Agents (all 14):
- Add confidence scoring (0-100) with clear action thresholds
- Add domain-specific anti-pattern lists
- Standardize structure across all specialists

Commands:
- deslop: Before/after examples, words to watch, false positive list
- design-audit: Confidence scoring, false positives, --thorough mode
- All long commands: Quick reference headers

Skills:
- verify-changes: Quick reference header

Full changelog in commit history.