learn-skill

name	learn-skill
description	Review the current session and propose new skills that could be created. Use when the user says "propose skills", "what skills could help", "find skill opportunities", "learn-skill", or after a long session to identify reusable patterns worth codifying into skills.
context	fork

Review this session and identify patterns that would benefit from being codified as reusable skills.

Look for:

• Repeated workflows: Multi-step procedures performed more than once or likely to recur
• Domain expertise applied: Specialized knowledge about tools, APIs, or systems that had to be recalled or looked up
• Workarounds: Non-obvious solutions to tool limitations or environment quirks
• Complex procedures: Multi-step processes where ordering, flags, or edge cases matter
• Tool orchestration: Patterns combining multiple tools in a specific way

For each candidate skill, determine:

1. Name: Hyphen-case identifier
2. Description: What the skill would do and when it triggers (this becomes frontmatter)
3. What it would contain: Instructions, scripts, references, or assets
4. Placement: Project-level (.claude/skills/) or user-level (~/.claude/skills/)
   • Project-level: Tied to this specific codebase, tech stack, or repo conventions
   • User-level: Generally useful across projects
   • If unclear, ask the user

Skip patterns that:

• Are already covered by an existing skill
• Are standard language/framework usage Claude already knows
• Would be used only once
• Are better captured as an AGENTS.md learning (use /learn for those)

Present each proposal in this format:

## <skill-name>

**Description:** <frontmatter description — what it does and when to trigger>
**Contains:** <brief list: instructions only, or scripts/references/assets needed>
**Placement:** <user-level | project-level> — <one-line rationale>
**Rationale:** <why this is worth a skill — what pain it prevents>

After presenting proposals, ask the user which (if any) they want to create, then hand off to /skill-creator with the relevant context.

$ARGUMENTS