agentsclimarketplace
Sign in

Skill proof

Skill emdysq/skill-proof/skills/skill-proof

Audit, repair, and prepare Agent Skill folders for publication by checking SKILL.md routing metadata, structure, portability, safety hazards, resource organization, and release evidence. Use when Codex needs to review a skill before GitHub publishing, generate a proof report, create trigger or anti-trigger eval prompts, or help an author make an Agent Skill more reusable and trustworthy.From its SKILL.md

Install
npx -y skills add emdysq/skill-proof --skill skill-proof

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.2 KB, 402 tokens by cl100k_base, as published. Nobody here has run it

Skill Proof

Workflow

  1. Identify the target skill folder. It must contain SKILL.md.

  2. Run the bundled proof script when Node.js is available:

    node scripts/skill_proof.mjs <target-skill-folder> --out skill-proof-output --fail-on high

  3. Read skill-proof-output/skill-proof-report.md first, then inspect skill-proof-output/skill-proof.json for machine-readable details.

  4. Fix high-severity findings before recommending publication.

  5. Use skill-proof-output/evals.json as starter trigger tests in a fresh agent context.

Manual Review

Use references/checklist.md when a finding needs judgment rather than a mechanical fix. Prioritize these questions:

  • Does the frontmatter description say when to use the skill?
  • Does the body stay procedural instead of becoming a long knowledge dump?
  • Are deterministic or risky operations moved into scripts?
  • Are references and assets discoverable from SKILL.md?
  • Are there hard-coded local paths, real secrets, or unsafe shell commands?
  • Could a stranger install the skill and understand the release evidence?

Repair Guidance

When editing a target skill, keep changes narrow and preserve the author's intent. Improve the release gate in this order:

  1. Fix invalid frontmatter, missing name, invalid name, or placeholder description.
  2. Move trigger conditions into frontmatter description.
  3. Remove secrets and replace local paths with placeholders.
  4. Guard destructive commands with explicit confirmation and path validation.
  5. Move large detailed guidance into references/.
  6. Add or update scripts for repeated checks.
  7. Regenerate the proof report.

Output Use

Treat the report as a publishing artifact, not only a lint result. Summarize the clean sections in the target repository README and use the generated eval prompts for release demos.

What ships with it: 3 files

28.4 KB alongside SKILL.md, 1 of them executable

agents/

references/

scripts/

Keep looking

Skills are one crate of 326,499. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.