Skip to content

Skill Creator

Helps an agent create, test, benchmark and refine other agent skills

Stars
84,876
Licence
MIT
Files
15 files

What it does

This skill teaches an AI agent how to build other skills, which are reusable instruction folders that tell an agent how to handle a task. Once installed, the agent walks you through drafting a skill, writing test prompts, running them, grading the results, and improving the skill based on feedback. It also includes local scripts that generate a browser-based review page, summarize benchmark statistics across runs, and help tune a skill's description so it triggers at the right times.

Great for

  • Developers who want to package a repeated workflow into a reusable skill
  • Teams that want measurable evidence on whether a skill actually improves agent output
  • People iterating on an existing skill's wording or trigger description

Skip it if

  • Casual users who only want to use existing skills; this is for authoring them (the scripts run local evaluations that can take time and cost model usage)
  • Anyone running it on untrusted skill folders, since the bundled scripts execute code on your machine

Try asking

  • “Create a new skill that formats my weekly status reports in our team's style”
  • “Run evals on my invoice-extraction skill and show me the results in a review page”
  • “Benchmark this skill across several runs and report the variance”
  • “Optimize the description of my code-review skill so it triggers more accurately”
Step-by-step guide: how to use Skill CreatorInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 15 original files, the MIT licence and an attribution note.

    skill-creator.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

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

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip skill-creator.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

  • Review the scripts before running them on a skill you did not write, since they execute locally
  • Start with a few test prompts before scaling up to a larger eval set
  • Ask for a plain-language explanation if the eval terms like assertion or benchmark are unfamiliar

What's inside

  • agents/analyzer.md
  • agents/comparator.md
  • agents/grader.md
  • assets/eval_review.html
  • eval-viewer/generate_review.py
  • eval-viewer/viewer.html
  • LICENSE.txt
  • references/output-patterns.md
  • references/schemas.md
  • references/workflows.md
  • scripts/aggregate_benchmark.py
  • scripts/generate_report.py
  • scripts/improve_description.py
  • scripts/init_skill.py
  • 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.