verification-checklist
Use before marking any task complete to ensure all requirements are met
When & Why to Use This Skill
This Claude skill provides a rigorous verification framework and 'Definition of Done' checklist to ensure task accuracy and code quality. It automates the quality assurance process by mandating diagnostic checks, build validations, and requirement cross-referencing, preventing AI agents from delivering incomplete or buggy solutions.
Use Cases
- Code Quality Assurance: Automatically running LSP diagnostics and linters on modified files to ensure no new errors or warnings are introduced.
- Build and Test Validation: Executing project-specific build commands and test suites (e.g., npm test, cargo build) to verify functional integrity before task completion.
- Requirement Alignment: Systematically re-evaluating the final output against the user's original prompt to ensure every specific constraint and edge case has been addressed.
- Agent Workflow Standardization: Implementing a mandatory safety gate for autonomous agents to prevent premature task termination and ensure all background processes are cleaned up.
| name | verification-checklist |
|---|---|
| description | Use before marking any task complete to ensure all requirements are met |
Verification Checklist
A task is NOT complete until all verification passes. No exceptions.
Completion Criteria
- All planned todo items marked done
- Diagnostics clean on changed files
- Build passes (if applicable)
- Tests pass (if applicable)
- User's original request fully addressed
Evidence Requirements
| Action | Required Evidence |
|---|---|
| File edit | lsp_diagnostics clean on changed files |
| Build command | Exit code 0 |
| Test run | Pass (or explicit note of pre-existing failures) |
| Delegation | Agent result received and verified |
NO EVIDENCE = NOT COMPLETE.
Verification Steps
1. Run Diagnostics
lsp_diagnostics on each changed file
2. Run Build (if project has build)
npm run build / cargo build / etc.
3. Run Tests (if project has tests)
npm test / cargo test / etc.
4. Check Against Original Request
Re-read the user's original request. Does your implementation:
- Address ALL requirements mentioned?
- Handle edge cases discussed?
- Follow any constraints specified?
If Verification Fails
- Fix issues caused by YOUR changes
- Do NOT fix pre-existing issues unless asked
- Report: "Done. Note: found N pre-existing lint errors unrelated to my changes."
Pre-existing Failures
If tests/build/lint were failing BEFORE your changes:
- Document the pre-existing failures
- Ensure you didn't make them worse
- Report them separately from your work
Before Final Delivery
- Cancel ALL running background tasks:
background_cancel(all=true) - Summarize what was done
- Note any caveats or follow-up items