meta-skill-hygiene

Skill: Meta Skill Hygiene

Context (Required)

• Scope + target files: references/context.md
• Execution checklist: templates/checklist.md

Use this skill to keep every skill asset in sync: SKILL.md, manifest.json, agents/openai.yaml, the Claude wrapper, and the registry sections in AGENTS.md / CLAUDE.md.

Workflow

1. Identify which source of truth changed:
   • SKILL.md for name and description
   • manifest.json for scope, path, and whenToUse
   • agents/openai.yaml for Codex/OpenAI metadata
2. For new skills or renames, scaffold/update the source of truth first.
3. Run bun run skills:sync to regenerate the registry and Claude wrapper.
4. Run bun run skills:validate to confirm no drift or missing files.
5. If validation fails, fix the source of truth, then sync and validate again.

Prohibitions

• NEVER treat the Claude wrapper or the generated registry sections as the source of truth.
• NEVER edit blocks between <!-- skill-registry:* --> or <!-- skill-links:* --> markers without re-running sync.
• NEVER fix drift by editing the generated output only; fix the source file.
• NEVER leave manifest.json without whenToUse for any skill.

Pre-Completion Checklist

• [ ] Source of truth for the change is clear
• [ ] bun run skills:sync run
• [ ] bun run skills:validate passes
• [ ] No drift between manifest, wrapper, and registry
• [ ] Every file ends with a newline (EOF)