test-repo/memory/SKILLS_AUDIT.md

86 lines
2.6 KiB
Markdown

# Skills Audit - Active vs Unused
**Date:** 2026-02-22
**Total Skills:** 58
**Purpose:** Identify token bloat from unused skills
---
## 🔥 Active Skills (Frequently Used)
| Skill | Category | Usage |
|-------|----------|-------|
| nextjs-expert | Web Dev | High - All web projects |
| shadcn-ui | UI | High - Component library |
| frontend-design | Design | Medium - Layout/styling |
| web-design-guidelines | Design | Medium - Standards |
| subagent-driven-development | Architecture | Medium - Complex tasks |
| agent-tools | Core | High - Always loaded |
| find-skills | Core | Medium - Skill discovery |
| using-superpowers | Core | Medium - Capabilities |
| web-research-fast | Research | Medium - Quick lookups |
| url-research-task-creator | Research | Low - Background tasks |
---
## ⚠️ Context-Specific Skills (Load on Demand)
**iOS Development (11 skills)**
- ios-17-role, ios-18-role, ios-26-role
- ios-debugger-agent, ios-project-structure
- swiftui-expert-skill, swiftui-modern, swiftui-mvvm
- swift-modern, swiftdata-cloudkit
- xcode-project-shell-ios
*These load automatically when iOS work detected. Keep but don't preload.*
**Firebase (6 skills)**
- firebase-basics, firebase-auth-basics
- firebase-firestore-basics, firebase-hosting-basics
- firebase-app-hosting-basics, firebase-ai-logic
*Only used when Firebase projects active. Consider on-demand loading.*
---
## 💡 Optimization Recommendations
### Immediate (Implemented 2026-02-22)
1. ✅ Created BRAIN.md for external memory
2. ✅ Added write discipline to AGENTS.md
3. ✅ Optimized HEARTBEAT.md for <3s checks
4. Created heartbeat-state.json for tracking
### Short-Term (Next Week)
1. **Audit iOS skills** - Many may be redundant (swiftui-modern vs swiftui-expert-skill)
2. **Consolidate Firebase** - 6 separate skills could be 2-3
3. **Document skill usage** - Add usage notes to each SKILL.md
### Long-Term (Next Month)
1. **Lazy loading** - Don't load skills until first use
2. **Project-specific profiles** - Web mode vs iOS mode
3. **Monthly re-audit** - Check /context detail quarterly
---
## Token Impact Estimate
**Assumption:** Each skill = ~50-200 tokens in system prompt
| Scenario | Skills Loaded | Est. Tokens |
|----------|---------------|-------------|
| Current | 58 | ~6,000-8,000 |
| Optimized (Web focus) | 15 | ~1,500-2,000 |
| Optimized (iOS focus) | 20 | ~2,000-2,500 |
**Potential savings:** 60-70% reduction in token bloat
---
## Action Items
- [ ] Consolidate swiftui-* skills into fewer files
- [ ] Merge firebase-* into core + advanced
- [ ] Add "Last Used" tracking to skills
- [ ] Create skill usage report monthly