About learn
learn is a skill for recording insights, corrections, preferences, or rules that an agent should retain for future tasks. It stores the learning in the memory system and writes it to `.cowork/MEMORY.md`.
Key features
- 01Record insights, corrections, and preferences
- 02Store rules for future tasks
- 03Write learning to `.cowork/MEMORY.md`
Use cases
- 01Save a recurring user preference
- 02Record a correction for later tasks
- 03Persist a rule the agent should remember
Install
Copy the command and run it in a terminal.
$
npx skills add https:/ / github.com/ cowork-os/ cowork-os --skill learnAfter installing, type/learnin Claude Code to confirm it loaded.
Details
From the project READMELearn
Purpose
Manually record an insight, correction, preference, or rule that the agent should remember for future tasks.
Routing
- Use when: Use when the user explicitly wants to teach the agent something, record a preference, save a correction, or store a rule for future tasks.
- Do not use when: Don't use for general conversation, one-time instructions that don't need to be remembered, or when the user is just chatting.
- Outputs: Persisted learning in memory system and .cowork/MEMORY.md file.
- Success criteria: Learning is stored in memory and written to .cowork/MEMORY.md with confirmation to the user.
Trigger Examples
Positive
- Use the learn skill for this request.
- Help me with learn.
- Use when the user explicitly wants to teach the agent something, record a preference, save a correction, or store a rule for future tasks.
- Learn: provide an actionable result.
Negative
- Don't use for general conversation, one-time instructions that don't need to be remembered, or when the user is just chatting.
- Do not use learn for unrelated requests.
- This request is outside learn scope.
- This is conceptual discussion only; no tool workflow is needed.
Parameters
| Name | Type | Required | Description |
|---|---|---|---|
| what | string | Yes | The insight, correction, preference, or rule to remember |
| category | select | No | Type of learning |
Runtime Prompt
- Current runtime prompt length: 1153 characters.
- Runtime prompt is defined directly in
../learn.json.
Comments
Markdown supported