skilldeps
Use this skill when an agent needs to audit one or more agent SKILL.md folders for missing references, unclear tool requirements, weak side-effect boundaries, or absent validation instructions.
Inputs
- A path to a
SKILL.mdfile, a skill directory, or a directory containing multiple skills. - Optional severity gate:
info,warning, orerror. - Optional output format:
jsonormarkdown.
Required Tools
- Local filesystem read access.
- Node.js 18 or newer when using the CLI.
Side-Effect Boundaries
This skill is read-only. Do not execute scripts mentioned by the inspected skill. Do not install dependencies, approve skills, write into live skill directories, or call external services as part of this audit.
Approval Requirements
No approval is needed for local read-only analysis. Ask for explicit approval before combining this report with any action that mutates a skill, installs it, applies a proposal, or writes to a remote service.
Workflow
- Run
skilldeps <path> --format markdown. - Review missing references and contract warnings.
- If using CI, rerun with
--fail-on warningor--fail-on error. - Fix the skill separately and rerun the audit.
Examples
skilldeps ~/.openclaw/agents/main/agent/codex-home/skills/.system/imagegen
skilldeps ./skills --format json --fail-on warning
Validation
Run:
npm test
npm run smoke