QA Handoff
SkillDev toolsQuality gate protocol between implementation and handoff.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the QA Handoff skill
What this skill tells your AI
The instructions your AI receives, as published by maxim-saplin/llm_chess in .agents/skills/qa-handoff/SKILL.md and read by ahel’s review.
Use this skill after implementation work and before reporting results to the user. This file is the entrypoint: keep it for overview, routing, and shared invariants, then use the mode docs in references/ for the actual review flow.
!!! WHEN APPLICABLE PREFFER RUNTIME EVIDENCE, live checks, any real-world verfication scenarious OVER static file analysis !!!
Do Not Use When
- The task is still mid-implementation.
- You are only exploring or planning and have not produced work to verify yet.
- There is nothing substantive to hand off for review.
Shared Invariants
- Work is not reported to the user until it passes QA or 3 rounds are exhausted.
- Ambiguous verdict (neither clear PASS nor FAIL) counts as FAIL.
- The QA reviewer has no conversation history. Reconstruct the needed context in the handoff package.
- Only the user can waive a requirement.
- Before assembling or reviewing evidence, read
references/evidence-taxonomy.md. - QA validates deliverable readiness, not only internal correctness.
Choose the Mode
- Default to
references/default-qa-mode.md. - Use
references/two-qa-mode.mdonly when the user asks for a stronger gate or when you intentionally want bothqa-1andqa-2. - In two-QA mode, both reviewers must PASS on the same round before handoff.
Signals
- GitHub stars
- 127
- Forks
- 14
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
qa-handoff- Source
- github.com/maxim-saplin/llm_chess