SKILL.md

name	description	metadata
skill-discovery-engine	Automatically identifies recurring patterns in the codebase and suggests or creates new skill files to institutionalize knowledge	author version oversold 1.0.0

Scan the codebase for recurring patterns and propose new skills to capture institutional knowledge.

---

Core scenarios

1. Session initialization - scan for patterns on startup
2. Complex task completed - analyze multi-file edits for reusable patterns
3. User asks "What have you learned about this project?"

Execution Steps

1. Environmental Scan

# Index existing skills
ls ./.claude/skills/

# Analyze recent activity
git log --oneline -10 --name-only

2. Pattern Detection

Search for these pattern categories:

Category	What to Look For
Refactoring	Repetitive boilerplate, complex logic transformations, repeated code structures
DevOps	Multi-step CLI sequences, configuration file patterns, deployment scripts
Testing	Mock setups, assertion patterns, test file structures
Security	Auth patterns, input validation, secret handling, CORS configs, rate limiting
Research	Data gathering workflows, API exploration, documentation synthesis
SEO	Meta tag patterns, structured data, sitemap generation, canonical URLs
Design	Component patterns, design tokens, styling conventions, responsive breakpoints
UX	Loading states, error boundaries, form validation, accessibility patterns

3. Validation Criteria

A pattern is "skill-worthy" if it meets ALL of these:

• Reusable across multiple contexts
• Non-obvious (requires domain knowledge)
• Occurs 2+ times in the codebase
• Would save time if documented

4. Skill Creation

When a valid pattern is found:

1. Propose the skill name (kebab-case)
2. Create directory structure:

.claude/skills/
└── [skill-name]/
    └── SKILL.md

mkdir -p ./.claude/skills/[skill-name]

3. Draft SKILL.md content using this template:

---
description: Brief one-line description
triggers:
  - trigger condition 1
  - trigger condition 2
---

# Skill Name

Brief overview of what this skill does.

## Steps

1. First step
2. Second step

## Code Examples

\`\`\`typescript
// Example code
\`\`\`

4. Prompt user: "I've detected a recurring pattern for [Pattern Name]. Save as .claude/skills/[skill-name]/SKILL.md?"

IMPORTANT: Always use the directory structure [skill-name]/SKILL.md, never a flat file.

5. Self-Update

If an existing skill could be improved:

• Suggest specific edits to the existing file
• Do NOT create duplicates

Output Format

## Skill Discovery Report

**Scanned**: [X files, Y recent commits]
**Existing Skills**: [list]

### Patterns Detected

1. **[Pattern Name]** (Category: Refactoring|DevOps|Testing|Security|Research|SEO|Design|UX)
   - Occurrences: N
   - Files: file1.ts, file2.ts
   - Recommendation: Create skill | Update existing | Skip (not reusable)

### Proposed Actions

- [ ] Create `.claude/skills/[skill-name]/SKILL.md`
- [ ] Update `.claude/skills/[existing-skill]/SKILL.md` with [improvement]