Specialist: Memory Manager
Trigger: /learn comment on any issue
Browser required: No
Default: ✅ Enabled
/learn manages what the agent has learned across sessions. It provides tools to review, search, prune, and export project-specific patterns and preferences that the agent has accumulated over time.
Key capabilities:
- Review learnings — See everything the agent has learned about your project
- Search — Find specific patterns, conventions, or past decisions
- Prune — Remove stale or incorrect learnings that no longer apply
- Export — Extract learnings as documentation or configuration
- When you wonder "didn't we fix this before?" or "what did we decide about X?"
- To audit what the agent knows about your project
- To clean up outdated patterns after a major refactor
- To onboard new team members by showing accumulated project knowledge
- After major changes to prune learnings that no longer apply
Comment on any issue:
/learn
You can also say "what have we learned", "show learnings", "prune stale learnings", or "export learnings".
Review all learnings:
/learn
Show me everything you've learned about this project.
Search for specific patterns:
/learn
What do you know about our authentication patterns?
Prune stale learnings:
/learn
We just migrated from REST to GraphQL. Prune any learnings about REST endpoints.
The agent's primary memory is Git — session transcripts committed to state/sessions/. /learn provides a higher-level view of the patterns extracted from those sessions, making accumulated knowledge searchable and manageable.
| Command | Relationship |
|---|---|
/retro |
Retrospectives generate new learnings from recent work |
/office-hours |
Discovery sessions often establish foundational project patterns |
/investigate |
Root cause investigations add debugging patterns to learnings |
{
"learn": { "enabled": true, "trigger": "issue_comment" }
}
