amnafarzy/ravie/skill-creator
Claude Code and agent skill by amnafarzy/ravie
- community
Checking that the SKILL.md file is still in the repository.
Use this whenever creating or improving a Ravie skill — the user says "make this a skill", "we keep repeating this", or a workflow or correction has recurred enough to codify. Also owns evaluating existing skills after real usage. NOT for one-off tasks.
Install
npx vskill@latest install amnafarzy/ravie/skill-creatorSKILL.md
The same SKILL.md was published earlier as amnafarzy/ravie/skill-creator.
skill-creator
Create new skills from observed patterns. When you find yourself re-explaining the same workflow 3+ times, it should become a skill.
Purpose
Turn a repeated pattern, workflow, or correction into a valid SKILL.md file with YAML frontmatter so future Claude Code sessions can match it from the description. Also used to improve existing skills based on real friction.
Process
Frontmatter requirements
Every skill must begin with YAML frontmatter. The description is the ENTIRE routing surface: what the skill does plus the exact contexts and phrases that should fire it, max 60 words, phrased assertively (Claude under-triggers skills), with explicit NOT-boundaries naming the neighboring skill. When-to-use info never goes in the body.
Example:
1. Identify the trigger
What situation should activate this skill? The answer belongs first in YAML frontmatter description because Claude Code uses that description for skill selection. Be specific:
- Bad: "when working on frontend code"
- Good: "when creating a new React component that needs to connect to Supabase data"
2. Extract the process
From your past sessions and repeated corrections, document what the correct process actually is. Include:
- Steps in order
- Commands to run
- Files to check
- Common mistakes to avoid
- What the output should look like
3. Write the SKILL.md
Follow this structure:
4. Test the skill
Ask the agent to read the skill and explain when it would use it and what it would do. Check:
- Does it understand the trigger conditions?
- Would it follow the steps correctly?
- Would it produce the right output format?
- Would it respect the hard rules under pressure?
If it fails any of these, tighten the language. Use mandatory framing ("You MUST", "NEVER", "ALWAYS") for critical rules.
5. Register in ROUTER.md
Add the new skill to the index with its group, depth tier, and purpose.
Output format
When creating or revising a skill, produce a complete skill package plan and the updated SKILL.md content:
Common failure modes
Duplicate skill creation — Creating a new skill without first checking ROUTER.md leads to overlapping workflows and unclear routing. Improve the existing skill instead when the trigger already exists.
Vague trigger description — A description like "Use for frontend work" does not give Claude Code enough routing signal. The description must state the exact situation, inputs, boundaries, and when not to use the skill.
Generic checklist content — A skill body that only says "review, implement, test" is not a reusable workflow. Add concrete commands, files to inspect, output templates, approval gates, and real failure modes.
No output contract — Without an output template, future sessions produce inconsistent handoffs. Every skill needs a visible shape for the final answer or artifact it should produce.
Hard rules
- Never create a skill for a one-time task — skills are for repeated patterns
- Never create a skill that duplicates an existing one — check ROUTER.md first
- Never put client-specific content in a reusable skill
- Always include valid YAML frontmatter before the title with
namematching the directory and a trigger-contractdescription(max 60 words — all when-to-use info lives here, never in the body) - Always include "Hard rules", "Connects to", and realistic failure modes; keep the body under 500 lines (use a references/ subfolder beyond that)
- Never let two skill descriptions overlap in trigger conditions — merge the skills or add an explicit "NOT for X — use [other]" boundary
- Always test with a fresh session before considering it done
- Use mandatory language ("MUST", "NEVER") for critical behavioral rules — advisory language gets ignored under pressure
Connects to
- This skill also owns pattern-mining and skill-health evaluation (deeper checklists archived in
archive/skills/pattern-learner/andarchive/skills/workflow-evaluator/) project-control-plane— for updating CLAUDE.md to reference new skills
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/skill-creator/SKILL.md
Popularity
- Repository stars
- 3
- 7-day trend score
- 7
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/amnafarzy/ravie/skill-creator)