asset-validation

Validate Claude Code AI assets for file size limits, frontmatter correctness, cross-reference integrity, runtime readiness, and naming conventions. Use after creating or modifying rules, skills, hooks

SKILL.md

GitHub

Asset Validation

Use this skill after modifying .agents/skills/ or .codex/ assets and before accepting a parity-impacting change.

Validation Layers

1. Structure

Confirm the expected runtime layout exists:

  • .agents/skills/<name>/SKILL.md
  • .codex/roles/*.md
  • .codex/rules/*.md
  • .codex/operations/*.md
  • .codex/templates/*.md and .codex/checklists/*.md when applicable

2. Frontmatter and Format

Check:

  • SKILL.md contains at least name and description
  • optional codex-roles values resolve to both role and overlay files
  • AGENTS.md stays plain Markdown
  • markdown assets stay within repository size guidance where required

3. Cross-References

Verify:

  • skill references resolve to existing folders under .agents/skills/
  • role references resolve to existing files under .codex/roles/
  • overlay references resolve under .codex/rules/role-overlays/
  • operation docs point to existing Codex files and not stale Claude runtime paths

4. Runtime Readiness

Confirm:

  • assets assume the Codex runtime layout under .agents/skills/ and .codex/
  • Claude hook and settings behavior is translated into visible operation docs, not implied hidden automation
  • no installer-specific or machine-specific commands remain

5. Naming

  • role files use kebab-case.md
  • overlay files use the same role id as the matching role file
  • skill folders use kebab-case/
  • template files use consistent Codex naming

Report Format

Return:

  • Summary
  • Passed checks
  • Findings
  • Required follow-up

Use validation-checklist.md for the detailed checklist.