Azure99 Skill Creator Skill Reviewer

Skill Reviewer

Review a given skill folder against the skill-creator standards and produce an actionable report (with severities and specific fixes).

Quick start

1. Get the skill folder path (or the full SKILL.md plus a directory tree).
2. Run the automated lint (when you have filesystem access):
   • From the skill-reviewer folder: python scripts/lint_skill.py /path/to/skill
3. Review manually using the checklist and standards references.
4. Produce the review report using references/report-template.md.
5. If fixing, follow references/remediation.md, re-run lint, and update the report.

Workflow

0) Intake (ask first, keep it lightweight)

• Ask for:
  • Skill folder path
  • Intended purpose and the top 3–5 example user prompts that should trigger it
  • Any constraints (offline/no-network, no deletions, target runtime, etc.)
  • Variants (providers/frameworks/domains)
• Confirm scope (review only vs review + fix).

1) Lint first (blockers)

• Run python scripts/lint_skill.py <skill_dir> and treat errors as blockers.
• Re-run with --strict when you want warnings to fail CI-like checks.

2) Manual review (follow checklist)

• Execute references/checklist.md top-to-bottom.
• Use references/standards.md when you need to map “skill-creator rules” to concrete criteria.

3) Packaging readiness (when applicable)

• If a packager/validator exists in the environment, run it and treat failures as blockers.
  • Example: python skills/skill-creator/scripts/package_skill.py <path/to/skill>
• Otherwise rely on the automated lint plus the manual checklist.

4) Fixes (only when requested)

• Apply minimal, targeted changes.
• Re-run lint/validation after fixes.
• Update the report sections: Findings, Fixes Applied, Remaining Issues, Tests / Validation.

Guardrails

• Use imperative/infinitive voice in edits.
• Keep SKILL.md under ~500 lines; move deep/variant details to references/.
• Keep references one level deep; add a TOC for files over 100 lines.
• Do not add auxiliary docs (README, CHANGELOG, QUICK_REFERENCE, etc.).
• If a fix could delete or overwrite user content, confirm or explain before doing it.

Bundled files

• scripts/lint_skill.py: Automated lint for structure/frontmatter/reference hygiene.
• references/checklist.md: The step-by-step review checklist (with severities).
• references/standards.md: The condensed skill-creator standards to enforce.
• references/report-template.md: The report structure to use.
• references/remediation.md: Common fixes and rewrite patterns.

Converted and distributed by TomeVault — claim your Tome and manage your conversions.