Skill Creator
Create new skills for Claude Code. A skill is just a markdown file — nothing more.
Trigger
Use when the user says "create a skill", "new skill", "skill for X", /skill-creator, or asks to build/improve a custom command.
Process
Phase 1 — Interview (extract intent)
Ask these questions one by one. Don't dump them all at once.
- What should this skill do? (one sentence)
- When should it trigger? (slash command, auto-detect, or both — list trigger phrases)
- What does "good output" look like? (ask for a concrete example or description)
- What should it NOT do? (boundaries, failure modes to avoid)
- Does it need tools? (file read/write, bash, web, agents, etc.)
If the user is vague, propose a concrete interpretation and ask "is this what you mean?"
Phase 2 — Generate first draft
Write a complete SKILL.md based on the interview. Structure it as:
# [Skill Name]
[One-line description]
## Trigger
[When this skill activates — command name + auto-trigger phrases]
## Process
[Numbered steps — what Claude does when this skill runs]
## Constraints
[What NOT to do, boundaries, edge cases]
## Output
[What the user gets at the end — format, location, style]
Place it in ~/.claude/skills/[skill-name]/SKILL.md.
Phase 3 — Smoke test
Immediately after generating:
- Propose 2-3 realistic test prompts that would trigger the new skill
- Ask the user to pick one (or suggest their own)
- Simulate how the skill would respond to that prompt
- Show the result and ask: "Is this what you wanted? What's wrong?"
Phase 4 — Iterate
Based on user feedback:
- Identify what to change
- Edit the SKILL.md directly
- Re-run the same test prompt (or a new one)
- Ask again: "Better? What's still off?"
Repeat until the user says it's good. Don't stop after one round.
Phase 5 — Finalize
Once satisfied:
- Confirm the skill location and trigger command
- Suggest 1-2 edge cases to watch for
- Remind the user they can invoke
/skill-creatoragain to improve it later
Constraints
- Never over-engineer. A skill is a markdown file, not a codebase.
- Keep generated skills short — if it doesn't fit on 2 screens, it's too complex. Split it.
- Don't add folders, scripts, or infrastructure unless the user explicitly asks.
- Don't generate evaluation frameworks or rubrics. The user is the evaluator.
- Respect the user's existing skill naming conventions if they have any.
Output
A working skill in ~/.claude/skills/[skill-name]/SKILL.md, tested and iterated to the user's satisfaction.