skill-reviewerlisted

# Skill Reviewer Compile skills, lint them, review against best practices from Claude/Codex/OpenCode, and produce a prioritized report. ## Process ### Step 1: Compile fresh skills ```bash python src/compile.py ``` Abort if compilation fails -- fix errors first. ### Step 2: Run autoskill lint ```bash autoskill lint ``` Capture the output for the report. ### Step 3: Fetch best practices Fetch these URLs and extract the key rules and checklists from each: - https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices.md - https://developers.openai.com/codex/learn/best-practices.md - https://opencode.ai/docs/skills.md These are the authoritative sources. Always fetch fresh -- do not rely on cached or bundled copies. ### Step 4: Review each skill in parallel Spawn one subagent per skill directory in `./skills/`. Pass each subagent the best practices content fetched in Step 3. Each subagent: 1. Read the compiled `SKILL.md` and any files in `references/` 2. Check against every applicable rule from all three best practice sources 3. Return a list of issues, each with: - **Priority**: P1 (violates a hard requirement), P2 (violates a strong recommendation), P3 (minor improvement) - **Rule source**: which best practice doc the rule comes from - **Issue**: what's wrong - **Suggestion**: how to fix it ### Step 5: Produce report Combine all subagent results into a single markdown report. Save to `./docs/reviews/YYYY-MM-DD.md`. Report struc