skill-improver
Audit and improve skills - preserve intent, fix frontmatter, convert tutorial prose to actionable rules, track usage.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- skill-improver
- triggers
- Skill improvement, audit skills, refactor skills, skill refresher, drift detection, auto-heal
- changelog
- docs/ciclos/cycle28-20260815.md
- description
- Audit and improve skills - preserve intent, fix frontmatter, convert tutorial prose to actionable rules, track usage.
- token_budget
- 2600
02 资源文件 1
- 参考文档references/audit-examples.md1.0 KB
03 SKILL.md 正文
When to Use
Audit and improve skills — preserve author intent, fix frontmatter, convert tutorial prose to actionable rules, track usage.
Rules
- Preserve author intent, critical rules, activation semantics.
- Audit-only by default.
- Never delete: content →
references/. - Don't invent triggers — mark ambiguous.
- Skills only — no opencode.json, no AGENTS.md.
Decision Gates
| Signal | Action |
|---|---|
| Invalid frontmatter | Fix metadata |
| Tutorial prose | Convert to rules, bg → references/ |
Over budget (>3200 B — check-token-budget.ps1 -Json) |
Rules stay, examples → references/ |
| Branching prose | Decision table |
| Conflicting rules | Report both — escalate |
Anti-Patterns
| Anti-Pattern | Fix |
|---|---|
| Rewrite entire skill | Audit → fix gaps → repeat |
| Add unintended triggers | Mark ambiguous |
| Delete instead of archive | references/ |
| Major bump for trivial | minor=content |
| Ignore Engram | mem_search first |
| Over-compress rules | Preserve rules, compress examples |
Anti-Rationalization
| Rationalization | Red Flag | Verification |
|---|---|---|
| "reescribir en vez de audit-fix preservando intención" | Rewrite completo borra intent original | Regla 1 Preserve intent + audit-fix gaps only |
| "borrar en vez de archivar" | Contenido eliminado sin mover a references/ | Regla 3 Never delete → references/ + changelog |
| "añadir triggers no solicitados" | Triggers nuevos no pedidos en spec | Regla 4 Don't invent triggers — mark ambiguous |
Red Flags
- Doing work without checking output format → STOP, re-read skill
- Second occurrence of same rationalization → force RED zone
Audit
- Read all SKILL.md in target. 2. Audit: metadata · triggers · budget · actionability · gates. 3.
mem_search(last 50)— flag 90d+ untouched. 4. Report grouped-by-skill with severity.
Regeneration
- Audit + Engram (10 sessions) → 2. Drift check → 3. Fix gaps → 4. Compress if >5% new → 5. Version + changelog.
Output
{"audited":["skill-a"],"issues":[{"skill":"skill-a","severity":"high"}],"drifted":["skill-b"]}
Verification
- Output matches skill ## Output contract + file:line citaton
- cross-ref-check.ps1 → SKILL.md OK
Refs
karpathy-loop · opencode-skill-creator · skill-testing · immune-system · dreaming · skill-registry · skill-graph
Reference
docs/skills/skill-improver/reference.md
04 同仓库更多 Skills
web-quality-audit
Comprehensive web audit: performance, a11y, SEO, responsive, animation, design tokens.
ui-engine
UI system — Grid/Flexbox/@layer/:has(), container queries, compositor-only animation, OKLCH tokens, component patterns
baseline-ui
Anti-slop UI — layout, typography, responsive, animation, tokens. Use for cleanup or polish.
skill-creator
Create new AI agent skills following Agent Skills spec -- auto-bootstrap unknown tasks, define structure, register in AGENTS.md
senior-engineer
Apply senior engineering competencies — system design, trade-off analysis, delegation, mentoring, and tech debt strategy
prompt-engineering
Design effective prompts using SPEARS framework — scope, principles, examples, assertions, refinements, with ReAct and multi-agent patterns