Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add UniM0cha/self-improving-skills --skill distill-skillgit clone --depth 1 https://github.com/UniM0cha/self-improving-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/unim0cha/self-improving-skills/distill-skill)<a href="https://agentmods.dev/skills/unim0cha/self-improving-skills/distill-skill"><img src="https://agentmods.dev/badge/skills/unim0cha/self-improving-skills/distill-skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/unim0cha/self-improving-skills/distill-skill"><img src="https://agentmods.dev/badge/skills/unim0cha/self-improving-skills/distill-skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00080 | $0.00707 |
| Opus 5.5 | $0.00032 | $0.00283 |
| Sonnet 5.5 | $0.00016 | $0.00141 |
| Haiku 4.5 | $0.00008 | $0.00071 |
Grade A, and why
distill-skill scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
방금 진행한 작업을 돌아보고, 재사용 가능한 기법·패턴·해결책이 있으면 학습 스킬로 증류하세요.
이것은 자기개선 루프의 수동 트리거입니다 (Stop 훅 자동 nudge와 동일한 일을 사용자가 원할 때 실행).
1단계 — skill-distiller 서브에이전트에 위임
Task(또는 Agent) 도구로 subagent_type="claude-cowork-self-improving-skills:skill-distiller" 를 호출하세요 (플러그인 네임스페이스 접두사를 포함한 정확한 이름입니다 — skill-distiller 만으로는 호출이 실패합니다). 프롬프트에 다음을 담으세요:
- 방금 세션에서 무엇을 했고 무엇이 비자명했는지 (어떤 문제를 어떻게 풀었는지, 어떤 API/라이브러리 동작을 발견했는지)
- 이 세션의 transcript 경로 (
~/.claude/projects/<프로젝트>/<세션ID>.jsonl— 서브에이전트가 증류 근거를 직접 읽을 수 있게) - 사용자가 인자를 함께 전달했다면 그 안내 (예: 특정 기법에 집중, 특정 기존 스킬 갱신)
서브에이전트가: 기존 스킬이 있으면 patch, 없고 class-level 로 재사용 가능할 때만 새 스킬 생성, 일회성·환경의존적 지식이면 아무것도 만들지 않고 이유를 한 줄 보고합니다.
환경변수 SIS_DISTILLER_MODEL 이 설정되어 있으면(예: sonnet) 호출에 model="<값>" 파라미터를 포함하세요. 값이 haiku 면 무시하세요(서브에이전트 Haiku 금지 정책).
2단계 — Cowork 필수: 저장까지 안내
서브에이전트 보고에 생성/patch 된 SKILL.md 경로가 있으면:
- 각 SKILL.md 를 SendUserFile 로 사용자에게 보내세요 (caption 에 스킬 이름과 기법 한 줄).
- 사용자에게 파일 카드의 '스킬 저장' 버튼을 눌러 claude.ai 에 등록하도록 안내하세요 — 이 컨테이너의
~/.claude/skills는 세션이 끝나면 사라지고, 등록된 스킬만 다음 세션 컨테이너에 자동 동기화됩니다.
서브에이전트가 "만들 것 없음"으로 판단했다면 그 이유를 사용자에게 한 줄로 전달하고 끝내세요 — 그것도 정상 결과입니다.
참고: 증류는 "재사용 가능한 기법"이 있을 때만 의미가 있습니다. 단순 기능 구현·리팩토링·일회성 수정만 했다면 만들지 않는 것이 맞습니다.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today First seen · 32 lines · 80 tokens per session scan A c51350c14974
distill-skill is a skill published in the GitHub repository UniM0cha/self-improving-skills (51 stars, last pushed 1mo ago), licensed MIT. It adds 80 tokens to every session and 707 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-07.
distill-skill compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| distill-skillUniM0cha/self-improving-skills | Skill | 51 | 1mo ago | MIT | A | 80 | Claude Code, Codex |
| browser-harness-authoringAtlasOmnia/hermes-custom-pack | Skill | 76 | 11d ago | MIT | A | 60 | Claude Code, Codex |
| notion-artifact-captureAtlasOmnia/hermes-custom-pack | Skill | 76 | 11d ago | MIT | A | 82 | Claude Code, Codex |
| obsidian-memory-architectureAtlasOmnia/hermes-custom-pack | Skill | 76 | 11d ago | MIT | A | 56 | Claude Code, Codex |
| hermes-diagnostic-reviewAtlasOmnia/hermes-custom-pack | Skill | 76 | 11d ago | MIT | A | 49 | Claude Code, Codex |
| reddit-browse-and-postAtlasOmnia/hermes-custom-pack | Skill | 76 | 11d ago | MIT | A | 54 | Claude Code, Codex |
| medical-bill-reconciliationAtlasOmnia/hermes-custom-pack | Skill | 76 | 11d ago | MIT | A | 19 | Claude Code, Codex |