Verify before done
Skill charlieviettq/awesome-agent-skill/.claude/skills/verify-before-done
Curated skill pack for LLM agents in engineer and science workflow (Cursor & Claude ready).
npx -y skills add charlieviettq/awesome-agent-skill --skill verify-before-doneAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 22 stars22 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Require fresh verification evidence before claiming tests pass, builds succeed, or work is complete. Use when finishing a task, creating a PR, or reporting status.
SKILL.md
2.0 KB, 409 tokens by cl100k_base, as published. Nobody here has run it
Verify before done
Iron rule
No completion claims without fresh verification evidence in the current turn.
Gate (before claiming success)
- Identify what command proves the claim (test, lint, build, repro).
- Run the full command (not a partial check from memory).
- Read exit code and full relevant output.
- Verify output matches the claim.
- Then state the claim with evidence.
Common claims vs proof
| Claim | Requires | Not sufficient |
|---|---|---|
| Tests pass | Test command, 0 failures | Previous run, "should pass" |
| Linter clean | Linter output, 0 errors | Build passed |
| Bug fixed | Repro test passes | Code changed only |
| Requirements met | Checklist vs plan | Tests pass alone |
Red flags (stop and verify)
- "Should", "probably", "seems to"
- Satisfaction before running checks
- Trusting subagent or prior message without re-running
- Partial verification only
Requirements met (not tests alone)
- Map plan/spec acceptance criteria to evidence (test output, screenshot, API response, manual step)
- A green test suite does not prove every requirement if scope included non-testable items
Regression discipline
For bugfixes: failing test or repro first, fix, verify pass, optionally revert fix and confirm fail again.
Branch completion (lightweight)
- Before claiming PR-ready: tests/lint/build per project norms, no known blockers, user informed of scope
- Do not discard branches, force-merge, or cleanup worktrees without explicit user approval
Verification gate inspired by obra/superpowers (MIT).
Output format
Ran: pytest tests/foo.py -q
Result: 12 passed, exit 0
Claim: Tests pass for foo module.