| name | description | disable-model-invocation |
|---|---|---|
learn |
Analyze the current session and extract key learnings worth preserving. |
true |
Analyze the current session and extract key takeaways worth preserving.
Separate findings into three categories:
Personal (~/CLAUDE.md):
- General patterns and approaches that work across projects
- Tool/language insights not specific to this codebase
- Workflow preferences discovered
- Errors caused by misunderstanding that were corrected during the session
Project (CLAUDE.md):
- Architecture decisions specific to this codebase
- Project-specific conventions and patterns
- Codebase quirks and gotchas
- Integration details with project's stack
- Errors and mistakes fixed during test implementation and runs
- Insights from code review if available
Memory (auto memory system):
- User preferences and feedback specific to this project (type: user, feedback)
- Project status, ongoing initiatives, deadlines (type: project)
- References to external systems: boards, channels, dashboards (type: reference)
- Gotchas and lessons learned that are personal experience, not codebase conventions
- Anything that is project-scoped but should NOT be committed to the repo
If the project has no CLAUDE.md (or user has no write access to repo):
- Everything from Project category goes to Memory instead
- Use memory types:
projectfor architecture/conventions,feedbackfor gotchas/corrections - Do NOT create CLAUDE.md without explicit user request
- Read target files before proposing changes (~/CLAUDE.md, CLAUDE.md, MEMORY.md)
- Check for duplicates or overlapping content across ALL targets — skip if already covered
- Skip findings if found in another file (e.g. project item already in ~/CLAUDE.md globally)
- Decide where each item fits best within existing structure (don't create new sections unless nothing fits)
- Formulate concisely, matching the style of existing content
Show proposed additions grouped by target:
~/CLAUDE.md:
- [where in file] addition text
CLAUDE.md:
- [where in file] addition text
Memory:
- [type: feedback] description of memory
- [type: project] description of memory
Ask for confirmation. Accept: "y", "yes", or selective like "only project" / "skip personal" / "only memory".
After confirmation, apply changes.