Skip to content
Get packs

Skill Reviewer

Audits an existing agent skill for readiness, routing, safety and evidence without editing it.

Source Download skill
Repository
bytedance/deer-flow
Stars
84,876
Licence
MIT
Files
13 files

What it does

When installed, the agent acts as a quality reviewer for other agent skills. It inspects a skill package as untrusted data, checks whether its description triggers correctly, flags safety and resource problems, and grades how much evidence exists for its claims. It reports a readiness verdict and suggests paste-ready fixes, but it does not apply changes; edits, creation or experiments are handed off to a separate skill-creator skill.

Great for

  • Developers maintaining a DeerFlow or other SKILL.md skill library who want a pre-publish check
  • Teams deciding whether a skill is ready to share internally or publicly
  • Authors diagnosing why a skill fires too often or not at all

Skip it if

  • Ordinary application-code review, which the skill explicitly excludes
  • Someone who wants a skill created or edited, since it only recommends changes and hands off that work
  • Anyone expecting proof of runtime behaviour: a 'publish_candidate' verdict only means no material issue was found in the assessed scope

Try asking

  • “Review the skill at skill://public/data-analysis for production readiness”
  • “Audit my pasted SKILL.md and tell me whether its description will over-trigger”
  • “Check whether the evals for this skill actually prove what it claims”
  • “Grade only the safety and resources dimensions of my team-helper skill”
Step-by-step guide: how to use Skill ReviewerInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 13 original files, the MIT licence and an attribution note.

    skill-reviewer.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip skill-reviewer.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip skill-reviewer.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give a precise scope (for example only triggers or safety) to get a focused, faster review.
  • Ask for deterministic facts only when you want a purely mechanical check without semantic judgement.
  • Treat the verdict as a checklist for human review; use the separate assurance level to see how much was actually tested.

What's inside

  • evals/evals.json
  • evals/fixtures/blocked/SKILL.md
  • evals/fixtures/needs-revision/SKILL.md
  • evals/fixtures/partial-package/SKILL.md
  • evals/fixtures/prompt-injection/SKILL.md
  • evals/fixtures/publish-candidate/SKILL.md
  • evals/fixtures/zh-output/SKILL.md
  • references/effect-verification.md
  • references/eval-design.md
  • references/report-rendering.md
  • references/review-checklist.md
  • references/review-rubric.md
  • SKILL.md

Redistributed unmodified from bytedance/deer-flow under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.