Extract as Skill
When asked to extract something as a skill, follow these steps:
1. Review the Conversation
- •Identify the workflow, approach, or process to extract
- •Find the core decision-making steps
- •Note any variations or edge cases discussed
2. Name It Clearly
- •Use kebab-case (e.g.,
python-optimization,code-review) - •Action-oriented names are best
- •Avoid generic names like
thing,helper - •Example good names:
- •
python-async-patterns - •
etsy-market-research - •
refactor-for-readability
- •
3. Determine When to Use It
- •What problem does this solve?
- •What's the trigger (e.g., "when optimizing network code")
- •What type of projects benefit? (backend, frontend, data, etc.)
4. Extract Core Steps
- •Break workflow into 3-7 main steps
- •Include decision points ("if X, do Y")
- •Reference relevant files/docs if applicable
- •Show concrete examples from conversation
5. Create the SKILL.md File
Location: ~/.claude/skills/<skill-name>/SKILL.md
Template:
yaml
--- name: <skill-name> description: <one-line: what it does and when to use> allowed-tools: Read, Bash, Grep, Edit, Write, Glob --- # <Skill Title> ## When to Use This - Trigger 1 - Trigger 2 - Example: "when optimizing..." ## Core Workflow ### Step 1: [Name] [Description and rationale] ### Step 2: [Name] [Description and rationale] ### Step 3: [Name] [Description and rationale] ## Examples [Real example from conversation] ## Common Pitfalls - Mistake 1 and how to avoid it - Mistake 2 and how to avoid it ## References - Link to related docs - Link to other skills
6. Create Supporting Files (Optional)
If the skill is complex, create:
- •
EXAMPLES.md— detailed real-world examples - •
CHECKLIST.md— validation steps - •
REFERENCE.md— deeper technical details - •
TEMPLATES/— code/template examples
7. Test the Skill
- •Check that it's clear and actionable
- •Verify it works for the context you extracted it from
- •Consider if it would work in other projects
Example Output
code
✓ Created skill: ~/.claude/skills/python-async-optimization/SKILL.md Description: Design async Python code for I/O-heavy operations You can now use this in any project: > /python-async-optimization or > Have me optimize this code using the async patterns approach