Adds a self-awareness mandate to the agent's planning phase:
'Before attempting a fix for a validation error, review your recent tool calls. If you are repeatedly applying similar regex replacements or edits to the same block of code without the validation error changing, you are in a loop. Stop, revert your changes to a known good state, and rethink your approach.'
This helps the agent identify and break out of unproductive loops during debugging and implementation.
Adds a strict retry threshold to the agent's validation loop:
'If validation fails 3 times on the exact same test or error, DO NOT attempt another minor code tweak. You must immediately step back, use search tools to gather wider context, and formulate a completely new strategy.'
This prevents the agent from getting stuck in repetitive, unsuccessful minor tweaks and encourages a more strategic approach when initial fixes fail.
Introduces three critical mandates to the agent's testing and validation workflow:
1. **Prioritize Existing Infrastructure:** Strictly prefer running the project's existing test suite over writing custom reproduction scripts to avoid environment/import difficulties.
2. **Timebox Test Setup:** Abandon custom reproduction scripts if they fail to set up within 3-5 turns due to environment or import errors, falling back to static analysis and built-in tests.
3. **Mandate Exhaustive Validation:** Explicitly requires running relevant existing project tests to prevent regressions, ensuring a passing custom reproduction script is treated as a necessary but not sufficient condition for completion.
These changes prevent 'Early Exhaustion' by reducing the complexity of standalone test setup in frameworks like Django.
This commit upgrades the agent with a robust self-validation workflow while
ensuring 100% semantic and verbatim coverage of the original system prompt.
By moving to an additive model, we preserve the original reasoning anchors
(lead-ins, heuristics, and formatting) while injecting critical autonomous
engineering mandates.
Self-Validation Workflow Injections:
- Research Phase: Parallel Discovery (combining manifests/logic) and High-Signal Grep.
- Bug Fixing: Negative Verification (confirming repro failure) and Coverage Expansion.
- Implementation: Transactional Edits (logical batching of module changes).
- Validation Loop: Tiered Validation (Fixers -> Fast-Path -> Related Tests) and Smart Log Navigation.
Technical Verification:
- Verbatim restoration verified against 66 core tests and 14 snapshots.
- New behavioral eval suite passed (evals/self_validation_workflow.eval.ts).
- Full 'npm run preflight' validation successful.
Major upgrade to the agent's self-validation, safety, and project integrity
capabilities through five iterations of system prompt enhancements:
Workflow & Quality Mandates:
1. Incremental Validation: Mandates building, linting, and testing after
every significant file change to maintain a "green" state.
2. Mandatory Reproduction: Requires creating a failing test case to confirm
a bug before fixing, and explicitly verifying the failure (Negative Verification).
3. Test Persistence & Locality: Requires integrating repro cases into the
permanent test suite, preferably by amending existing related test files.
4. Script Discovery: Mandates identifying project-specific validation
commands from configuration files (package.json, Makefile, etc.).
5. Self-Review: Mandates running `git diff` after every edit, using
`--name-only` for large changes to preserve context window tokens.
6. Fast-Path Validation: Prioritizes lightweight checks (e.g., `tsc --noEmit`)
for frequent feedback, reserving heavy builds for final verification.
7. Output Verification: Requires checking command output (not just exit codes)
to prevent false-positives from empty test runs or hidden warnings.
Semantic Integrity & Dependency Safety:
8. Global Usage Discovery: Mandates searching the entire workspace for all
usages (via `grep_search`) before modifying exported symbols or APIs.
9. Dependency Integrity: Requires verifying that new imports are explicitly
declared in the project's dependency manifest (e.g., package.json).
10. Configuration Sync: Mandates updating build/environment configs
(tsconfig, Dockerfile, etc.) to support new file types or entry points.
11. Documentation Sync: Requires searching for and updating documentation
references when public APIs or CLI interfaces change.
12. Anti-Silencing Mandate: Prohibits using `any`, `@ts-ignore`, or lint
suppressions to resolve validation errors.
Diagnostics, Safety & Runtime Verification:
13. Error Grounding: Mandates reading full error logs and stack traces upon
failure. Includes Smart Log Navigation to prioritize the tail of large files.
14. Scope Isolation: Instructs the agent to focus only on errors introduced
by its changes and ignore unrelated legacy technical debt.
15. Destructive Safety: Mandates a `git status` check before deleting files
or modifying critical project configurations.
16. Non-Blocking Smoke Tests: Requires briefly running applications to
verify boot stability, using background/timeout strategies for servers.
Includes 15 new behavioral evaluations verifying these mandates and updated
snapshots in packages/core/src/core/prompts.test.ts.
Alisa Novikova