Learn Skill

You are conducting deep research to build lasting knowledge on a topic. This is NOT a quick answer - spend real time investigating thoroughly.

Input Parsing

Parse $ARGUMENTS to determine mode:

• Research mode: Just a topic (e.g., "React Server Components", "product research best practices")
• Document mode: Topic + "from" + path/URL (e.g., "API design from ./docs/api-guide.md")

Research Process

For Research Mode (no document provided):

1. Web Search: Search for current best practices, official docs, expert opinions
2. Multiple Sources: Gather from 3-5 authoritative sources minimum
3. Deep Reading: Actually read and synthesize, don't just skim
4. Current Info: Prioritize recent content (last 1-2 years)

For Document Mode (document provided):

1. Read Thoroughly: Read the entire document carefully
2. Extract Framework: Identify the methodology, principles, process
3. Capture Nuance: Get the "why" not just the "what"
4. Preserve Voice: Maintain the original perspective/philosophy

Output Structure

Create a comprehensive knowledge artifact with this structure:

# [Topic Name]

> Learned: [Date]
> Mode: [Research | Document: source-name]
> Sources: [List of sources]

## TL;DR
[2-3 sentence executive summary]

## Core Concepts
[Key ideas, definitions, mental models]

## Best Practices
[What TO do - with reasoning]

## Anti-Patterns & Pitfalls
[What NOT to do - common mistakes]

## Process/Framework
[Step-by-step methodology if applicable]

## Code Examples
[If technical - real, working examples]

## Key Insights
[Non-obvious learnings, expert tips]

## When to Apply
[Context for when this knowledge is relevant]

## Sources & References
[Links, citations for deeper reading]

Saving the Knowledge

1. Convert topic to kebab-case filename (e.g., "React Server Components" → "react-server-components.md")
2. Save to: ~/.claude/knowledge/[filename].md
3. If project-specific, also mention they can move to .claude/knowledge/ for project scope

Data warning: Knowledge files persist indefinitely on disk with no encryption or automatic cleanup. Avoid storing customer names, financial figures, deal values, or confidential strategic information in knowledge files. Keep learned knowledge general — frameworks, patterns, and public information are fine; customer-specific data is not.

After Saving

1. Confirm what was learned and where it was saved
2. Give a brief summary of the key takeaways
3. Mention they can use /recall [topic] to retrieve this knowledge later

Quality Standards

• Depth over breadth: Better to know one thing well than many things superficially
• Actionable: Knowledge should be usable, not just theoretical
• Opinionated: Capture best practices, not just options
• Examples: Always include concrete examples
• Sources: Always cite where information came from