Skip to content
AI Dev Toolkit.
Esc
  • AI Token CounterCount tokens for GPT, Claude, Gemini, DeepSeek, Qwen and more.Tool
  • LLM API Cost CalculatorEstimate per-request, daily and monthly API costs.Tool
  • AI Model ComparisonCompare prices, context windows and features across models.Tool
  • AI Model Pricing PagesSpecs, real costs and cheaper alternatives for popular models.Tool
  • Context Window CheckerSee whether your text fits each model’s context window.Tool
  • Subscription vs API CalculatorFind out whether a chat plan or the API is cheaper for you.Tool
  • GPU / VRAM CalculatorCheck how much VRAM a local model needs and which GPUs fit.Tool
  • Prompt Caching CalculatorEstimate savings from prompt caching.Tool

Claude skill · Development · Anthropic

Skill creator skill for Claude

Anthropic’s skill for writing skills: draft a SKILL.md, test it against a baseline, review the results, and tune the description so it triggers reliably.

Command
/skill-creator
Licence
Apache-2.0, copying allowed with attribution
Folder
18 files, 220 KB
Runs code
10 scripts

Overview

What it does

It goes well beyond a template. After interviewing you and drafting the skill, it runs test prompts with and without the skill (or against its previous version), grades the outputs, and opens a viewer where you compare them and leave feedback. The loop repeats until the skill clearly helps.

A separate stage optimises the description: it generates prompts that should and shouldn’t trigger the skill and adjusts the wording until Claude picks it up at the right times. Our skill generator covers the first step, writing a valid SKILL.md.

Triggers

When Claude uses it

  • Creating a skill from scratch, improving one, running evals on it, or making its description trigger more reliably.
Its description, which Claude matches against
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

What it needs

  • Python for the evaluation and packaging scripts. The full loop works best in Claude Code: it runs tests in parallel subagents, and description tuning needs the `claude -p` command.

Install

How to install it

Claude Code

Plugin (type in Claude Code)
/plugin marketplace add anthropics/skills
/plugin install example-skills@anthropic-agent-skills

The example-skills plugin installs 12 skills from this repository, including this one. Its skills are then namespaced, such as /example-skills:skill-creator.

Or copy the folder (shell)
git clone --depth 1 https://github.com/anthropics/skills.git
mkdir -p ~/.claude/skills
cp -r skills/skills/skill-creator ~/.claude/skills/

That makes it a personal skill, available as /skill-creator in all your projects. Copy it into .claude/skills/ in a repository instead to share it with your team.

claude.ai and the Claude API

Zip the skill folder, then in claude.ai open Customize > Skills, choose + and Create skill, then Upload a skill. Code execution must be turned on.

Upload with the Python SDK
from anthropic import Anthropic
from anthropic.lib import files_from_dir

client = Anthropic()
skill = client.skills.create(files=files_from_dir("skill-creator"))
print(skill.id)  # use as a custom skill_id with the code execution tool

Checks

SKILL.md validator results

Its SKILL.md (485 lines) run through our SKILL.md validator, with the folder’s file list, for both targets. These check the file’s format, not whether the skill works well.

Claude Code

0 errors, 0 warnings, 1 tip

  • tipSKILL.md doesn’t mention agents/analyzer.md, agents/comparator.md, agents/grader.md. Say what each file is for and when to read or run it, or Claude won’t know to use it.

claude.ai and the API

0 errors, 0 warnings, 2 tips

  • tipThe Claude Help Center’s upload guide gives a 200-character description limit for claude.ai, though the platform docs say 1,024. Keep it under 200 if you’ll upload it there.
  • tipSKILL.md doesn’t mention agents/analyzer.md, agents/comparator.md, agents/grader.md. Say what each file is for and when to read or run it, or Claude won’t know to use it.

Folder

What’s in the folder

  • LICENSE.txt11 KB
  • SKILL.md32 KB
  • agents/3 files, 26 KB
  • assets/1 file, 6.9 KB
  • eval-viewer/2 files, 60 KB
  • references/1 file, 12 KB
  • scripts/9 files, 72 KB

SKILL.md sections

  1. Skill Creator
  2. Communicating with the user
  3. Creating a skill
  4. Capture Intent
  5. Interview and Research
  6. Write the SKILL.md
  7. Skill Writing Guide
  8. Writing Style
  9. Test Cases
  10. Running and evaluating test cases
  11. Step 1: Spawn all runs (with-skill AND baseline) in the same turn
  12. Step 2: While runs are in progress, draft assertions
  13. Step 3: As runs complete, capture timing data
  14. Step 4: Grade, aggregate, and launch the viewer
  15. What the user sees in the viewer
  16. Step 5: Read the feedback
  17. Improving the skill
  18. How to think about improvements
  19. The iteration loop
  20. Advanced: Blind comparison
  21. Description Optimization
  22. Step 1: Generate trigger eval queries
  23. Step 2: Review with user
  24. Step 3: Run the optimization loop
  25. How skill triggering works
  26. Step 4: Apply the result
  27. Package and Present (only if present_files tool is available)
  28. Claude.ai-specific instructions
  29. Cowork-Specific Instructions
  30. Reference files

Source

Licence and source

Related

More skills

All Claude skillsWrite your own skillSlash command builderGuide: subagents, skills and commands