learn
Analyze conversation for learnings and save to docs folder
When & Why to Use This Skill
This Claude skill automates the extraction of project-specific insights and technical knowledge from active conversations. It identifies reusable patterns, architectural decisions, and troubleshooting solutions, then formats and saves them directly into your project's documentation folder to prevent knowledge loss and improve team onboarding.
Use Cases
- Capturing Architectural Decisions: Automatically document the 'why' behind system design choices made during a brainstorming session to maintain a clear ADR (Architecture Decision Record).
- Troubleshooting and Bug Fixes: Record non-obvious solutions to complex bugs and 'gotchas' encountered during debugging, creating a searchable reference for future incidents.
- Standardizing Project Conventions: Formalize new coding patterns or workflow agreements discussed during pair programming into the project's official documentation.
- Knowledge Retention for Teams: Ensure that valuable insights discovered during AI-assisted development are preserved in the codebase rather than being lost in the chat history.
| name | learn |
|---|---|
| description | Analyze conversation for learnings and save to docs folder |
| argument-hint | [optional topic hint] |
| model | claude-opus-4-5-20251101 |
| allowed-tools | Read, Write, Edit, Glob, Grep, AskUserQuestion |
Learn from Conversation
Analyze this conversation for insights worth preserving in the project's documentation.
If a topic hint was provided via $ARGUMENTS, focus on capturing that specific learning.
If no hint provided, analyze the full conversation for valuable insights.
Phase 1: Deep Analysis
Think deeply about what was learned in this conversation:
- What new patterns or approaches were discovered?
- What gotchas or pitfalls were encountered?
- What architecture decisions were made and why?
- What conventions were established?
- What troubleshooting solutions were found?
Only capture insights that are:
- Reusable - Will help in future similar situations
- Non-obvious - Not already common knowledge
- Project-specific - Relevant to this codebase
If nothing valuable was learned, say so and exit gracefully.
Phase 2: Categorize & Locate
Read existing docs to find the best home. Look for a docs/ folder or similar documentation directory in the project.
If no existing doc fits, propose a new doc file with kebab-case naming.
Note: CLAUDE.md stays stable as the entry point. All detailed learnings go to /docs only.
Phase 3: Draft the Learning
Format the insight to match existing doc style:
- Clear heading describing the topic
- Concise explanation of the insight
- Code examples if applicable
- Context on when this applies
Phase 4: User Approval (BLOCKING)
Present your proposed changes:
- What insight you identified
- Where you'll save it (existing doc + section, or new file)
- The exact content to add
Wait for explicit user approval before saving.
Phase 5: Save
After approval, save the learning and confirm what was captured.