Read .claude/findings.jsonl, show pending skill findings, and apply approved changes back to the relevant skill file.
Instructions
-
Read
.claude/findings.jsonl. If the file does not exist or has no entries wherereviewed: false, say "No pending findings." and stop. -
Filter to
reviewed: falseentries. Group byskillfield. Print a summary table:
| Skill | Count | Sources |
|---------------|-------|-------------------|
| fastapi | 2 | web, gap |
| (unmatched) | 1 | web |
-
Ask the user which skill to process. Work on one skill per invocation.
-
For entries where
skillisnull: list thequery/url/notefor each, suggest the most likely skill match by comparing keywords against folder names underskills/*/skill.md, and confirm with the user before proceeding. -
Read the target
skills/<name>/skill.md. -
For each unreviewed finding for this skill, determine the minimal change:
source: "web"— new info from a search: add a checklist item, update a code example, or add a###subsection if the topic is genuinely newsource: "gap"— pattern Claude had to invent: add a code example + checklist item in the most relevant existing sectionsource: "bug"— incorrect content: show the current block and the corrected version side by side as# BEFORE/# AFTERin a fenced block
-
Show every proposed change in full before touching any file. Ask for explicit approval:
- "Apply these changes? (yes / skip / edit)"
- "skip" marks the finding reviewed without editing the skill
- "edit" lets the user paste a corrected version
-
On approval:
- Edit
skills/<name>/skill.mdin-place — never rewrite the whole file, never change frontmatter or section order - For each processed finding, update its
reviewedfield totrueinfindings.jsonl(rewrite that line only)
- Edit
-
After editing, run
/skill-validateto confirm the file still passes CI format checks. If it fails, show the error and offer to fix it before proceeding.
Format rules (enforced in every proposed change)
- Frontmatter: unchanged (
name:,description:as-is) - Checklist items:
- [ ] verb-first, present tense - Code examples for language-agnostic skills: Python, TypeScript, and Go as siblings
- Anti-patterns:
# BADcomment on the bad block,# GOODon the corrected block — always paired - No meta-commentary ("as of 2026", "Claude found this", "new finding") — write as if it was always there
Generateur de Documentation API
Documentation
Genere automatiquement de la documentation API OpenAPI/Swagger.
Rédacteur Technique
Documentation
Rédige de la documentation technique claire selon les meilleurs style guides.
Synchronisation de la documentation avec les changements de code
Documentation
Synchroniser la documentation projet avec les derniers changements de code, refactorisations et fonctionnalités de la branche courante.