| name | learner |
| description | Extract a reusable skill from the current conversation — learn from what just worked |
| tags | ["meta","skill-creation"] |
When to Use
- A workflow was just executed successfully and should be captured as a reusable pattern
- User says "learn this", "save as skill", "remember how we did this"
- A multi-step process worked well and should be repeatable
Workflow
1. Analyze Conversation
Review the conversation to identify the successful pattern:
- What was the goal?
- What steps were taken?
- Which agents were involved?
- What tools were used?
- What made it succeed vs. generic approaches?
2. Extract Pattern
Identify the reusable elements:
- Trigger conditions: When should this skill activate?
- Steps: Ordered workflow (not just a list of actions)
- Agent routing: Which agents for which steps, with model hints
- Persistence: What state to save and where
- Success criteria: How to know it worked
- Failure modes: What can go wrong and how to handle it
3. Generate SKILL.md
Create a complete SKILL.md with proper structure:
---
name: {skill-name}
description: "{one-line description}"
tags:
- {relevant-tags}
---
## When to Use
{trigger conditions}
## When NOT to Use
{anti-patterns}
## Workflow
{step-by-step protocol}
## Tool Usage
{which tools and agents, with task() examples}
## Examples
{good and bad examples from the conversation}
## Checklist
{success verification}
4. Save
Write to plugin/skills/{name}/SKILL.md via edit/create.
Report: "New skill saved. Reinstall plugin to activate: copilot plugin install ./plugin"
What Makes a Good Learned Skill
- Specific enough to be useful (not "do good things")
- General enough to apply beyond this one conversation
- Protocol-driven with concrete steps (not vague advice)
- Tested — the conversation proved it works
Checklist
Trigger Keywords
learn this, save as skill, remember how
Quality Contract
- Generates reusable SKILL.md from conversation patterns