moonlight-lupin/agent-skills / plugins/dev-pipeline/skills/review-lessons
review-lessons
VerifiedLibrary of lessons from past build-review loops — not a stage. Use when a review or fix round hits a specific case: a reviewer BLOCKING that contradicts the brief, a test that passes but checks nothing, a feature dead in production, SQLite/WAL migrations, MCP tool surfaces, REST API authz, app ↔ external-source sync, Android/APK verification, fill-template extraction, CLI/git mechanics, reviewer outages. Load the one file the case needs; never load the whole library.
references/evidence-pairs.md
View on GitHub# Evidence Pairs
Every delivered batch carries an **evidence pair**: the state before the
change and the state after, captured the same way. The pair is embedded in
the delivery — PR body, commit message thread, or final report — not stored
beside it. A claim about the change without its pair is unverified prose.
Ported concept (Sep 2026) from michaelshimeles/skills "prove" beat
(evidence-driven-testing). See *Adaptation* below for what was NOT taken.
## Capture rules
- Capture the **before** while reproducing the issue — before the fix, when
it is cheapest. A before captured after the fix is a reconstruction, not
evidence.
- Capture **after** with the same method as the before. A screenshot vs a
screenshot; a measured number vs the same measurement re-run. Method
change = invalid pair.
- No visual surface? Use **measured numbers or output pairs** instead:
the failing check output vs the passing one, the slow timing vs the fast
one, the wrong total vs the right one.
- Keep pairs **local** (workspace, `/tmp`, repo `docs/` when meant to
ship). Never upload to public hosts — screenshots of customer or
internal surfaces are sensitive.
## Where the pair lives
| Delivery | How the pair is embedded |
|---|---|
| PR | Markdown table `\| Before \| After \|` in the body; images via repo-relative paths or uploaded assets |
| Delivery message | Before/after blocks side by side |
| Commit for a small fix | Before/after output pasted in the message body |
The loop's dual review (Codex + Opus) judges the diff; the pair is for the
human — it answers "does the change do what it claims" in one look.
## Completion criterion
A batch's delivery step is done only when every user-visible claim in it
has a pair beside it. `--gate` open + attestation verified + evidence pairs
embedded = deliverable.
## Adaptation (source: michaelshimeles/skills)
Ported: the before/after pair as a delivery requirement, and the
"capture before while reproducing" timing rule.
NOT adopted: their `@vercel/before-and-after` CLI (uploads to a public
host by default — 0x0.st; unacceptable for production/customer surfaces),
`greploop` (needs Greptile; our Codex+Opus dual review is stricter),
`unslop` (controlled-English is our writing standard), `new-feature`
worktree flow (our parallel dispatches use disjoint brief ownership,
pitfall (parallel dispatches); worktree isolation remains an option for true conflicts),
`code-structure` service-layer skill (already covered by our
architecture review checklist).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |