Skill Stocktake

/skill-stocktake — Skill Quality Audit

Systematic audit of all skills to identify what to keep, improve, update, retire, or merge. Prevents skill bloat and maintains collection quality.

Usage

/skill-stocktake              # Full audit of all skills
/skill-stocktake quick        # Only skills changed since last stocktake
/skill-stocktake <skill>      # Audit a single skill

Audit Dimensions

Each skill is evaluated against four dimensions:

1. Actionability (Weight: 30%)

Does the skill provide clear, executable guidance?

2. Scope Fit (Weight: 25%)

Does the skill belong in Hangar's mission?

3. Uniqueness (Weight: 25%)

Is this skill distinct from other skills and rules?

4. Currency (Weight: 20%)

Is the skill up-to-date with current tools and practices?

Verdicts

Based on the weighted score, each skill receives a verdict:

Workflow

Step 1: Inventory

# Count and list all skills
ls -d core/skills/*/

Step 2: Per-Skill Analysis

For each skill, read the SKILL.md and evaluate:

1. Read the skill content
2. Score each dimension (1-10)
3. Check for overlap with other skills (grep for similar keywords)
4. Check for outdated references (tool versions, deprecated APIs)
5. Assign verdict

Step 3: Cross-Skill Analysis

After individual audits:

• Overlap detection: Find skills that cover similar ground
• Gap detection: Find workflow stages with no skill coverage
• Dependency mapping: Which skills reference or complement each other

Step 4: Report

Output Format

## Skill Stocktake Report

**Date:** 2026-04-05
**Skills Audited:** 28
**Mode:** Full Stocktake

### Summary

| Verdict | Count | Skills |
|---------|-------|--------|
| Keep | 18 | scan, verify, context-budget, ... |
| Improve | 5 | polish, handoff, ... |
| Update | 3 | lighthouse-quick, capture-pdf, ... |
| Merge | 1 | audit-runner → audit |
| Retire | 1 | (none) |

### Detailed Results

#### scan — KEEP (Score: 88%)
| Dimension | Score | Notes |
|-----------|-------|-------|
| Actionability | 9/10 | Clear 4-phase workflow |
| Scope Fit | 9/10 | Core Hangar functionality |
| Uniqueness | 9/10 | Only onboarding skill |
| Currency | 8/10 | Current patterns |

#### polish — IMPROVE (Score: 72%)
| Dimension | Score | Notes |
|-----------|-------|-------|
| Actionability | 8/10 | Good workflow |
| Scope Fit | 7/10 | Frontend-specific |
| Uniqueness | 6/10 | Overlaps with design-system |
| Currency | 7/10 | Mostly current |

**Improvements needed:**
1. Clarify distinction from design-system skill
2. Add non-frontend polish checks (API response times, DB queries)

Quick Scan Mode

When using quick mode:

1. Check git log for skills changed since last stocktake
2. Only audit changed skills
3. Cross-reference changes against existing verdicts
4. Report changes in verdict (e.g., "Improved: polish moved from IMPROVE to KEEP")

# Find changed skills since last stocktake
git log --name-only --since="2026-03-01" -- core/skills/

Reason Quality Requirements

Every verdict MUST include:

• Specific evidence — Quote the exact line or section that supports the score
• Actionable recommendation — What specifically to change (not just "improve")
• Comparison reference — How this compares to similar skills in the collection

Vague verdicts like "could be better" are not acceptable.

When to Use

• Monthly maintenance — Run full stocktake to catch drift
• After adding skills — Quick scan to ensure new skills meet standards
• Before releases — Verify collection quality
• After rules-distill — Complete the governance cycle

Inspired by ECC's skill-stocktake with strict reason quality requirements.