagentProgress,
   type SubagentActivityItem,
   AgentTerminateMode,
+  SubagentActivityErrorType,
+  SUBAGENT_REJECTED_ERROR_PREFIX,
+  SUBAGENT_CANCELLED_ERROR_MESSAGE,
 } from './types.js';
 import { randomUUID } from 'node:crypto';
 import type { MessageBus } from '../confirmation-bus/message-bus.js';
@@ -120,7 +123,7 @@ export class LocalSubagentInvocation extends BaseToolInvocation<
               lastItem.type === 'thought' &&
               lastItem.status === 'running'
             ) {
-              lastItem.content += text;
+              lastItem.content = text;
             } else {
               recentActivity.push({
                 id: randomUUID(),
@@ -171,12 +174,19 @@ export class LocalSubagentInvocation extends BaseToolInvocation<
           }
           case 'ERROR': {
             const error = String(activity.data['error']);
-            const isCancellation = error === 'Request cancelled.';
+            const errorType = activity.data['errorType'];
+            const isCancellation =
+              errorType === SubagentActivityErrorType.CANCELLED ||
+              error === SUBAGENT_CANCELLED_ERROR_MESSAGE;
+            const isRejection =
+              errorType === SubagentActivityErrorType.REJECTED ||
+              error.startsWith(SUBAGENT_REJECTED_ERROR_PREFIX);
+
             const toolName = activity.data['name']
               ? String(activity.data['name'])
               : undefined;
 
-            if (toolName && isCancellation) {
+            if (toolName && (isCancellation || isRejection)) {
               for (let i = recentActivity.length - 1; i >= 0; i--) {
                 if (
                   recentActivity[i].type === 'tool_call' &&
@@ -188,13 +198,27 @@ export class LocalSubagentInvocation extends BaseToolInvocation<
                   break;
                 }
               }
+            } else if (toolName) {
+              // Mark non-rejection/non-cancellation errors as 'error'
+              for (let i = recentActivity.length - 1; i >= 0; i--) {
+                if (
+                  recentActivity[i].type === 'tool_call' &&
+                  recentActivity[i].content === toolName &&
+                  recentActivity[i].status === 'running'
+                ) {
+                  recentActivity[i].status = 'error';
+                  updated = true;
+                  break;
+                }
+              }
             }
 
             recentActivity.push({
               id: randomUUID(),
-              type: 'thought', // Treat errors as thoughts for now, or add an error type
-              content: isCancellation ? error : `Error: ${error}`,
-              status: isCancellation ? 'cancelled' : 'error',
+              type: 'thought',
+              content:
+                isCancellation || isRejection ? error : `Error: ${error}`,
+              status: isCancellation || isRejection ? 'cancelled' : 'error',
             });
             updated = true;
             break;

packages/core/src/agents/types.ts

@@ -65,6 +65,18 @@ export type RemoteAgentInputs = { query: string };
 /**
  * Structured events emitted during subagent execution for user observability.
  */
+export enum SubagentActivityErrorType {
+  REJECTED = 'REJECTED',
+  CANCELLED = 'CANCELLED',
+  GENERIC = 'GENERIC',
+}
+
+/**
+ * Standard error messages for subagent activities.
+ */
+export const SUBAGENT_REJECTED_ERROR_PREFIX = 'User rejected this operation.';
+export const SUBAGENT_CANCELLED_ERROR_MESSAGE = 'Request cancelled.';
+
 export interface SubagentActivityEvent {
   isSubagentActivityEvent: true;
   agentName: string;

packages/core/src/config/config.ts

@@ -33,8 +33,6 @@ import { WebFetchTool } from '../tools/web-fetch.js';
 import { MemoryTool, setGeminiMdFilename } from '../tools/memoryTool.js';
 import { WebSearchTool } from '../tools/web-search.js';
 import { AskUserTool } from '../tools/ask-user.js';
-import { ScheduleWorkTool } from '../tools/schedule-work.js';
-import { WorkScheduler } from '../services/work-scheduler.js';
 import { ExitPlanModeTool } from '../tools/exit-plan-mode.js';
 import { EnterPlanModeTool } from '../tools/enter-plan-mode.js';
 import { GeminiClient } from '../core/client.js';
@@ -642,7 +640,6 @@ export interface ConfigParameters {
   mcpEnabled?: boolean;
   extensionsEnabled?: boolean;
   agents?: AgentSettings;
-  isForeverMode?: boolean;
   onReload?: () => Promise<{
     disabledSkills?: string[];
     adminSkillsEnabled?: boolean;
@@ -850,8 +847,6 @@ export class Config implements McpContext, AgentLoopContext {
 
   private readonly enableAgents: boolean;
   private agents: AgentSettings;
-  private readonly isForeverMode: boolean;
-  private readonly workScheduler: WorkScheduler;
   private readonly enableEventDrivenScheduler: boolean;
   private readonly skillsSupport: boolean;
   private disabledSkills: string[];
@@ -964,8 +959,6 @@ export class Config implements McpContext, AgentLoopContext {
     this._activeModel = params.model;
     this.enableAgents = params.enableAgents ?? true;
     this.agents = params.agents ?? {};
-    this.isForeverMode = params.isForeverMode ?? false;
-    this.workScheduler = new WorkScheduler();
     this.disableLLMCorrection = params.disableLLMCorrection ?? true;
     this.planEnabled = params.plan ?? true;
     this.trackerEnabled = params.tracker ?? false;
@@ -2719,14 +2712,6 @@ export class Config implements McpContext, AgentLoopContext {
     return remoteThreshold;
   }
 
-  getIsForeverMode(): boolean {
-    return this.isForeverMode;
-  }
-
-  getWorkScheduler(): WorkScheduler {
-    return this.workScheduler;
-  }
-
   async getUserCaching(): Promise<boolean | undefined> {
     await this.ensureExperimentsLoaded();
 
@@ -2878,7 +2863,6 @@ export class Config implements McpContext, AgentLoopContext {
   }
 
   isInteractiveShellEnabled(): boolean {
-    if (this.isForeverMode) return false;
     return (
       this.interactive &&
       this.ptyInfo !== 'child_process' &&
@@ -3200,24 +3184,15 @@ export class Config implements McpContext, AgentLoopContext {
     maybeRegister(ShellTool, () =>
       registry.registerTool(new ShellTool(this, this.messageBus)),
     );
-    if (!this.isForeverMode) {
-      maybeRegister(MemoryTool, () =>
-        registry.registerTool(new MemoryTool(this.messageBus)),
-      );
-    }
+    maybeRegister(MemoryTool, () =>
+      registry.registerTool(new MemoryTool(this.messageBus)),
+    );
     maybeRegister(WebSearchTool, () =>
       registry.registerTool(new WebSearchTool(this, this.messageBus)),
     );
     maybeRegister(AskUserTool, () =>
       registry.registerTool(new AskUserTool(this.messageBus)),
     );
-    if (this.isForeverMode) {
-      maybeRegister(ScheduleWorkTool, () =>
-        registry.registerTool(
-          new ScheduleWorkTool(this.messageBus, this.workScheduler),
-        ),
-      );
-    }
     if (this.getUseWriteTodos()) {
       maybeRegister(WriteTodosTool, () =>
         registry.registerTool(new WriteTodosTool(this.messageBus)),
@@ -3227,11 +3202,9 @@ export class Config implements McpContext, AgentLoopContext {
       maybeRegister(ExitPlanModeTool, () =>
         registry.registerTool(new ExitPlanModeTool(this, this.messageBus)),
       );
-      if (!this.isForeverMode) {
-        maybeRegister(EnterPlanModeTool, () =>
-          registry.registerTool(new EnterPlanModeTool(this, this.messageBus)),
-        );
-      }
+      maybeRegister(EnterPlanModeTool, () =>
+        registry.registerTool(new EnterPlanModeTool(this, this.messageBus)),
+      );
     }
 
     if (this.isTrackerEnabled()) {

packages/core/src/core/client.test.ts

@@ -143,7 +143,6 @@ const mockHookSystem = {
   fireBeforeAgentEvent: vi.fn().mockResolvedValue(undefined),
   fireAfterAgentEvent: vi.fn().mockResolvedValue(undefined),
   firePreCompressEvent: vi.fn().mockResolvedValue(undefined),
-  fireIdleEvent: vi.fn().mockResolvedValue(undefined),
 };
 
 /**
@@ -453,7 +452,6 @@ describe('Gemini Client (client.ts)', () => {
         getChatRecordingService: vi.fn().mockReturnValue({
           getConversation: vi.fn().mockReturnValue(null),
           getConversationFilePath: vi.fn().mockReturnValue(null),
-          recordCompressionPoint: vi.fn(),
         }),
       };
       client['chat'] = mockOriginalChat as GeminiChat;
@@ -688,7 +686,6 @@ describe('Gemini Client (client.ts)', () => {
       const mockRecordingService = {
         getConversation: vi.fn().mockReturnValue(mockConversation),
         getConversationFilePath: vi.fn().mockReturnValue(mockFilePath),
-        recordCompressionPoint: vi.fn(),
       };
       vi.mocked(mockOriginalChat.getChatRecordingService!).mockReturnValue(
         mockRecordingService as unknown as ChatRecordingService,

packages/core/src/core/client.ts

@@ -879,12 +879,6 @@ export class GeminiClient {
       this.hookStateMap.delete(this.lastPromptId);
       this.lastPromptId = prompt_id;
       this.currentSequenceModel = null;
-
-      // In Forever Mode, refresh the system instruction so memory changes
-      // (e.g. GEMINI.md updates from hooks/agents) are picked up immediately.
-      if (this.config.getIsForeverMode()) {
-        this.updateSystemInstruction();
-      }
     }
 
     if (hooksEnabled && messageBus) {
@@ -1184,7 +1178,6 @@ export class GeminiClient {
         // capture current session data before resetting
         const currentRecordingService =
           this.getChat().getChatRecordingService();
-        currentRecordingService.recordCompressionPoint();
         const conversation = currentRecordingService.getConversation();
         const filePath = currentRecordingService.getConversationFilePath();
 

packages/core/src/core/prompts.test.ts

@@ -116,7 +116,6 @@ describe('Core System Prompt (prompts.ts)', () => {
       getApprovalMode: vi.fn().mockReturnValue(ApprovalMode.DEFAULT),
       getApprovedPlanPath: vi.fn().mockReturnValue(undefined),
       isTrackerEnabled: vi.fn().mockReturnValue(false),
-      getIsForeverMode: vi.fn().mockReturnValue(false),
       get config() {
         return this;
       },
@@ -437,7 +436,6 @@ describe('Core System Prompt (prompts.ts)', () => {
         }),
         getApprovedPlanPath: vi.fn().mockReturnValue(undefined),
         isTrackerEnabled: vi.fn().mockReturnValue(false),
-        getIsForeverMode: vi.fn().mockReturnValue(false),
         get config() {
           return this;
         },

packages/core/src/core/turn.ts

@@ -182,9 +182,6 @@ export enum CompressionStatus {
 
   /** The compression was skipped due to previous failure, but content was truncated to budget */
   CONTENT_TRUNCATED,
-
-  /** The compression was replaced by a PreCompress hook */
-  HOOK_REPLACED,
 }
 
 export interface ChatCompressionInfo {

packages/core/src/hooks/hookAggregator.test.ts

@@ -472,33 +472,5 @@ describe('HookAggregator', () => {
         aggregated.finalOutput?.hookSpecificOutput?.['additionalContext'],
       ).toBe('Context from hook 1\nContext from hook 2');
     });
-
-    it('should propagate refreshContext with any-true-wins logic', () => {
-      const results: HookExecutionResult[] = [
-        createHookExecutionResult({ refreshContext: false }),
-        createHookExecutionResult({ refreshContext: true }),
-      ];
-
-      const aggregated = aggregator.aggregateResults(
-        results,
-        HookEventName.AfterAgent,
-      );
-
-      expect(aggregated.finalOutput?.refreshContext).toBe(true);
-    });
-
-    it('should not set refreshContext when no hook requests it', () => {
-      const results: HookExecutionResult[] = [
-        createHookExecutionResult({}),
-        createHookExecutionResult({}),
-      ];
-
-      const aggregated = aggregator.aggregateResults(
-        results,
-        HookEventName.AfterAgent,
-      );
-
-      expect(aggregated.finalOutput?.refreshContext).toBeUndefined();
-    });
   });
 });

packages/core/src/hooks/hookAggregator.ts

@@ -158,11 +158,6 @@ export class HookAggregator {
         merged.suppressOutput = true;
       }
 
-      // Handle refreshContext (any true wins)
-      if (output.refreshContext) {
-        merged.refreshContext = true;
-      }
-
       // Handle clearContext (any true wins) - for AfterAgent hooks
       if (output.hookSpecificOutput?.['clearContext'] === true) {
         merged.hookSpecificOutput = {
@@ -360,7 +355,7 @@ export class HookAggregator {
     // Extract additionalContext from various hook types
     if (
       'additionalContext' in specific &&
-       
+      // eslint-disable-next-line no-restricted-syntax
       typeof specific['additionalContext'] === 'string'
     ) {
       contexts.push(specific['additionalContext']);

packages/core/src/hooks/hookEventHandler.test.ts

@@ -16,7 +16,6 @@ import {
   SessionStartSource,
   HookEventName,
   HookType,
-  DefaultHookOutput,
   type HookConfig,
   type HookExecutionResult,
 } from './types.js';
@@ -892,100 +891,4 @@ describe('HookEventHandler', () => {
       );
     });
   });
-
-  describe('refreshContext handling', () => {
-    it('should call updateSystemInstructionIfInitialized when refreshContext is true', async () => {
-      const hookConfig = {
-        type: HookType.Command,
-        command: './after-agent.sh',
-      } as HookConfig;
-      const mockResults: HookExecutionResult[] = [
-        {
-          success: true,
-          duration: 100,
-          hookConfig,
-          eventName: HookEventName.AfterAgent,
-          output: { refreshContext: true },
-        },
-      ];
-      const mockAggregated = {
-        success: true,
-        finalOutput: new DefaultHookOutput({ refreshContext: true }),
-        allOutputs: mockResults.map((r) => r.output!),
-        errors: [],
-        totalDuration: 100,
-      };
-
-      vi.mocked(mockHookPlanner.createExecutionPlan).mockReturnValue({
-        hookConfigs: [hookConfig],
-        sequential: false,
-        eventName: HookEventName.AfterAgent,
-      });
-      vi.mocked(mockHookRunner.executeHooksParallel).mockResolvedValue(
-        mockResults,
-      );
-      vi.mocked(mockHookAggregator.aggregateResults).mockReturnValue(
-        mockAggregated,
-      );
-
-      const updateSpy = vi.fn();
-      (mockConfig as unknown as Record<string, unknown>)[
-        'updateSystemInstructionIfInitialized'
-      ] = updateSpy;
-
-      await hookEventHandler.fireAfterAgentEvent(
-        'test prompt',
-        'test response',
-      );
-
-      expect(updateSpy).toHaveBeenCalled();
-    });
-
-    it('should not call updateSystemInstructionIfInitialized when refreshContext is not set', async () => {
-      const hookConfig = {
-        type: HookType.Command,
-        command: './after-agent.sh',
-      } as HookConfig;
-      const mockResults: HookExecutionResult[] = [
-        {
-          success: true,
-          duration: 100,
-          hookConfig,
-          eventName: HookEventName.AfterAgent,
-          output: {},
-        },
-      ];
-      const mockAggregated = {
-        success: true,
-        finalOutput: new DefaultHookOutput({}),
-        allOutputs: mockResults.map((r) => r.output!),
-        errors: [],
-        totalDuration: 100,
-      };
-
-      vi.mocked(mockHookPlanner.createExecutionPlan).mockReturnValue({
-        hookConfigs: [hookConfig],
-        sequential: false,
-        eventName: HookEventName.AfterAgent,
-      });
-      vi.mocked(mockHookRunner.executeHooksParallel).mockResolvedValue(
-        mockResults,
-      );
-      vi.mocked(mockHookAggregator.aggregateResults).mockReturnValue(
-        mockAggregated,
-      );
-
-      const updateSpy = vi.fn();
-      (mockConfig as unknown as Record<string, unknown>)[
-        'updateSystemInstructionIfInitialized'
-      ] = updateSpy;
-
-      await hookEventHandler.fireAfterAgentEvent(
-        'test prompt',
-        'test response',
-      );
-
-      expect(updateSpy).not.toHaveBeenCalled();
-    });
-  });
 });

packages/core/src/hooks/hookEventHandler.ts

@@ -29,7 +29,6 @@ import {
   type PreCompressTrigger,
   type HookExecutionResult,
   type McpToolContext,
-  type IdleInput,
 } from './types.js';
 import { defaultHookTranslator } from './hookTranslator.js';
 import type {
@@ -205,30 +204,16 @@ export class HookEventHandler {
    */
   async firePreCompressEvent(
     trigger: PreCompressTrigger,
-    history: Array<{ role: string; parts: Array<{ text?: string }> }>,
   ): Promise<AggregatedHookResult> {
     const input: PreCompressInput = {
       ...this.createBaseInput(HookEventName.PreCompress),
       trigger,
-      history,
     };
 
     const context: HookEventContext = { trigger };
     return this.executeHooks(HookEventName.PreCompress, input, context);
   }
 
-  /**
-   * Fire an Idle event
-   */
-  async fireIdleEvent(idleSeconds: number): Promise<AggregatedHookResult> {
-    const input: IdleInput = {
-      ...this.createBaseInput(HookEventName.Idle),
-      idle_seconds: idleSeconds,
-    };
-
-    return this.executeHooks(HookEventName.Idle, input);
-  }
-
   /**
    * Fire a BeforeModel event
    * Called by handleHookExecutionRequest - executes hooks directly
@@ -508,11 +493,6 @@ export class HookEventHandler {
       // This is just logging the request centrally
     }
 
-    // Handle refreshContext - reload the system instruction to pick up context file changes
-    if (aggregated.finalOutput.refreshContext) {
-      this.context.config.updateSystemInstructionIfInitialized();
-    }
-
     // Other common fields like decision/reason are handled by specific hook output classes
   }
 

packages/core/src/hooks/hookRegistry.ts

@@ -26,7 +26,6 @@ export interface HookRegistryEntry {
   matcher?: string;
   sequential?: boolean;
   enabled: boolean;
-  idleTimeout?: number;
 }
 
 /**
@@ -280,7 +279,6 @@ please review the project settings (.gemini/settings.json) and remove them.`;
           matcher: definition.matcher,
           sequential: definition.sequential,
           enabled: !isDisabled,
-          idleTimeout: definition.idleTimeout,
         });
       } else {
         // Invalid hooks are logged and discarded here, they won't reach HookRunner

packages/core/src/hooks/hookSystem.ts

@@ -232,15 +232,8 @@ export class HookSystem {
 
   async firePreCompressEvent(
     trigger: PreCompressTrigger,
-    history: Array<{ role: string; parts: Array<{ text?: string }> }>,
   ): Promise<AggregatedHookResult | undefined> {
-    return this.hookEventHandler.firePreCompressEvent(trigger, history);
-  }
-
-  async fireIdleEvent(
-    idleSeconds: number,
-  ): Promise<AggregatedHookResult | undefined> {
-    return this.hookEventHandler.fireIdleEvent(idleSeconds);
+    return this.hookEventHandler.firePreCompressEvent(trigger);
   }
 
   async fireBeforeAgentEvent(

packages/core/src/hooks/types.test.ts

@@ -57,7 +57,6 @@ describe('Hook Types', () => {
         'BeforeModel',
         'AfterModel',
         'BeforeToolSelection',
-        'Idle',
       ];
 
       for (const event of expectedEvents) {
@@ -113,7 +112,6 @@ describe('Hook Output Classes', () => {
         systemMessage: 'test system message',
         decision: 'block' as HookDecision,
         reason: 'test reason',
-        refreshContext: true,
         hookSpecificOutput: { key: 'value' },
       };
       const output = new DefaultHookOutput(data);
@@ -123,7 +121,6 @@ describe('Hook Output Classes', () => {
       expect(output.systemMessage).toBe(data.systemMessage);
       expect(output.decision).toBe(data.decision);
       expect(output.reason).toBe(data.reason);
-      expect(output.refreshContext).toBe(true);
       expect(output.hookSpecificOutput).toEqual(data.hookSpecificOutput);
     });
 

packages/core/src/hooks/types.ts

@@ -43,7 +43,6 @@ export enum HookEventName {
   BeforeModel = 'BeforeModel',
   AfterModel = 'AfterModel',
   BeforeToolSelection = 'BeforeToolSelection',
-  Idle = 'Idle',
 }
 
 /**
@@ -105,8 +104,6 @@ export interface HookDefinition {
   matcher?: string;
   sequential?: boolean;
   hooks: HookConfig[];
-  /** Seconds before the Idle hook fires. Only meaningful for Idle hooks. */
-  idleTimeout?: number;
 }
 
 /**
@@ -150,8 +147,6 @@ export interface HookOutput {
   systemMessage?: string;
   decision?: HookDecision;
   reason?: string;
-  /** When true, refreshes the system instruction after hook execution to pick up context file changes (e.g. GEMINI.md). */
-  refreshContext?: boolean;
   hookSpecificOutput?: Record<string, unknown>;
 }
 
@@ -189,7 +184,6 @@ export class DefaultHookOutput implements HookOutput {
   systemMessage?: string;
   decision?: HookDecision;
   reason?: string;
-  refreshContext?: boolean;
   hookSpecificOutput?: Record<string, unknown>;
 
   constructor(data: Partial<HookOutput> = {}) {
@@ -199,7 +193,6 @@ export class DefaultHookOutput implements HookOutput {
     this.systemMessage = data.systemMessage;
     this.decision = data.decision;
     this.reason = data.reason;
-    this.refreshContext = data.refreshContext;
     this.hookSpecificOutput = data.hookSpecificOutput;
   }
 
@@ -649,37 +642,14 @@ export enum PreCompressTrigger {
  */
 export interface PreCompressInput extends HookInput {
   trigger: PreCompressTrigger;
-  history: Array<{ role: string; parts: Array<{ text?: string }> }>;
 }
+
 /**
  * PreCompress hook output
  */
 export interface PreCompressOutput {
   suppressOutput?: boolean;
   systemMessage?: string;
-  hookSpecificOutput?: {
-    hookEventName: 'PreCompress';
-    newHistory?: Array<{ role: string; parts: Array<{ text?: string }> }>;
-  };
-}
-
-/**
- * Idle hook input
- */
-export interface IdleInput extends HookInput {
-  idle_seconds: number;
-}
-
-/**
- * Idle hook output
- */
-export interface IdleOutput {
-  suppressOutput?: boolean;
-  systemMessage?: string;
-  hookSpecificOutput?: {
-    hookEventName: 'Idle';
-    prompt?: string;
-  };
 }
 
 /**

packages/core/src/index.ts

@@ -125,7 +125,6 @@ export * from './services/fileDiscoveryService.js';
 export * from './services/gitService.js';
 export * from './services/FolderTrustDiscoveryService.js';
 export * from './services/chatRecordingService.js';
-export * from './services/work-scheduler.js';
 export * from './services/fileSystemService.js';
 export * from './services/sessionSummaryUtils.js';
 export * from './services/contextManager.js';

packages/core/src/policy/policies/plan.toml

@@ -33,6 +33,13 @@
 toolName = "enter_plan_mode"
 decision = "ask_user"
 priority = 50
+interactive = true
+
+[[rule]]
+toolName = "enter_plan_mode"
+decision = "allow"
+priority = 50
+interactive = false
 
 [[rule]]
 toolName = "enter_plan_mode"
@@ -46,6 +53,13 @@ toolName = "exit_plan_mode"
 decision = "ask_user"
 priority = 70
 modes = ["plan"]
+interactive = true
+
+[[rule]]
+toolName = "exit_plan_mode"
+decision = "allow"
+priority = 70
+interactive = false
 
 [[rule]]
 toolName = "exit_plan_mode"

packages/core/src/policy/policies/yolo.toml

@@ -45,6 +45,7 @@ toolName = ["enter_plan_mode", "exit_plan_mode"]
 decision = "deny"
 priority = 999
 modes = ["yolo"]
+interactive = true
 
 # Allow everything else in YOLO mode
 [[rule]]

packages/core/src/policy/policy-engine.test.ts

@@ -3343,4 +3343,121 @@ describe('PolicyEngine', () => {
       expect(excluded.has('test-tool')).toBe(false);
     });
   });
+
+  describe('interactive matching', () => {
+    it('should ignore interactive rules in non-interactive mode', async () => {
+      const engine = new PolicyEngine({
+        rules: [
+          {
+            toolName: 'my_tool',
+            decision: PolicyDecision.ALLOW,
+            interactive: true,
+          },
+        ],
+        nonInteractive: true,
+        defaultDecision: PolicyDecision.DENY,
+      });
+
+      const result = await engine.check(
+        { name: 'my_tool', args: {} },
+        undefined,
+      );
+      expect(result.decision).toBe(PolicyDecision.DENY);
+    });
+
+    it('should allow interactive rules in interactive mode', async () => {
+      const engine = new PolicyEngine({
+        rules: [
+          {
+            toolName: 'my_tool',
+            decision: PolicyDecision.ALLOW,
+            interactive: true,
+          },
+        ],
+        nonInteractive: false,
+        defaultDecision: PolicyDecision.DENY,
+      });
+
+      const result = await engine.check(
+        { name: 'my_tool', args: {} },
+        undefined,
+      );
+      expect(result.decision).toBe(PolicyDecision.ALLOW);
+    });
+
+    it('should ignore non-interactive rules in interactive mode', async () => {
+      const engine = new PolicyEngine({
+        rules: [
+          {
+            toolName: 'my_tool',
+            decision: PolicyDecision.ALLOW,
+            interactive: false,
+          },
+        ],
+        nonInteractive: false,
+        defaultDecision: PolicyDecision.DENY,
+      });
+
+      const result = await engine.check(
+        { name: 'my_tool', args: {} },
+        undefined,
+      );
+      expect(result.decision).toBe(PolicyDecision.DENY);
+    });
+
+    it('should allow non-interactive rules in non-interactive mode', async () => {
+      const engine = new PolicyEngine({
+        rules: [
+          {
+            toolName: 'my_tool',
+            decision: PolicyDecision.ALLOW,
+            interactive: false,
+          },
+        ],
+        nonInteractive: true,
+        defaultDecision: PolicyDecision.DENY,
+      });
+
+      const result = await engine.check(
+        { name: 'my_tool', args: {} },
+        undefined,
+      );
+      expect(result.decision).toBe(PolicyDecision.ALLOW);
+    });
+
+    it('should apply rules without interactive flag to both', async () => {
+      const rule: PolicyRule = {
+        toolName: 'my_tool',
+        decision: PolicyDecision.ALLOW,
+      };
+
+      const engineInteractive = new PolicyEngine({
+        rules: [rule],
+        nonInteractive: false,
+        defaultDecision: PolicyDecision.DENY,
+      });
+      const engineNonInteractive = new PolicyEngine({
+        rules: [rule],
+        nonInteractive: true,
+        defaultDecision: PolicyDecision.DENY,
+      });
+
+      expect(
+        (
+          await engineInteractive.check(
+            { name: 'my_tool', args: {} },
+            undefined,
+          )
+        ).decision,
+      ).toBe(PolicyDecision.ALLOW);
+      expect(
+        (
+          await engineNonInteractive.check(
+            { name: 'my_tool', args: {} },
+            undefined,
+          )
+        ).decision,
+      ).toBe(PolicyDecision.ALLOW);
+    });
+  });
 });

packages/core/src/policy/policy-engine.ts

@@ -74,6 +74,7 @@ function ruleMatches(
   stringifiedArgs: string | undefined,
   serverName: string | undefined,
   currentApprovalMode: ApprovalMode,
+  nonInteractive: boolean,
   toolAnnotations?: Record<string, unknown>,
   subagent?: string,
 ): boolean {
@@ -146,6 +147,16 @@ function ruleMatches(
     }
   }
 
+  // Check interactive if specified
+  if ('interactive' in rule && rule.interactive !== undefined) {
+    if (rule.interactive && nonInteractive) {
+      return false;
+    }
+    if (!rule.interactive && !nonInteractive) {
+      return false;
+    }
+  }
+
   return true;
 }
 
@@ -443,6 +454,7 @@ export class PolicyEngine {
           stringifiedArgs,
           serverName,
           this.approvalMode,
+          this.nonInteractive,
           toolAnnotations,
           subagent,
         ),
@@ -521,6 +533,7 @@ export class PolicyEngine {
             stringifiedArgs,
             serverName,
             this.approvalMode,
+            this.nonInteractive,
             toolAnnotations,
             subagent,
           )
@@ -713,6 +726,7 @@ export class PolicyEngine {
           undefined, // stringifiedArgs
           serverName,
           this.approvalMode,
+          this.nonInteractive,
           annotations,
         );
 

packages/core/src/policy/toml-loader.ts

@@ -61,6 +61,7 @@ const PolicyRuleSchema = z.object({
         'priority must be <= 999 to prevent tier overflow. Priorities >= 1000 would jump to the next tier.',
     }),
   modes: z.array(z.nativeEnum(ApprovalMode)).optional(),
+  interactive: z.boolean().optional(),
   toolAnnotations: z.record(z.any()).optional(),
   allow_redirection: z.boolean().optional(),
   deny_message: z.string().optional(),
@@ -475,6 +476,7 @@ export async function loadPoliciesFromToml(
                   decision: rule.decision,
                   priority: transformPriority(rule.priority, tier),
                   modes: rule.modes,
+                  interactive: rule.interactive,
                   toolAnnotations: rule.toolAnnotations,
                   allowRedirection: rule.allow_redirection,
                   source: `${tierName.charAt(0).toUpperCase() + tierName.slice(1)}: ${file}`,

packages/core/src/policy/types.ts

@@ -152,6 +152,13 @@ export interface PolicyRule {
    */
   modes?: ApprovalMode[];
 
+  /**
+   * If true, this rule only applies to interactive environments.
+   * If false, this rule only applies to non-interactive environments.
+   * If undefined, it applies to both interactive and non-interactive environments.
+   */
+  interactive?: boolean;
+
   /**
    * If true, allows command redirection even if the policy engine would normally
    * downgrade ALLOW to ASK_USER for redirected commands.

packages/core/src/prompts/promptProvider.test.ts

@@ -71,7 +71,6 @@ describe('PromptProvider', () => {
       getApprovedPlanPath: vi.fn().mockReturnValue(undefined),
       getApprovalMode: vi.fn(),
       isTrackerEnabled: vi.fn().mockReturnValue(false),
-      getIsForeverMode: vi.fn().mockReturnValue(false),
     } as unknown as Config;
   });
 

packages/core/src/prompts/promptProvider.ts

@@ -175,6 +175,7 @@ export class PromptProvider {
         planningWorkflow: this.withSection(
           'planningWorkflow',
           () => ({
+            interactive: interactiveMode,
             planModeToolsList,
             plansDir: context.config.storage.getPlansDir(),
             approvedPlanPath: context.config.getApprovedPlanPath(),

packages/core/src/prompts/snippets.ts

@@ -88,6 +88,7 @@ export interface GitRepoOptions {
 }
 
 export interface PlanningWorkflowOptions {
+  interactive: boolean;
   planModeToolsList: string;
   plansDir: string;
   approvedPlanPath?: string;
@@ -513,7 +514,7 @@ export function renderPlanningWorkflow(
   return `
 # Active Approval Mode: Plan
 
-You are operating in **Plan Mode**. Your goal is to produce an implementation plan in \`${options.plansDir}/\` and get user approval before editing source code.
+You are operating in **Plan Mode**. Your goal is to produce an implementation plan in \`${options.plansDir}/\` and ${options.interactive ? 'get user approval before editing source code.' : 'create a design document before proceeding autonomously.'}
 
 ## Available Tools
 The following tools are available in Plan Mode:
@@ -550,7 +551,7 @@ Write the implementation plan to \`${options.plansDir}/\`. The plan's structure
 - **Complex Tasks:** Include **Background & Motivation**, **Scope & Impact**, **Proposed Solution**, **Alternatives Considered**, a phased **Implementation Plan**, **Verification**, and **Migration & Rollback** strategies.
 
 ### 4. Review & Approval
-Use the ${formatToolName(EXIT_PLAN_MODE_TOOL_NAME)} tool to present the plan and formally request approval.
+Use the ${formatToolName(EXIT_PLAN_MODE_TOOL_NAME)} tool to present the plan and ${options.interactive ? 'formally request approval.' : 'begin implementation.'}
 
 ${renderApprovedPlanSection(options.approvedPlanPath)}`.trim();
 }
@@ -711,7 +712,7 @@ function newApplicationSteps(options: PrimaryWorkflowsOptions): string {
   // standard 'Execution' loop handle implementation once the plan is approved.
   if (options.enableEnterPlanModeTool) {
     return `
-1. **Mandatory Planning:** You MUST use the ${formatToolName(ENTER_PLAN_MODE_TOOL_NAME)} tool to draft a comprehensive design document and obtain user approval before writing any code.
+1. **Mandatory Planning:** You MUST use the ${formatToolName(ENTER_PLAN_MODE_TOOL_NAME)} tool to draft a comprehensive design document${options.interactive ? ' and obtain user approval' : ''} before writing any code.
 2. **Design Constraints:** When drafting your plan, adhere to these defaults unless explicitly overridden by the user:
    - **Goal:** Autonomously design a visually appealing, substantially complete, and functional prototype with rich aesthetics. Users judge applications by their visual impact; ensure they feel modern, "alive," and polished through consistent spacing, typography, and interactive feedback.
    - **Visuals:** Describe your strategy for sourcing or generating placeholders (e.g., stylized CSS shapes, gradients, procedurally generated patterns) to ensure a visually complete prototype. Never plan for assets that cannot be locally generated.

packages/core/src/sandbox/macos/MacOsSandboxManager.integration.test.ts

@@ -0,0 +1,202 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+import { describe, it, expect, beforeAll, afterAll } from 'vitest';
+import { MacOsSandboxManager } from './MacOsSandboxManager.js';
+import { ShellExecutionService } from '../../services/shellExecutionService.js';
+import { getSecureSanitizationConfig } from '../../services/environmentSanitization.js';
+import { type SandboxedCommand } from '../../services/sandboxManager.js';
+import { execFile } from 'node:child_process';
+import { promisify } from 'node:util';
+import os from 'node:os';
+import fs from 'node:fs';
+import path from 'node:path';
+import http from 'node:http';
+
+/**
+ * A simple asynchronous wrapper for execFile that returns the exit status,
+ * stdout, and stderr. Unlike spawnSync, this does not block the Node.js
+ * event loop, allowing the local HTTP test server to function.
+ */
+async function runCommand(command: SandboxedCommand) {
+  try {
+    const { stdout, stderr } = await promisify(execFile)(
+      command.program,
+      command.args,
+      {
+        cwd: command.cwd,
+        env: command.env,
+        encoding: 'utf-8',
+      },
+    );
+    return { status: 0, stdout, stderr };
+  } catch (error: unknown) {
+    const err = error as {
+      code?: number;
+      stdout?: string;
+      stderr?: string;
+    };
+    return {
+      status: err.code ?? 1,
+      stdout: err.stdout ?? '',
+      stderr: err.stderr ?? '',
+    };
+  }
+}
+
+describe.skipIf(os.platform() !== 'darwin')(
+  'MacOsSandboxManager Integration',
+  () => {
+    describe('Basic Execution', () => {
+      it('should execute commands within the workspace', async () => {
+        const manager = new MacOsSandboxManager({ workspace: process.cwd() });
+        const command = await manager.prepareCommand({
+          command: 'echo',
+          args: ['sandbox test'],
+          cwd: process.cwd(),
+          env: process.env,
+        });
+
+        const execResult = await runCommand(command);
+
+        expect(execResult.status).toBe(0);
+        expect(execResult.stdout.trim()).toBe('sandbox test');
+      });
+
+      it('should support interactive pseudo-terminals (node-pty)', async () => {
+        const manager = new MacOsSandboxManager({ workspace: process.cwd() });
+        const abortController = new AbortController();
+
+        // Verify that node-pty file descriptors are successfully allocated inside the sandbox
+        // by using the bash [ -t 1 ] idiom to check if stdout is a TTY.
+        const handle = await ShellExecutionService.execute(
+          'bash -c "if [ -t 1 ]; then echo True; else echo False; fi"',
+          process.cwd(),
+          () => {},
+          abortController.signal,
+          true,
+          {
+            sanitizationConfig: getSecureSanitizationConfig(),
+            sandboxManager: manager,
+          },
+        );
+
+        const result = await handle.result;
+        expect(result.error).toBeNull();
+        expect(result.exitCode).toBe(0);
+        expect(result.output).toContain('True');
+      });
+    });
+
+    describe('File System Access', () => {
+      it('should block file system access outside the workspace', async () => {
+        const manager = new MacOsSandboxManager({ workspace: process.cwd() });
+        const blockedPath = '/Users/Shared/.gemini_test_sandbox_blocked';
+
+        const command = await manager.prepareCommand({
+          command: 'touch',
+          args: [blockedPath],
+          cwd: process.cwd(),
+          env: process.env,
+        });
+        const execResult = await runCommand(command);
+
+        expect(execResult.status).not.toBe(0);
+        expect(execResult.stderr).toContain('Operation not permitted');
+      });
+
+      it('should grant file system access to explicitly allowed paths', async () => {
+        // Create a unique temporary directory to prevent artifacts and test flakiness
+        const allowedDir = fs.mkdtempSync(
+          path.join(os.tmpdir(), 'gemini-sandbox-test-'),
+        );
+
+        try {
+          const manager = new MacOsSandboxManager({
+            workspace: process.cwd(),
+            allowedPaths: [allowedDir],
+          });
+          const testFile = path.join(allowedDir, 'test.txt');
+
+          const command = await manager.prepareCommand({
+            command: 'touch',
+            args: [testFile],
+            cwd: process.cwd(),
+            env: process.env,
+          });
+
+          const execResult = await runCommand(command);
+
+          expect(execResult.status).toBe(0);
+        } finally {
+          fs.rmSync(allowedDir, { recursive: true, force: true });
+        }
+      });
+    });
+
+    describe('Network Access', () => {
+      let testServer: http.Server;
+      let testServerUrl: string;
+
+      beforeAll(async () => {
+        testServer = http.createServer((_, res) => {
+          // Ensure connections are closed immediately to prevent hanging
+          res.setHeader('Connection', 'close');
+          res.writeHead(200);
+          res.end('ok');
+        });
+
+        await new Promise<void>((resolve, reject) => {
+          testServer.on('error', reject);
+          testServer.listen(0, '127.0.0.1', () => {
+            const address = testServer.address() as import('net').AddressInfo;
+            testServerUrl = `http://127.0.0.1:${address.port}`;
+            resolve();
+          });
+        });
+      });
+
+      afterAll(async () => {
+        if (testServer) {
+          await new Promise<void>((resolve) => {
+            testServer.close(() => resolve());
+          });
+        }
+      });
+
+      it('should block network access by default', async () => {
+        const manager = new MacOsSandboxManager({ workspace: process.cwd() });
+        const command = await manager.prepareCommand({
+          command: 'curl',
+          args: ['-s', '--connect-timeout', '1', testServerUrl],
+          cwd: process.cwd(),
+          env: process.env,
+        });
+
+        const execResult = await runCommand(command);
+
+        expect(execResult.status).not.toBe(0);
+      });
+
+      it('should grant network access when explicitly allowed', async () => {
+        const manager = new MacOsSandboxManager({
+          workspace: process.cwd(),
+          networkAccess: true,
+        });
+        const command = await manager.prepareCommand({
+          command: 'curl',
+          args: ['-s', '--connect-timeout', '1', testServerUrl],
+          cwd: process.cwd(),
+          env: process.env,
+        });
+
+        const execResult = await runCommand(command);
+
+        expect(execResult.status).toBe(0);
+        expect(execResult.stdout.trim()).toBe('ok');
+      });
+    });
+  },
+);

packages/core/src/sandbox/macos/MacOsSandboxManager.test.ts

@@ -0,0 +1,107 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+import {
+  describe,
+  it,
+  expect,
+  vi,
+  beforeEach,
+  afterEach,
+  type MockInstance,
+} from 'vitest';
+import { MacOsSandboxManager } from './MacOsSandboxManager.js';
+import * as seatbeltArgsBuilder from './seatbeltArgsBuilder.js';
+
+describe('MacOsSandboxManager', () => {
+  const mockWorkspace = '/test/workspace';
+  const mockAllowedPaths = ['/test/allowed'];
+  const mockNetworkAccess = true;
+
+  let manager: MacOsSandboxManager;
+  let buildArgsSpy: MockInstance<typeof seatbeltArgsBuilder.buildSeatbeltArgs>;
+
+  beforeEach(() => {
+    manager = new MacOsSandboxManager({
+      workspace: mockWorkspace,
+      allowedPaths: mockAllowedPaths,
+      networkAccess: mockNetworkAccess,
+    });
+
+    buildArgsSpy = vi
+      .spyOn(seatbeltArgsBuilder, 'buildSeatbeltArgs')
+      .mockReturnValue([
+        '-p',
+        '(mock profile)',
+        '-D',
+        'WORKSPACE=/test/workspace',
+      ]);
+  });
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  it('should correctly invoke buildSeatbeltArgs with the configured options', async () => {
+    await manager.prepareCommand({
+      command: 'echo',
+      args: ['hello'],
+      cwd: mockWorkspace,
+      env: {},
+    });
+
+    expect(buildArgsSpy).toHaveBeenCalledWith({
+      workspace: mockWorkspace,
+      allowedPaths: mockAllowedPaths,
+      networkAccess: mockNetworkAccess,
+    });
+  });
+
+  it('should format the executable and arguments correctly for sandbox-exec', async () => {
+    const result = await manager.prepareCommand({
+      command: 'echo',
+      args: ['hello'],
+      cwd: mockWorkspace,
+      env: {},
+    });
+
+    expect(result.program).toBe('/usr/bin/sandbox-exec');
+    expect(result.args).toEqual([
+      '-p',
+      '(mock profile)',
+      '-D',
+      'WORKSPACE=/test/workspace',
+      '--',
+      'echo',
+      'hello',
+    ]);
+  });
+
+  it('should correctly pass through the cwd to the resulting command', async () => {
+    const result = await manager.prepareCommand({
+      command: 'echo',
+      args: ['hello'],
+      cwd: '/test/different/cwd',
+      env: {},
+    });
+
+    expect(result.cwd).toBe('/test/different/cwd');
+  });
+
+  it('should apply environment sanitization via the default mechanisms', async () => {
+    const result = await manager.prepareCommand({
+      command: 'echo',
+      args: ['hello'],
+      cwd: mockWorkspace,
+      env: {
+        SAFE_VAR: '1',
+        GITHUB_TOKEN: 'sensitive',
+      },
+    });
+
+    expect(result.env['SAFE_VAR']).toBe('1');
+    expect(result.env['GITHUB_TOKEN']).toBeUndefined();
+  });
+});

packages/core/src/sandbox/macos/MacOsSandboxManager.ts

@@ -0,0 +1,60 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import {
+  type SandboxManager,
+  type SandboxRequest,
+  type SandboxedCommand,
+} from '../../services/sandboxManager.js';
+import {
+  sanitizeEnvironment,
+  getSecureSanitizationConfig,
+  type EnvironmentSanitizationConfig,
+} from '../../services/environmentSanitization.js';
+import { buildSeatbeltArgs } from './seatbeltArgsBuilder.js';
+
+/**
+ * Options for configuring the MacOsSandboxManager.
+ */
+export interface MacOsSandboxOptions {
+  /** The primary workspace path to allow access to within the sandbox. */
+  workspace: string;
+  /** Additional paths to allow access to within the sandbox. */
+  allowedPaths?: string[];
+  /** Whether network access is allowed. */
+  networkAccess?: boolean;
+  /** Optional base sanitization config. */
+  sanitizationConfig?: EnvironmentSanitizationConfig;
+}
+
+/**
+ * A SandboxManager implementation for macOS that uses Seatbelt.
+ */
+export class MacOsSandboxManager implements SandboxManager {
+  constructor(private readonly options: MacOsSandboxOptions) {}
+
+  async prepareCommand(req: SandboxRequest): Promise<SandboxedCommand> {
+    const sanitizationConfig = getSecureSanitizationConfig(
+      req.config?.sanitizationConfig,
+      this.options.sanitizationConfig,
+    );
+
+    const sanitizedEnv = sanitizeEnvironment(req.env, sanitizationConfig);
+
+    const sandboxArgs = buildSeatbeltArgs({
+      workspace: this.options.workspace,
+      allowedPaths: this.options.allowedPaths,
+      networkAccess: this.options.networkAccess,
+    });
+
+    return {
+      program: '/usr/bin/sandbox-exec',
+      args: [...sandboxArgs, '--', req.command, ...req.args],
+      env: sanitizedEnv,
+      cwd: req.cwd,
+    };
+  }
+}

packages/core/src/sandbox/macos/baseProfile.ts

@@ -0,0 +1,94 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+/**
+ * The base macOS Seatbelt (SBPL) profile for tool execution.
+ *
+ * This uses a strict allowlist (deny default) but imports Apple's base system profile
+ * to handle undocumented internal dependencies, sysctls, and IPC mach ports required
+ * by standard tools to avoid "Abort trap: 6".
+ */
+export const BASE_SEATBELT_PROFILE = `(version 1)
+(deny default)
+
+(import "system.sb")
+
+; Core execution requirements
+(allow process-exec)
+(allow process-fork)
+(allow signal (target same-sandbox))
+(allow process-info* (target same-sandbox))
+
+; Allow basic read access to system frameworks and libraries required to run
+(allow file-read*
+  (subpath "/System")
+  (subpath "/usr/lib")
+  (subpath "/usr/share")
+  (subpath "/usr/bin")
+  (subpath "/bin")
+  (subpath "/sbin")
+  (subpath "/usr/local/bin")
+  (subpath "/opt/homebrew")
+  (subpath "/Library")
+  (subpath "/private/var/run")
+  (subpath "/private/var/db")
+  (subpath "/private/etc")
+)
+
+; PTY and Terminal support
+(allow pseudo-tty)
+(allow file-read* file-write* file-ioctl (literal "/dev/ptmx"))
+(allow file-read* file-write* file-ioctl (regex #"^/dev/ttys[0-9]+"))
+
+; Allow read/write access to temporary directories and common device nodes
+(allow file-read* file-write*
+  (literal "/dev/null")
+  (literal "/dev/zero")
+  (subpath "/tmp")
+  (subpath "/private/tmp")
+  (subpath (param "TMPDIR"))
+)
+
+; Workspace access using parameterized paths
+(allow file-read* file-write*
+  (subpath (param "WORKSPACE"))
+)
+`;
+
+/**
+ * The network-specific macOS Seatbelt (SBPL) profile rules.
+ *
+ * These rules are appended to the base profile when network access is enabled,
+ * allowing standard socket creation, DNS resolution, and TLS certificate validation.
+ */
+export const NETWORK_SEATBELT_PROFILE = `
+; Network Access
+(allow network*)
+
+(allow system-socket
+  (require-all
+    (socket-domain AF_SYSTEM)
+    (socket-protocol 2)
+  )
+)
+
+(allow mach-lookup
+    (global-name "com.apple.bsd.dirhelper")
+    (global-name "com.apple.system.opendirectoryd.membership")
+    (global-name "com.apple.SecurityServer")
+    (global-name "com.apple.networkd")
+    (global-name "com.apple.ocspd")
+    (global-name "com.apple.trustd.agent")
+    (global-name "com.apple.mDNSResponder")
+    (global-name "com.apple.mDNSResponderHelper")
+    (global-name "com.apple.SystemConfiguration.DNSConfiguration")
+    (global-name "com.apple.SystemConfiguration.configd")
+)
+
+(allow sysctl-read
+  (sysctl-name-regex #"^net.routetable")
+)
+`;

packages/core/src/sandbox/macos/seatbeltArgsBuilder.test.ts

@@ -0,0 +1,97 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+import { describe, it, expect, vi } from 'vitest';
+import { buildSeatbeltArgs } from './seatbeltArgsBuilder.js';
+import fs from 'node:fs';
+import os from 'node:os';
+
+describe('seatbeltArgsBuilder', () => {
+  it('should build a strict allowlist profile allowing the workspace via param', () => {
+    // Mock realpathSync to just return the path for testing
+    vi.spyOn(fs, 'realpathSync').mockImplementation((p) => p as string);
+
+    const args = buildSeatbeltArgs({ workspace: '/Users/test/workspace' });
+
+    expect(args[0]).toBe('-p');
+    const profile = args[1];
+    expect(profile).toContain('(version 1)');
+    expect(profile).toContain('(deny default)');
+    expect(profile).toContain('(allow process-exec)');
+    expect(profile).toContain('(subpath (param "WORKSPACE"))');
+    expect(profile).not.toContain('(allow network*)');
+
+    expect(args).toContain('-D');
+    expect(args).toContain('WORKSPACE=/Users/test/workspace');
+    expect(args).toContain(`TMPDIR=${os.tmpdir()}`);
+
+    vi.restoreAllMocks();
+  });
+
+  it('should allow network when networkAccess is true', () => {
+    const args = buildSeatbeltArgs({ workspace: '/test', networkAccess: true });
+    const profile = args[1];
+    expect(profile).toContain('(allow network*)');
+  });
+
+  it('should parameterize allowed paths and normalize them', () => {
+    vi.spyOn(fs, 'realpathSync').mockImplementation((p) => {
+      if (p === '/test/symlink') return '/test/real_path';
+      return p as string;
+    });
+
+    const args = buildSeatbeltArgs({
+      workspace: '/test',
+      allowedPaths: ['/custom/path1', '/test/symlink'],
+    });
+
+    const profile = args[1];
+    expect(profile).toContain('(subpath (param "ALLOWED_PATH_0"))');
+    expect(profile).toContain('(subpath (param "ALLOWED_PATH_1"))');
+
+    expect(args).toContain('-D');
+    expect(args).toContain('ALLOWED_PATH_0=/custom/path1');
+    expect(args).toContain('ALLOWED_PATH_1=/test/real_path');
+
+    vi.restoreAllMocks();
+  });
+
+  it('should resolve parent directories if a file does not exist', () => {
+    vi.spyOn(fs, 'realpathSync').mockImplementation((p) => {
+      if (p === '/test/symlink/nonexistent.txt') {
+        const error = new Error('ENOENT');
+        Object.assign(error, { code: 'ENOENT' });
+        throw error;
+      }
+      if (p === '/test/symlink') {
+        return '/test/real_path';
+      }
+      return p as string;
+    });
+
+    const args = buildSeatbeltArgs({
+      workspace: '/test/symlink/nonexistent.txt',
+    });
+
+    expect(args).toContain('WORKSPACE=/test/real_path/nonexistent.txt');
+    vi.restoreAllMocks();
+  });
+
+  it('should throw if realpathSync throws a non-ENOENT error', () => {
+    vi.spyOn(fs, 'realpathSync').mockImplementation(() => {
+      const error = new Error('Permission denied');
+      Object.assign(error, { code: 'EACCES' });
+      throw error;
+    });
+
+    expect(() =>
+      buildSeatbeltArgs({
+        workspace: '/test/workspace',
+      }),
+    ).toThrow('Permission denied');
+
+    vi.restoreAllMocks();
+  });
+});

packages/core/src/sandbox/macos/seatbeltArgsBuilder.ts

@@ -0,0 +1,80 @@
+/**
+ * @license
+ * Copyright 2026 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import {
+  BASE_SEATBELT_PROFILE,
+  NETWORK_SEATBELT_PROFILE,
+} from './baseProfile.js';
+
+/**
+ * Options for building macOS Seatbelt arguments.
+ */
+export interface SeatbeltArgsOptions {
+  /** The primary workspace path to allow access to. */
+  workspace: string;
+  /** Additional paths to allow access to. */
+  allowedPaths?: string[];
+  /** Whether to allow network access. */
+  networkAccess?: boolean;
+}
+
+/**
+ * Resolves symlinks for a given path to prevent sandbox escapes.
+ * If a file does not exist (ENOENT), it recursively resolves the parent directory.
+ * Other errors (e.g. EACCES) are re-thrown.
+ */
+function tryRealpath(p: string): string {
+  try {
+    return fs.realpathSync(p);
+  } catch (e) {
+    if (e instanceof Error && 'code' in e && e.code === 'ENOENT') {
+      const parentDir = path.dirname(p);
+      if (parentDir === p) {
+        return p;
+      }
+      return path.join(tryRealpath(parentDir), path.basename(p));
+    }
+    throw e;
+  }
+}
+
+/**
+ * Builds the arguments array for sandbox-exec using a strict allowlist profile.
+ * It relies on parameters passed to sandbox-exec via the -D flag to avoid
+ * string interpolation vulnerabilities, and normalizes paths against symlink escapes.
+ *
+ * Returns arguments up to the end of sandbox-exec configuration (e.g. ['-p', '<profile>', '-D', ...])
+ * Does not include the final '--' separator or the command to run.
+ */
+export function buildSeatbeltArgs(options: SeatbeltArgsOptions): string[] {
+  let profile = BASE_SEATBELT_PROFILE + '\n';
+  const args: string[] = [];
+
+  const workspacePath = tryRealpath(options.workspace);
+  args.push('-D', `WORKSPACE=${workspacePath}`);
+
+  const tmpPath = tryRealpath(os.tmpdir());
+  args.push('-D', `TMPDIR=${tmpPath}`);
+
+  if (options.allowedPaths) {
+    for (let i = 0; i < options.allowedPaths.length; i++) {
+      const allowedPath = tryRealpath(options.allowedPaths[i]);
+      args.push('-D', `ALLOWED_PATH_${i}=${allowedPath}`);
+      profile += `(allow file-read* file-write* (subpath (param "ALLOWED_PATH_${i}")))\n`;
+    }
+  }
+
+  if (options.networkAccess) {
+    profile += NETWORK_SEATBELT_PROFILE;
+  }
+
+  args.unshift('-p', profile);
+
+  return args;
+}

packages/core/src/services/chatCompressionService.test.ts

@@ -186,7 +186,7 @@ describe('ChatCompressionService', () => {
       }),
       getEnableHooks: vi.fn().mockReturnValue(false),
       getMessageBus: vi.fn().mockReturnValue(undefined),
-      getHookSystem: vi.fn().mockReturnValue(undefined),
+      getHookSystem: () => undefined,
       getNextCompressionTruncationId: vi.fn().mockReturnValue(1),
       getTruncateToolOutputThreshold: vi.fn().mockReturnValue(40000),
       storage: {
@@ -897,151 +897,4 @@ describe('ChatCompressionService', () => {
       );
     });
   });
-
-  describe('PreCompress hook replacement', () => {
-    it('should use hook-provided newHistory and skip built-in compression', async () => {
-      const history: Content[] = [
-        { role: 'user', parts: [{ text: 'msg1' }] },
-        { role: 'model', parts: [{ text: 'msg2' }] },
-        { role: 'user', parts: [{ text: 'msg3' }] },
-        { role: 'model', parts: [{ text: 'msg4' }] },
-      ];
-      vi.mocked(mockChat.getHistory).mockReturnValue(history);
-      vi.mocked(mockChat.getLastPromptTokenCount).mockReturnValue(600000);
-      vi.mocked(tokenLimit).mockReturnValue(1_000_000);
-
-      const hookReplacementHistory = [
-        {
-          role: 'user',
-          parts: [{ text: 'Archive summary: topics discussed...' }],
-        },
-        {
-          role: 'model',
-          parts: [{ text: 'Understood, continuing from archive.' }],
-        },
-      ];
-
-      const mockHookSystem = {
-        firePreCompressEvent: vi.fn().mockResolvedValue({
-          success: true,
-          finalOutput: {
-            hookSpecificOutput: {
-              hookEventName: 'PreCompress',
-              newHistory: hookReplacementHistory,
-            },
-          },
-          allOutputs: [],
-          errors: [],
-          totalDuration: 100,
-        }),
-      };
-
-      vi.mocked(mockConfig.getHookSystem).mockReturnValue(
-        mockHookSystem as unknown as ReturnType<Config['getHookSystem']>,
-      );
-
-      const result = await service.compress(
-        mockChat,
-        mockPromptId,
-        true,
-        mockModel,
-        mockConfig,
-        false,
-      );
-
-      expect(result.info.compressionStatus).toBe(
-        CompressionStatus.HOOK_REPLACED,
-      );
-      expect(result.newHistory).not.toBeNull();
-      expect(result.newHistory!.length).toBe(2);
-      expect(result.newHistory![0].parts![0].text).toBe(
-        'Archive summary: topics discussed...',
-      );
-      // Built-in LLM compression should NOT have been called
-      expect(
-        mockConfig.getBaseLlmClient().generateContent,
-      ).not.toHaveBeenCalled();
-    });
-
-    it('should proceed with built-in compression when hook returns no newHistory', async () => {
-      const history: Content[] = [
-        { role: 'user', parts: [{ text: 'msg1' }] },
-        { role: 'model', parts: [{ text: 'msg2' }] },
-        { role: 'user', parts: [{ text: 'msg3' }] },
-        { role: 'model', parts: [{ text: 'msg4' }] },
-      ];
-      vi.mocked(mockChat.getHistory).mockReturnValue(history);
-      vi.mocked(mockChat.getLastPromptTokenCount).mockReturnValue(600000);
-      vi.mocked(tokenLimit).mockReturnValue(1_000_000);
-
-      const mockHookSystem = {
-        firePreCompressEvent: vi.fn().mockResolvedValue({
-          success: true,
-          finalOutput: {
-            systemMessage: 'Compression starting...',
-          },
-          allOutputs: [],
-          errors: [],
-          totalDuration: 50,
-        }),
-      };
-
-      vi.mocked(mockConfig.getHookSystem).mockReturnValue(
-        mockHookSystem as unknown as ReturnType<Config['getHookSystem']>,
-      );
-
-      const result = await service.compress(
-        mockChat,
-        mockPromptId,
-        true,
-        mockModel,
-        mockConfig,
-        false,
-      );
-
-      // Should fall through to normal compression
-      expect(result.info.compressionStatus).toBe(CompressionStatus.COMPRESSED);
-      expect(mockConfig.getBaseLlmClient().generateContent).toHaveBeenCalled();
-    });
-
-    it('should pass history to the hook', async () => {
-      const history: Content[] = [
-        { role: 'user', parts: [{ text: 'hello' }] },
-        { role: 'model', parts: [{ text: 'world' }] },
-      ];
-      vi.mocked(mockChat.getHistory).mockReturnValue(history);
-      vi.mocked(mockChat.getLastPromptTokenCount).mockReturnValue(600000);
-      vi.mocked(tokenLimit).mockReturnValue(1_000_000);
-
-      const mockHookSystem = {
-        firePreCompressEvent: vi.fn().mockResolvedValue({
-          success: true,
-          allOutputs: [],
-          errors: [],
-          totalDuration: 10,
-        }),
-      };
-
-      vi.mocked(mockConfig.getHookSystem).mockReturnValue(
-        mockHookSystem as unknown as ReturnType<Config['getHookSystem']>,
-      );
-
-      await service.compress(
-        mockChat,
-        mockPromptId,
-        true,
-        mockModel,
-        mockConfig,
-        false,
-      );
-
-      expect(mockHookSystem.firePreCompressEvent).toHaveBeenCalledWith(
-        'manual',
-        [
-          { role: 'user', parts: [{ text: 'hello' }] },
-          { role: 'model', parts: [{ text: 'world' }] },
-        ],
-      );
-    });
-  });
 });

packages/core/src/services/chatCompressionService.ts

@@ -156,13 +156,13 @@ async function truncateHistoryToBudget(
           } else if (responseObj && typeof responseObj === 'object') {
             if (
               'output' in responseObj &&
-               
+              // eslint-disable-next-line no-restricted-syntax
               typeof responseObj['output'] === 'string'
             ) {
               contentStr = responseObj['output'];
             } else if (
               'content' in responseObj &&
-               
+              // eslint-disable-next-line no-restricted-syntax
               typeof responseObj['content'] === 'string'
             ) {
               contentStr = responseObj['content'];
@@ -254,6 +254,11 @@ export class ChatCompressionService {
       };
     }
 
+    // Fire PreCompress hook before compression
+    // This fires for both manual and auto compression attempts
+    const trigger = force ? PreCompressTrigger.Manual : PreCompressTrigger.Auto;
+    await config.getHookSystem()?.firePreCompressEvent(trigger);
+
     const originalTokenCount = chat.getLastPromptTokenCount();
 
     // Don't compress if not forced and we are under the limit.
@@ -273,63 +278,6 @@ export class ChatCompressionService {
       }
     }
 
-    // Fire PreCompress hook — only when compression will actually proceed
-    const trigger = force ? PreCompressTrigger.Manual : PreCompressTrigger.Auto;
-
-    // Serialize history for the hook: strip non-text parts to keep payload manageable
-    const curatedForHook = curatedHistory.map((c) => ({
-      role: c.role ?? 'user',
-      parts: (c.parts ?? [])
-        .filter((p): p is { text: string } => typeof p.text === 'string')
-        .map((p) => ({ text: p.text })),
-    }));
-
-    const hookResult = await config
-      .getHookSystem()
-      ?.firePreCompressEvent(trigger, curatedForHook);
-
-    // If a hook provided replacement history, use it and skip built-in compression
-    const hookNewHistory =
-      hookResult?.finalOutput?.hookSpecificOutput?.['newHistory'];
-    if (Array.isArray(hookNewHistory) && hookNewHistory.length > 0) {
-      // Convert hook output back to Content[]
-      const replacementHistory: Content[] = hookNewHistory.map(
-        (entry: { role?: string; parts?: Array<{ text?: string }> }) => {
-          const role =
-            entry.role === 'model' || entry.role === 'user'
-              ? entry.role
-              : 'user';
-          return {
-            role,
-            parts: (entry.parts ?? []).map((p: { text?: string }) => ({
-              text: p.text ?? '',
-            })),
-          };
-        },
-      );
-
-      const newTokenCount = estimateTokenCountSync(
-        replacementHistory.flatMap((c) => c.parts || []),
-      );
-
-      logChatCompression(
-        config,
-        makeChatCompressionEvent({
-          tokens_before: originalTokenCount,
-          tokens_after: newTokenCount,
-        }),
-      );
-
-      return {
-        newHistory: replacementHistory,
-        info: {
-          originalTokenCount,
-          newTokenCount,
-          compressionStatus: CompressionStatus.HOOK_REPLACED,
-        },
-      };
-    }
-
     // Apply token-based truncation to the entire history before splitting.
     // This ensures that even the "to compress" portion is within safe limits for the summarization model.
     const truncatedHistory = await truncateHistoryToBudget(

packages/core/src/services/chatRecordingService.ts

@@ -20,7 +20,6 @@ import type {
 import { debugLogger } from '../utils/debugLogger.js';
 import type { ToolResultDisplay } from '../tools/tools.js';
 import type { AgentLoopContext } from '../config/agent-loop-context.js';
-import type { SerializedScheduledItem } from './work-scheduler.js';
 
 export const SESSION_FILE_PREFIX = 'session-';
 
@@ -105,10 +104,6 @@ export interface ConversationRecord {
   directories?: string[];
   /** The kind of conversation (main agent or subagent) */
   kind?: 'main' | 'subagent';
-  /** Index into messages[] after the last compression, used to skip pre-compressed messages on resume */
-  lastCompressionIndex?: number;
-  /** Pending scheduled work items persisted for session resume */
-  scheduledWork?: SerializedScheduledItem[];
 }
 
 /**
@@ -572,23 +567,6 @@ export class ChatRecordingService {
     }
   }
 
-  /**
-   * Records pending scheduled work items to the session file.
-   * Called when the work schedule changes so items survive session restart.
-   */
-  recordScheduledWork(items: SerializedScheduledItem[]): void {
-    if (!this.conversationFile) return;
-
-    try {
-      this.updateConversation((conversation) => {
-        conversation.scheduledWork = items.length > 0 ? items : undefined;
-      });
-    } catch (error) {
-      debugLogger.error('Error saving scheduled work to chat history.', error);
-      // Don't throw - we want graceful degradation
-    }
-  }
-
   /**
    * Gets the current conversation data (for summary generation).
    */
@@ -749,17 +727,6 @@ export class ChatRecordingService {
     }
   }
 
-  /**
-   * Stamps the current end of the messages array so that future session
-   * resumes can skip the pre-compression portion of the history.
-   */
-  recordCompressionPoint(): void {
-    if (!this.conversationFile) return;
-    this.updateConversation((conversation) => {
-      conversation.lastCompressionIndex = conversation.messages.length;
-    });
-  }
-
   /**
    * Rewinds the conversation to the state just before the specified message ID.
    * All messages from (and including) the specified ID onwards are removed.
@@ -792,39 +759,37 @@ export class ChatRecordingService {
   updateMessagesFromHistory(history: readonly Content[]): void {
     if (!this.conversationFile) return;
 
-    // Build the partsMap before touching the file — skip I/O entirely when
-    // there are no tool results to sync.
-    const partsMap = new Map<string, Part[]>();
-    for (const content of history) {
-      if (content.role === 'user' && content.parts) {
-        // Find all unique call IDs in this message
-        const callIds = content.parts
-          .map((p) => p.functionResponse?.id)
-          .filter((id): id is string => !!id);
-
-        if (callIds.length === 0) continue;
-
-        // Use the first ID as a seed to capture any "leading" non-ID parts
-        // in this specific content block.
-        let currentCallId = callIds[0];
-        for (const part of content.parts) {
-          if (part.functionResponse?.id) {
-            currentCallId = part.functionResponse.id;
-          }
-
-          if (!partsMap.has(currentCallId)) {
-            partsMap.set(currentCallId, []);
-          }
-          partsMap.get(currentCallId)!.push(part);
-        }
-      }
-    }
-
-    // No tool results to update — skip file I/O entirely.
-    if (partsMap.size === 0) return;
-
     try {
       this.updateConversation((conversation) => {
+        // Create a map of tool results from the API history for quick lookup by call ID.
+        // We store the full list of parts associated with each tool call ID to preserve
+        // multi-modal data and proper trajectory structure.
+        const partsMap = new Map<string, Part[]>();
+        for (const content of history) {
+          if (content.role === 'user' && content.parts) {
+            // Find all unique call IDs in this message
+            const callIds = content.parts
+              .map((p) => p.functionResponse?.id)
+              .filter((id): id is string => !!id);
+
+            if (callIds.length === 0) continue;
+
+            // Use the first ID as a seed to capture any "leading" non-ID parts
+            // in this specific content block.
+            let currentCallId = callIds[0];
+            for (const part of content.parts) {
+              if (part.functionResponse?.id) {
+                currentCallId = part.functionResponse.id;
+              }
+
+              if (!partsMap.has(currentCallId)) {
+                partsMap.set(currentCallId, []);
+              }
+              partsMap.get(currentCallId)!.push(part);
+            }
+          }
+        }
+
         // Update the conversation records tool results if they've changed.
         for (const message of conversation.messages) {
           if (message.type === 'gemini' && message.toolCalls) {

packages/core/src/services/sandboxManager.test.ts

@@ -12,6 +12,7 @@ import {
   createSandboxManager,
 } from './sandboxManager.js';
 import { LinuxSandboxManager } from '../sandbox/linux/LinuxSandboxManager.js';
+import { MacOsSandboxManager } from '../sandbox/macos/MacOsSandboxManager.js';
 
 describe('NoopSandboxManager', () => {
   const sandboxManager = new NoopSandboxManager();
@@ -124,23 +125,20 @@ describe('createSandboxManager', () => {
     expect(manager).toBeInstanceOf(NoopSandboxManager);
   });
 
-  it('should return LinuxSandboxManager if sandboxing is enabled and platform is linux', () => {
-    const osSpy = vi.spyOn(os, 'platform').mockReturnValue('linux');
-    try {
-      const manager = createSandboxManager(true, '/workspace');
-      expect(manager).toBeInstanceOf(LinuxSandboxManager);
-    } finally {
-      osSpy.mockRestore();
-    }
-  });
-
-  it('should return LocalSandboxManager if sandboxing is enabled and platform is not linux', () => {
-    const osSpy = vi.spyOn(os, 'platform').mockReturnValue('darwin');
-    try {
-      const manager = createSandboxManager(true, '/workspace');
-      expect(manager).toBeInstanceOf(LocalSandboxManager);
-    } finally {
-      osSpy.mockRestore();
-    }
-  });
+  it.each([
+    { platform: 'linux', expected: LinuxSandboxManager },
+    { platform: 'darwin', expected: MacOsSandboxManager },
+    { platform: 'win32', expected: LocalSandboxManager },
+  ] as const)(
+    'should return $expected.name if sandboxing is enabled and platform is $platform',
+    ({ platform, expected }) => {
+      const osSpy = vi.spyOn(os, 'platform').mockReturnValue(platform);
+      try {
+        const manager = createSandboxManager(true, '/workspace');
+        expect(manager).toBeInstanceOf(expected);
+      } finally {
+        osSpy.mockRestore();
+      }
+    },
+  );
 });

packages/core/src/services/sandboxManager.ts

@@ -11,6 +11,7 @@ import {
   type EnvironmentSanitizationConfig,
 } from './environmentSanitization.js';
 import { LinuxSandboxManager } from '../sandbox/linux/LinuxSandboxManager.js';
+import { MacOsSandboxManager } from '../sandbox/macos/MacOsSandboxManager.js';
 
 /**
  * Request for preparing a command to run in a sandbox.
@@ -98,6 +99,9 @@ export function createSandboxManager(
     if (os.platform() === 'linux') {
       return new LinuxSandboxManager({ workspace });
     }
+    if (os.platform() === 'darwin') {
+      return new MacOsSandboxManager({ workspace });
+    }
     return new LocalSandboxManager();
   }
   return new NoopSandboxManager();

packages/core/src/services/work-scheduler.test.ts

@@ -1,498 +0,0 @@
-/**
- * @license
- * Copyright 2026 Google LLC
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
-import { WorkScheduler } from './work-scheduler.js';
-import type { SerializedScheduledItem } from './work-scheduler.js';
-
-describe('WorkScheduler', () => {
-  let scheduler: WorkScheduler;
-
-  beforeEach(() => {
-    vi.useFakeTimers();
-    scheduler = new WorkScheduler();
-  });
-
-  afterEach(() => {
-    scheduler.dispose();
-    vi.useRealTimers();
-  });
-
-  describe('add', () => {
-    it('should add an item and return it with pending status', () => {
-      const fireAt = new Date(Date.now() + 60_000);
-      const item = scheduler.add('do something', fireAt);
-
-      expect(item.id).toBeDefined();
-      expect(item.prompt).toBe('do something');
-      expect(item.fireAt).toEqual(fireAt);
-      expect(item.status).toBe('pending');
-      expect(item.createdAt).toBeInstanceOf(Date);
-    });
-
-    it('should keep items sorted by fireAt', () => {
-      const now = Date.now();
-      scheduler.add('third', new Date(now + 300_000));
-      scheduler.add('first', new Date(now + 60_000));
-      scheduler.add('second', new Date(now + 120_000));
-
-      const pending = scheduler.getPendingItems();
-      expect(pending).toHaveLength(3);
-      expect(pending[0].prompt).toBe('first');
-      expect(pending[1].prompt).toBe('second');
-      expect(pending[2].prompt).toBe('third');
-    });
-
-    it('should emit changed event on add', () => {
-      const changedSpy = vi.fn();
-      scheduler.on('changed', changedSpy);
-
-      scheduler.add('test', new Date(Date.now() + 60_000));
-
-      expect(changedSpy).toHaveBeenCalledTimes(1);
-    });
-
-    it('should fire past-due items immediately on add', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      const pastDate = new Date(Date.now() - 10_000);
-      const item = scheduler.add('overdue task', pastDate);
-
-      expect(fireSpy).toHaveBeenCalledWith('overdue task');
-      expect(item.status).toBe('fired');
-    });
-
-    it('should rearm timer when adding an item sooner than the current next', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      scheduler.addRelative('ten min', 10);
-      scheduler.addRelative('five min', 5);
-
-      vi.advanceTimersByTime(5 * 60_000);
-      expect(fireSpy).toHaveBeenCalledTimes(1);
-      expect(fireSpy).toHaveBeenCalledWith('five min');
-
-      vi.advanceTimersByTime(5 * 60_000);
-      expect(fireSpy).toHaveBeenCalledTimes(2);
-      expect(fireSpy).toHaveBeenCalledWith('ten min');
-    });
-  });
-
-  describe('addRelative', () => {
-    it('should add an item relative to current time', () => {
-      const now = Date.now();
-      const item = scheduler.addRelative('in 5 minutes', 5);
-
-      expect(item.fireAt.getTime()).toBe(now + 5 * 60_000);
-      expect(item.status).toBe('pending');
-    });
-
-    it('should appear in pending items', () => {
-      scheduler.addRelative('relative item', 10);
-
-      const pending = scheduler.getPendingItems();
-      expect(pending).toHaveLength(1);
-      expect(pending[0].prompt).toBe('relative item');
-    });
-  });
-
-  describe('cancel', () => {
-    it('should cancel a pending item and return true', () => {
-      const item = scheduler.add('cancel me', new Date(Date.now() + 60_000));
-
-      const result = scheduler.cancel(item.id);
-
-      expect(result).toBe(true);
-      expect(scheduler.getPendingItems()).toHaveLength(0);
-    });
-
-    it('should return false for a non-existent id', () => {
-      const result = scheduler.cancel('non-existent-id');
-
-      expect(result).toBe(false);
-    });
-
-    it('should return false for an already cancelled item', () => {
-      const item = scheduler.add('cancel me', new Date(Date.now() + 60_000));
-      scheduler.cancel(item.id);
-
-      const result = scheduler.cancel(item.id);
-
-      expect(result).toBe(false);
-    });
-
-    it('should emit changed event on successful cancel', () => {
-      const item = scheduler.add('cancel me', new Date(Date.now() + 60_000));
-      const changedSpy = vi.fn();
-      scheduler.on('changed', changedSpy);
-
-      scheduler.cancel(item.id);
-
-      expect(changedSpy).toHaveBeenCalledTimes(1);
-    });
-
-    it('should not emit changed event on failed cancel', () => {
-      const changedSpy = vi.fn();
-      scheduler.on('changed', changedSpy);
-
-      scheduler.cancel('non-existent-id');
-
-      expect(changedSpy).not.toHaveBeenCalled();
-    });
-
-    it('should rearm timer to next item when cancelling the currently armed item', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      const first = scheduler.addRelative('five min', 5);
-      scheduler.addRelative('ten min', 10);
-
-      scheduler.cancel(first.id);
-
-      vi.advanceTimersByTime(5 * 60_000);
-      expect(fireSpy).not.toHaveBeenCalled();
-
-      vi.advanceTimersByTime(5 * 60_000);
-      expect(fireSpy).toHaveBeenCalledTimes(1);
-      expect(fireSpy).toHaveBeenCalledWith('ten min');
-    });
-  });
-
-  describe('getPendingItems', () => {
-    it('should return empty array when no items', () => {
-      expect(scheduler.getPendingItems()).toHaveLength(0);
-    });
-
-    it('should exclude cancelled and fired items', () => {
-      const item1 = scheduler.add('keep', new Date(Date.now() + 60_000));
-      scheduler.add('cancel', new Date(Date.now() + 120_000));
-      scheduler.add('also keep', new Date(Date.now() + 180_000));
-
-      scheduler.cancel(scheduler.getPendingItems()[1].id);
-
-      const pending = scheduler.getPendingItems();
-      expect(pending).toHaveLength(2);
-      expect(pending[0].id).toBe(item1.id);
-      expect(pending[1].prompt).toBe('also keep');
-    });
-  });
-
-  describe('getNextPending', () => {
-    it('should return undefined when no items', () => {
-      expect(scheduler.getNextPending()).toBeUndefined();
-    });
-
-    it('should return the soonest pending item', () => {
-      const now = Date.now();
-      scheduler.add('later', new Date(now + 120_000));
-      scheduler.add('sooner', new Date(now + 60_000));
-
-      const next = scheduler.getNextPending();
-      expect(next?.prompt).toBe('sooner');
-    });
-  });
-
-  describe('fire event', () => {
-    it('should emit fire event when timer expires', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      scheduler.addRelative('fire me', 5);
-
-      expect(fireSpy).not.toHaveBeenCalled();
-
-      vi.advanceTimersByTime(5 * 60_000);
-
-      expect(fireSpy).toHaveBeenCalledWith('fire me');
-    });
-
-    it('should emit changed event when item fires via timer', () => {
-      scheduler.addRelative('fire me', 5);
-      const changedSpy = vi.fn();
-      scheduler.on('changed', changedSpy);
-
-      vi.advanceTimersByTime(5 * 60_000);
-
-      expect(changedSpy).toHaveBeenCalled();
-    });
-
-    it('should fire multiple items in sequence as their timers expire', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      scheduler.addRelative('first', 1);
-      scheduler.addRelative('second', 2);
-
-      vi.advanceTimersByTime(60_000);
-      expect(fireSpy).toHaveBeenCalledTimes(1);
-      expect(fireSpy).toHaveBeenCalledWith('first');
-
-      vi.advanceTimersByTime(60_000);
-      expect(fireSpy).toHaveBeenCalledTimes(2);
-      expect(fireSpy).toHaveBeenCalledWith('second');
-    });
-
-    it('should not fire cancelled items', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      const item = scheduler.addRelative('cancel me', 5);
-      scheduler.cancel(item.id);
-
-      vi.advanceTimersByTime(5 * 60_000);
-
-      expect(fireSpy).not.toHaveBeenCalled();
-    });
-  });
-
-  describe('serialize', () => {
-    it('should serialize pending items to ISO string format', () => {
-      const now = Date.now();
-      scheduler.add('item 1', new Date(now + 60_000));
-      scheduler.add('item 2', new Date(now + 120_000));
-
-      const serialized = scheduler.serialize();
-
-      expect(serialized).toHaveLength(2);
-      expect(serialized[0].prompt).toBe('item 1');
-      expect(serialized[0].fireAt).toBe(new Date(now + 60_000).toISOString());
-      expect(serialized[0].id).toBeDefined();
-      expect(serialized[0].createdAt).toBeDefined();
-      expect(serialized[1].prompt).toBe('item 2');
-    });
-
-    it('should not include cancelled items', () => {
-      const item = scheduler.add('cancel me', new Date(Date.now() + 60_000));
-      scheduler.add('keep me', new Date(Date.now() + 120_000));
-      scheduler.cancel(item.id);
-
-      const serialized = scheduler.serialize();
-
-      expect(serialized).toHaveLength(1);
-      expect(serialized[0].prompt).toBe('keep me');
-    });
-
-    it('should return empty array when no pending items', () => {
-      expect(scheduler.serialize()).toEqual([]);
-    });
-
-    it('should exclude fired items', () => {
-      scheduler.addRelative('will fire', 5);
-
-      vi.advanceTimersByTime(5 * 60_000);
-
-      const serialized = scheduler.serialize();
-      expect(serialized).toEqual([]);
-    });
-  });
-
-  describe('restore', () => {
-    it('should restore items from serialized data', () => {
-      const now = Date.now();
-      const serialized: SerializedScheduledItem[] = [
-        {
-          id: 'restored-1',
-          prompt: 'restored item',
-          fireAt: new Date(now + 60_000).toISOString(),
-          createdAt: new Date(now - 10_000).toISOString(),
-        },
-      ];
-
-      scheduler.restore(serialized);
-
-      const pending = scheduler.getPendingItems();
-      expect(pending).toHaveLength(1);
-      expect(pending[0].id).toBe('restored-1');
-      expect(pending[0].prompt).toBe('restored item');
-      expect(pending[0].status).toBe('pending');
-    });
-
-    it('should emit changed event on restore', () => {
-      const changedSpy = vi.fn();
-      scheduler.on('changed', changedSpy);
-
-      scheduler.restore([
-        {
-          id: 'id-1',
-          prompt: 'test',
-          fireAt: new Date(Date.now() + 60_000).toISOString(),
-          createdAt: new Date().toISOString(),
-        },
-      ]);
-
-      expect(changedSpy).toHaveBeenCalledTimes(1);
-    });
-
-    it('should fire past-due items immediately on restore', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      const serialized: SerializedScheduledItem[] = [
-        {
-          id: 'past-due-1',
-          prompt: 'overdue restored',
-          fireAt: new Date(Date.now() - 30_000).toISOString(),
-          createdAt: new Date(Date.now() - 120_000).toISOString(),
-        },
-      ];
-
-      scheduler.restore(serialized);
-
-      expect(fireSpy).toHaveBeenCalledWith('overdue restored');
-    });
-
-    it('should handle round-trip serialize and restore', () => {
-      const now = Date.now();
-      scheduler.add('round trip 1', new Date(now + 60_000));
-      scheduler.add('round trip 2', new Date(now + 120_000));
-
-      const serialized = scheduler.serialize();
-
-      const newScheduler = new WorkScheduler();
-      newScheduler.restore(serialized);
-
-      const pending = newScheduler.getPendingItems();
-      expect(pending).toHaveLength(2);
-      expect(pending[0].prompt).toBe('round trip 1');
-      expect(pending[1].prompt).toBe('round trip 2');
-
-      newScheduler.dispose();
-    });
-
-    it('should sort restored items by fireAt', () => {
-      const now = Date.now();
-      const serialized: SerializedScheduledItem[] = [
-        {
-          id: 'later',
-          prompt: 'later',
-          fireAt: new Date(now + 120_000).toISOString(),
-          createdAt: new Date().toISOString(),
-        },
-        {
-          id: 'sooner',
-          prompt: 'sooner',
-          fireAt: new Date(now + 60_000).toISOString(),
-          createdAt: new Date().toISOString(),
-        },
-      ];
-
-      scheduler.restore(serialized);
-
-      const pending = scheduler.getPendingItems();
-      expect(pending[0].prompt).toBe('sooner');
-      expect(pending[1].prompt).toBe('later');
-    });
-
-    it('should handle mix of past-due and future items', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      const now = Date.now();
-      const serialized: SerializedScheduledItem[] = [
-        {
-          id: 'past-due',
-          prompt: 'overdue',
-          fireAt: new Date(now - 30_000).toISOString(),
-          createdAt: new Date(now - 120_000).toISOString(),
-        },
-        {
-          id: 'future',
-          prompt: 'upcoming',
-          fireAt: new Date(now + 5 * 60_000).toISOString(),
-          createdAt: new Date(now - 60_000).toISOString(),
-        },
-      ];
-
-      scheduler.restore(serialized);
-
-      expect(fireSpy).toHaveBeenCalledTimes(1);
-      expect(fireSpy).toHaveBeenCalledWith('overdue');
-
-      const pending = scheduler.getPendingItems();
-      expect(pending).toHaveLength(1);
-      expect(pending[0].prompt).toBe('upcoming');
-
-      vi.advanceTimersByTime(5 * 60_000);
-      expect(fireSpy).toHaveBeenCalledTimes(2);
-      expect(fireSpy).toHaveBeenCalledWith('upcoming');
-    });
-  });
-
-  describe('formatScheduleSummary', () => {
-    it('should include current time and timezone', () => {
-      const summary = scheduler.formatScheduleSummary();
-
-      expect(summary).toContain('Current time:');
-      const tz = Intl.DateTimeFormat().resolvedOptions().timeZone;
-      expect(summary).toContain(tz);
-    });
-
-    it('should show no scheduled items message when empty', () => {
-      const summary = scheduler.formatScheduleSummary();
-
-      expect(summary).toContain('No scheduled items.');
-    });
-
-    it('should list all pending items', () => {
-      scheduler.add('first task', new Date(Date.now() + 60_000));
-      scheduler.add('second task', new Date(Date.now() + 120_000));
-
-      const summary = scheduler.formatScheduleSummary();
-
-      expect(summary).toContain('Active schedule:');
-      expect(summary).toContain('first task');
-      expect(summary).toContain('second task');
-      expect(summary).toContain('1.');
-      expect(summary).toContain('2.');
-    });
-
-    it('should show truncated item IDs', () => {
-      const item = scheduler.add('test', new Date(Date.now() + 60_000));
-
-      const summary = scheduler.formatScheduleSummary();
-
-      expect(summary).toContain(`[${item.id.slice(0, 8)}]`);
-    });
-
-    it('should not contain ISO UTC format (toISOString)', () => {
-      scheduler.add('check format', new Date(Date.now() + 60_000));
-
-      const summary = scheduler.formatScheduleSummary();
-
-      expect(summary).not.toMatch(/\.\d{3}Z/);
-    });
-  });
-
-  describe('dispose', () => {
-    it('should prevent items from firing after dispose', () => {
-      const fireSpy = vi.fn();
-      scheduler.on('fire', fireSpy);
-
-      scheduler.addRelative('should not fire', 5);
-      scheduler.dispose();
-
-      vi.advanceTimersByTime(5 * 60_000);
-
-      expect(fireSpy).not.toHaveBeenCalled();
-    });
-
-    it('should be safe to call dispose multiple times', () => {
-      scheduler.dispose();
-      scheduler.dispose();
-      // No error thrown
-    });
-
-    it('should not throw when adding items after dispose', () => {
-      scheduler.dispose();
-
-      expect(() => {
-        scheduler.add('after dispose', new Date(Date.now() + 60_000));
-      }).not.toThrow();
-    });
-  });
-});

packages/core/src/services/work-scheduler.ts

@@ -1,223 +0,0 @@
-/**
- * @license
- * Copyright 2026 Google LLC
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { randomUUID } from 'node:crypto';
-import { EventEmitter } from 'node:events';
-
-/**
- * A single scheduled work item.
- */
-export interface ScheduledItem {
-  id: string;
-  prompt: string;
-  fireAt: Date;
-  createdAt: Date;
-  status: 'pending' | 'fired' | 'cancelled';
-}
-
-/**
- * Serializable representation of a scheduled item for persistence.
- */
-export interface SerializedScheduledItem {
-  id: string;
-  prompt: string;
-  fireAt: string;
-  createdAt: string;
-}
-
-export interface WorkSchedulerEvents {
-  fire: [prompt: string];
-  changed: [];
-}
-
-/**
- * Manages a time-based list of scheduled work items.
- * Emits 'fire' when a scheduled item's time arrives, with the prompt text.
- * Emits 'changed' whenever the schedule is mutated.
- */
-export class WorkScheduler extends EventEmitter<WorkSchedulerEvents> {
-  private items: ScheduledItem[] = [];
-  private timer: ReturnType<typeof setTimeout> | null = null;
-
-  /**
-   * Add a scheduled item at an absolute time.
-   * @returns The created ScheduledItem.
-   */
-  add(prompt: string, fireAt: Date): ScheduledItem {
-    const item: ScheduledItem = {
-      id: randomUUID(),
-      prompt,
-      fireAt,
-      createdAt: new Date(),
-      status: 'pending',
-    };
-    this.items.push(item);
-    this.items.sort((a, b) => a.fireAt.getTime() - b.fireAt.getTime());
-    this.rearm();
-    this.emit('changed');
-    return item;
-  }
-
-  /**
-   * Add a scheduled item using a relative delay in minutes.
-   * @returns The created ScheduledItem.
-   */
-  addRelative(prompt: string, inMinutes: number): ScheduledItem {
-    const fireAt = new Date(Date.now() + inMinutes * 60 * 1000);
-    return this.add(prompt, fireAt);
-  }
-
-  /**
-   * Cancel a scheduled item by ID.
-   * @returns true if the item was found and cancelled, false otherwise.
-   */
-  cancel(id: string): boolean {
-    const item = this.items.find((i) => i.id === id && i.status === 'pending');
-    if (!item) {
-      return false;
-    }
-    item.status = 'cancelled';
-    this.rearm();
-    this.emit('changed');
-    return true;
-  }
-
-  /**
-   * Get all pending items, sorted by fireAt.
-   */
-  getPendingItems(): readonly ScheduledItem[] {
-    return this.items.filter((i) => i.status === 'pending');
-  }
-
-  /**
-   * Get the next pending item (soonest fireAt).
-   */
-  getNextPending(): ScheduledItem | undefined {
-    return this.items.find((i) => i.status === 'pending');
-  }
-
-  /**
-   * Serialize pending items for session persistence.
-   */
-  serialize(): SerializedScheduledItem[] {
-    return this.getPendingItems().map((item) => ({
-      id: item.id,
-      prompt: item.prompt,
-      fireAt: item.fireAt.toISOString(),
-      createdAt: item.createdAt.toISOString(),
-    }));
-  }
-
-  /**
-   * Restore scheduled items from serialized data (e.g. session resume).
-   * Items whose fireAt is in the past fire immediately (queued sequentially).
-   * Items in the future get timers re-armed.
-   */
-  restore(serialized: SerializedScheduledItem[]): void {
-    for (const s of serialized) {
-      const item: ScheduledItem = {
-        id: s.id,
-        prompt: s.prompt,
-        fireAt: new Date(s.fireAt),
-        createdAt: new Date(s.createdAt),
-        status: 'pending',
-      };
-      this.items.push(item);
-    }
-    this.items.sort((a, b) => a.fireAt.getTime() - b.fireAt.getTime());
-    this.rearm();
-    this.emit('changed');
-  }
-
-  /**
-   * Format a human-readable summary of the current schedule, including current time.
-   */
-  formatScheduleSummary(): string {
-    const now = new Date();
-    const pending = this.getPendingItems();
-    const tz = Intl.DateTimeFormat().resolvedOptions().timeZone;
-
-    const localTime = now.toLocaleString([], {
-      year: 'numeric',
-      month: '2-digit',
-      day: '2-digit',
-      hour: '2-digit',
-      minute: '2-digit',
-      second: '2-digit',
-    });
-    const lines: string[] = [`Current time: ${localTime} (${tz})`];
-
-    if (pending.length === 0) {
-      lines.push('No scheduled items.');
-    } else {
-      lines.push('');
-      lines.push('Active schedule:');
-      for (let i = 0; i < pending.length; i++) {
-        const item = pending[i];
-        const diffMs = item.fireAt.getTime() - now.getTime();
-        const diffMins = Math.max(0, Math.ceil(diffMs / 60000));
-        const itemTime = item.fireAt.toLocaleTimeString([], {
-          hour: '2-digit',
-          minute: '2-digit',
-        });
-        lines.push(
-          `  ${i + 1}. [${item.id.slice(0, 8)}] ${itemTime} (in ${diffMins}m) — "${item.prompt}"`,
-        );
-      }
-    }
-    return lines.join('\n');
-  }
-
-  /**
-   * Stop all timers. Call on cleanup/shutdown.
-   */
-  dispose(): void {
-    if (this.timer) {
-      clearTimeout(this.timer);
-      this.timer = null;
-    }
-  }
-
-  /**
-   * Re-arm the internal timer to point at the next pending item.
-   */
-  private rearm(): void {
-    if (this.timer) {
-      clearTimeout(this.timer);
-      this.timer = null;
-    }
-
-    // Fire all past-due items immediately
-    const now = Date.now();
-    const pastDue = this.items.filter(
-      (i) => i.status === 'pending' && i.fireAt.getTime() <= now,
-    );
-    for (const item of pastDue) {
-      item.status = 'fired';
-      this.emit('fire', item.prompt);
-    }
-
-    // Find next future pending item
-    const next = this.items.find(
-      (i) => i.status === 'pending' && i.fireAt.getTime() > now,
-    );
-    if (!next) {
-      return;
-    }
-
-    const delayMs = Math.max(0, next.fireAt.getTime() - Date.now());
-    this.timer = setTimeout(() => {
-      this.timer = null;
-      if (next.status === 'pending') {
-        next.status = 'fired';
-        this.emit('fire', next.prompt);
-        this.emit('changed');
-      }
-      // Re-arm for the next item after this one
-      this.rearm();
-    }, delayMs);
-  }
-}

packages/core/src/tools/exit-plan-mode.test.ts

@@ -47,6 +47,7 @@ describe('ExitPlanModeTool', () => {
       storage: {
         getPlansDir: vi.fn().mockReturnValue(mockPlansDir),
       } as unknown as Config['storage'],
+      isInteractive: vi.fn().mockReturnValue(true),
     };
     tool = new ExitPlanModeTool(
       mockConfig as Config,
@@ -359,6 +360,36 @@ Ask the user for specific feedback on how to improve the plan.`,
     });
   });
 
+  describe('getAllowApprovalMode (internal)', () => {
+    it('should return YOLO when config.isInteractive() is false', async () => {
+      mockConfig.isInteractive = vi.fn().mockReturnValue(false);
+      const planRelativePath = createPlanFile('test.md', '# Content');
+      const invocation = tool.build({ plan_path: planRelativePath });
+
+      // Directly call execute to trigger the internal getAllowApprovalMode
+      const result = await invocation.execute(new AbortController().signal);
+
+      expect(result.llmContent).toContain('YOLO mode');
+      expect(mockConfig.setApprovalMode).toHaveBeenCalledWith(
+        ApprovalMode.YOLO,
+      );
+    });
+
+    it('should return DEFAULT when config.isInteractive() is true', async () => {
+      mockConfig.isInteractive = vi.fn().mockReturnValue(true);
+      const planRelativePath = createPlanFile('test.md', '# Content');
+      const invocation = tool.build({ plan_path: planRelativePath });
+
+      // Directly call execute to trigger the internal getAllowApprovalMode
+      const result = await invocation.execute(new AbortController().signal);
+
+      expect(result.llmContent).toContain('Default mode');
+      expect(mockConfig.setApprovalMode).toHaveBeenCalledWith(
+        ApprovalMode.DEFAULT,
+      );
+    });
+  });
+
   describe('getApprovalModeDescription (internal)', () => {
     it('should handle all valid approval modes', async () => {
       const planRelativePath = createPlanFile('test.md', '# Content');
@@ -387,6 +418,10 @@ Ask the user for specific feedback on how to improve the plan.`,
         ApprovalMode.DEFAULT,
         'Default mode (edits will require confirmation)',
       );
+      await testMode(
+        ApprovalMode.YOLO,
+        'YOLO mode (all tool calls auto-approved)',
+      );
     });
 
     it('should throw for invalid post-planning modes', async () => {
@@ -409,7 +444,6 @@ Ask the user for specific feedback on how to improve the plan.`,
         ).rejects.toThrow(/Unexpected approval mode/);
       };
 
-      await testInvalidMode(ApprovalMode.YOLO);
       await testInvalidMode(ApprovalMode.PLAN);
     });
   });