learner

name	learner
description	Extract learned skills from conversations. Use for "learn this", "extract skill", and "capture knowledge".
trigger	/oma:learner

Skill: learner

Extract reusable skills and patterns from work sessions.

When to Use

• After successful problem-solving
• When discovering useful patterns
• Before ending sessions with learnings
• To capture tribal knowledge
• When doing something novel well

Skillify: 5-Step Extraction Workflow

When a repeatable task pattern emerges during a session, follow these steps to capture it as a reusable OMA skill.

Step 1: Identify Repeatable Task

• Is this something done more than once?
• Does the context (codebase, config, environment) repeat?
• Would another agent benefit from this knowledge?

Step 2: Extract Inputs, Steps, and Criteria

• List every input the task requires (file paths, IDs, flags, etc.)
• Document the ordered steps to complete it
• Note success criteria and failure modes

Step 3: Decide Placement

Choose the right scope based on reusability:

Scope	Path	When to Use
Built-in	plugins/oma/skills/	Universal, high-value, plugin-wide
User	~/.augment/plugins/marketplaces/oh-my-auggie/plugins/oma/skills-learned/	Personal, machine-specific
Project	.oma/skills/	Team-shared, codebase-specific

Step 4: Draft SKILL.md with YAML Frontmatter

---
name: {skill-name}
description: {one-line description}
trigger: /oma:{skill-name}
allowed-tools: [Read, Glob, Bash, Edit, Write]
model: sonnet4.6
---

## Skill: {skill-name}

{body content}

Step 5: Flag Fuzzy Items for Human Review

Mark anything uncertain with a note:

<!-- TODO: Verify {item} with human before publishing -->

Items to flag: hard-coded values, version-dependent steps, untested edge cases.

---

Learning Capture Process

1. Identify Learnings

• What worked well?
• What did we discover?
• What would we do differently?
• What patterns emerged?

2. Extract Pattern

• Generalize from specific
• Create reusable form
• Document context
• Add examples

3. Create Skill Entry

• Name the skill
• Define trigger
• Write content
• Add constraints

4. Integrate

• Place in skill catalog
• Update AGENTS.md
• Test with examples
• Share if valuable

What to Learn

Techniques

• Useful algorithms
• Efficient approaches
• Clever solutions
• Tool combinations

Patterns

• Common structures
• Repeated solutions
• Best practices
• Anti-patterns to avoid

Context

• When to use X vs Y
• Trade-offs made
• Constraints that matter
• Environmental factors

Mistakes

• What went wrong
• Why it went wrong
• How to avoid
• Recovery approach

Skill Template

---
name: {skill-name}
description: {one-line description}
trigger: /oma:{skill-name}
---

## Skill: {skill-name}

{descriptive content}

## When to Use
{context}

## Examples

{examples}

## Constraints
{limitations}

Output Format

## Learner: Extracted Skill

### Source Session
**Topic:** {what was learned}
**Date:** {when}
**Outcome:** {result}

### What We Learned
{key insight 1}
{key insight 2}
{key insight 3}

### Pattern Extracted
**Name:** {skill-name}
**Trigger:** /oma:{name}
**When to use:** {context}

### Skill Content
```markdown
{generated skill content}

Integration

• Skill created
• Tested with examples
• Added to catalog
• Documented in project

Next Steps

{how to apply}

## Quality Gate: 3 Questions

Before finalizing any extracted skill, answer all three questions. A skill should pass at least 2 of 3 to be worth capturing.

### 1. Non-Googleable?
Would a web search not readily find this solution?

- If **yes**: high value to capture (hard-won knowledge)
- If **no**: the web already has it; consider linking instead of capturing

### 2. Codebase-Specific?
Does it depend on project context, local paths, custom tooling, or undocumented behavior?

- If **yes**: strong candidate for project/project scope
- If **no**: candidate for built-in or user scope

### 3. Hard-Won?
Did it require debugging, iteration, or repeated failure to discover?

- If **yes**: high value to capture so others don't repeat the effort
- If **no**: the knowledge is likely obvious or already well-documented

---

## Skillify Output Format

When skillify is complete, report:

Skillify Complete

Skill Name: {name} Target Location: {path} Trigger: /oma:{name}

Draft Workflow

{steps extracted}

Open Questions

• {fuzzy item 1}
• {fuzzy item 2}

---

## Constraints

- Extract from real work
- Generalize appropriately
- Don't overfit to specifics
- Test extracted patterns
- Keep skills focused