Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

omd:release-hygiene · kwakseongjae | DecimalAI

kwakseongjae/omd:release-hygiene

SkillScore86Maintained

오픈소스 릴리스 위생 루틴. 새 기능/스킬/에이전트를 추가했을 때 (1) 부산물(v1/v2 .bak, _source, 로그, scratch dir)이 changeset에 없는지, (2) 문서(README/docs/home)가 동기화됐는지, (3) npm 배포 대상인지(skills/ + files vs .claude local-preview) 판단했는지 체크리스트로 강제. '릴리스 준비', '커밋 전 점검', 'push 전에 정리', '배포 위생', 'release hygiene' 류 트리거. husky pre-commit이 부산물 게이트를 자동 실행.

Code Reviewby kwakseongjae

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

DevOps & SRE pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/kwakseongjae-omd-release-hygiene/SKILL.md

decimalai skills pull kwakseongjae-omd-release-hygiene --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/kwakseongjae-omd-release-hygiene/SKILL.md https://app.decimal.ai/s/kwakseongjae-omd-release-hygiene/SKILL.md

Raw file: /s/kwakseongjae-omd-release-hygiene/SKILL.md ↗ · machine-readable summary at /s/kwakseongjae-omd-release-hygiene.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 8d ago

—+49% tokens0% turns86% of whole cases pass with skill (19/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-08	✗→✓	▲ Improved	37%	0%
case-09	✗→✓	▲ Improved	28%	0%
case-11	✗→✓	▲ Improved	-19%	0%
case-12	✗→✓	▲ Improved	69%	0%
case-15	✗→✓	▲ Improved	47%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

kwakseongjae / oh-my-design

.claude/skills/omd-release-hygiene/SKILL.md

View on GitHub ↗

License MITImported 8/3/2026Version v1

Skill contents

SKILL.md

omd:release-hygiene

오픈소스 레포가 "최적의 결과물만 push"하도록 강제하는 루틴. 실험 과정에서 생긴 부산물이 public repo로 새는 것을 막고, 새 deliverable이 문서/패키징과 어긋나지 않게 한다.

자동 게이트는 scripts/check-release-hygiene.sh가 husky pre-commit에서 돌린다. 이 스킬은 그 게이트가 잡지 못하는 판단(문서 동기화, npm 배포 여부)을 사람/에이전트가 책임지게 한다.

0. 언제 트리거하나

새 스킬 / 에이전트 / CLI 플래그 / 큰 기능을 추가한 직후
커밋 또는 push 직전 ("정리하고 올리자")
실험(experiments/, .promo/)을 돌린 세션의 마무리

1. 4-항목 체크리스트 (전부 통과해야 push)

1] 부산물 0건 (자동 + 수동)

자동: bash scripts/check-release-hygiene.sh staged — pre-commit이 이미 실행. 수동 확인 대상 (게이트 정규식 밖일 수 있는 것):

반복 드래프트: *.v1.html, *-draft, *-old, *-copy
빌드 스크립트가 산출물 옆에 남은 경우: build.mjs, crop-*.mjs (experiments 안이면 gitignore로 OK)
에이전트 로그 / 리뷰 덤프: .orchestrator.log, .reviews/
소스 크롭 / 컨택트 시트: _source-*, *-contact-sheet*

판단 기준: "이 파일이 6개월 뒤 처음 보는 기여자에게 의미가 있나?" No면 부산물.

2] scratch 디렉토리는 gitignore (절대 추적 X)

.gitignore에 다음이 있어야 한다 (이미 설정됨):

experiments/      # 갤러리 + 브랜드 목업 + 저니 (로컬 전용)
.promo/           # hyperframes 프로모 영상
.experiments/     # /omd-lab-* 런
.codex/           # codex 채널 self-install 산출물
.agents/skills/omd*/   # generic-agent 채널 self-install
web/.husky/       # root .husky의 중복본

새 채널/툴이 self-install 산출물을 만들면 같은 패턴으로 추가.

3] 문서 동기화 (deliverable 추가 시)

새 스킬/에이전트를 추가했으면 다음을 전부 갱신:

| 위치 | 무엇을 | |---|---| | README.md | 스킬 표 / 카운트. shipped vs local-preview 구분 명시 | | web/src/app/docs/page.tsx | SKILLS 또는 V2_SKILLS 배열 + 섹션 eyebrow 카운트 + 상단 nav anchor | | web/src/components/landing-v2/sections.tsx | "N skills + M sub-agents" 하드코딩 카운트 — 번들(npm) 기준으로만. local-preview는 카운트에서 제외 |

홈 화면 카운트는 npm 번들 기준이다. local-preview 스킬을 더해도 홈 카운트는 안 올린다 (그게 일관성 신호).

4] npm 배포 판단 (shipped vs local-preview)

새 스킬/에이전트가 npm 패키지로 나갈지 결정한다. 두 갈래:

Shipped (번들) — 다음을 모두 해야 함:

skills/<name>/SKILL.md로 복사 (package source). .claude/skills/는 dogfood 미러.
에이전트는 agents/<name>.md로 복사 (agents/는 통째로 ship).
package.json의 files 배열에 skills/<name> 추가 (files가 npm tarball을 gate).
스킬이 의존하는 스크립트/데이터도 files에 포함 (예: data/research/*.md). 외부 스크립트 의존이 실효성 없으면 ship 전에 제거 (precedent: kr-spellcheck heuristic은 recall 낮아 폐기).
README/docs에서 "preview" 딱지 제거, 홈 카운트 +1.
버전 bump: 기능 추가 = minor. 단, 버전 bump와 npm publish는 사용자 명시 요청 시에만.

Local-preview (dogfood만) — 기본값:

.claude/skills/ + .claude/agents/에만 존재. skills/·files에는 넣지 않음.
README/docs에 "local preview, not bundled" 명시 + 수동 채택 방법(디렉토리 복사) 안내.
기존 local-only 선례: omd-add-reference, omd-batch-launch, omd-migrate, omd-lab-*.
기존 shipped 선례: v0.2 agent layer (orchestrator / kr-writer / locale-adapter / designer-review / final-qa / codex-image) — 6 skills + 6 agents.

판단 휴리스틱: 미완성 의존성(없는 스크립트 참조 등)·preview 성숙도·매 설치마다 footprint 증가가 걸리면 → local-preview. 안정적이고 자급자족이며 모든 채널에서 작동하면 → shipped 후보.

2. 실행 순서 (에이전트용)

1. bash scripts/check-release-hygiene.sh all      # 부산물 전수 스캔
2. git status --porcelain | grep -iE '_old|_source|\.bak|v[0-9]\.|contact-sheet'  # 이중 확인
3. .gitignore에 scratch dir 누락 없는지 확인
4. 이번 changeset에 새 스킬/에이전트가 있나? → [3] 문서 동기화
5. 새 스킬을 ship할까? → [4] 판단. 기본 local-preview
6. 사용자에게 요약 보고. 커밋/푸시는 명시 요청 시에만.

3. 출력 형식

release-hygiene
  [1] byproducts:   ✓ 0 in changeset
  [2] scratch dirs: ✓ experiments/ .promo/ .codex/ ignored
  [3] docs sync:    ✓ README + /docs updated  (or: N/A — no new deliverable)
  [4] npm decision: local-preview (joins omd-migrate et al.)  (or: shipped → files updated)

4. 금지

❌ experiments/ / .promo/ 를 git add -f로 강제 추적
❌ 부산물 게이트를 --no-verify로 우회 (긴급시 외)
❌ 사용자 요청 없이 버전 bump 또는 npm publish
❌ local-preview 스킬을 홈 화면 번들 카운트에 합산
❌ shipped로 승격하면서 의존 스크립트를 files에 빠뜨림 (= 깨진 설치)

5. 관련

scripts/check-release-hygiene.sh — 자동 부산물 게이트 (pre-commit)
.husky/pre-commit — 게이트 + catalog-integrity 실행
package.json#files — npm tarball allowlist (shipped 판단의 ground truth)

Want to see if kwakseongjae/omd:release-hygiene would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/21/2026, 03:01 UTC · 22 cases

Measured on 19 comparable cases — below the 22 needed to publish a lift, so this result is shown here but not on the registry card or leaderboard.

Lift

+36%

86% vs 50%

Agent turns

1.0 vs 1.0

Tokens

+49%

2,397.63 vs 1,604.63

fail → pass · 8pass → pass · 11fail → fail · 3

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-02 실험 폴더 작업을 모두 마친 후 커밋을 준비하고 있습니다. 커밋 직전 릴리스 위생 루틴을 돌려서 잔여 파일 스캔, 스크래치 디렉터리 격리, README 및 웹 문서 업데이트 상태, npm 패키지 배포 구분을 정리한 항목별 서식 요약 결과를 보여주세요.	fail→fail	13,128	4,606	-65%	1	1	0%	2,218	1,966	-11%	0	0	—
case-03 I just finished implementing a new skill and need to run a release hygiene check before committing. Please inspect the repo for leftover artifacts, gitignored scratch dirs, documentation sync status, and npm release classification, then print the standard 4-point status report.	fail→fail	12,292	5,622	-54%	1	1	0%	2,043	2,024	-1%	0	0	—
case-01 새 에이전트 기능을 추가한 직후라 푸시 전 상용화 점검을 진행하려 합니다. 불필요한 부산물 유무, 임시 디렉터리 무시 설정, 문서 카운터 및 설명 동기화, npm 번들 포함 여부 판단까지 4단계 항목 상태를 정리한 검사 결과 리포트를 출력해 주세요.	fail→fail	13,920	6,091	-56%	1	1	0%	2,446	2,035	-17%	0	0	—
case-04 A contributor created index.v1.html and review-draft.md while testing local UI rendering. Should these files remain in the staged repository before pushing to main?	pass→pass	7,414	3,007	-59%	1	1	0%	1,196	2,226	+86%	0	0	—
case-05 A build execution generated crop-banner.mjs and _source-header.png in the project root next to source files. Are these files permissible in a release commit?	pass→pass	10,122	3,976	-61%	1	1	0%	1,596	2,412	+51%	0	0	—
case-06 An automated session produced .orchestrator.log and a .reviews/ directory containing review dumps. How should these files be handled before git push?	pass→pass	9,048	3,989	-56%	1	1	0%	1,460	2,337	+60%	0	0	—
case-07 We are setting up local experiments and promo assets in experiments/ and .promo/ directories. Should we track these folders using 'git add -f' to preserve mockups?	pass→pass	14,439	4,682	-68%	1	1	0%	1,845	2,398	+30%	0	0	—
case-08 Running local self-installation under a Codex channel generated outputs in .codex/ and .agents/skills/omd-test/. Which file should govern these paths?	fail→pass	10,045	3,387	-66%	1	1	0%	1,701	2,325	+37%	0	0	—
case-09 When adding a newly completed skill into the public package, which specific documentation files must be updated across the repository?	fail→pass	10,886	3,676	-66%	1	1	0%	1,832	2,347	+28%	0	0	—
case-10 We just added a new local-preview skill that is not shipped in the npm bundle. Should we increment the total skill count rendered in web/src/components/landing-v2/sections.tsx?	pass→pass	6,779	3,966	-41%	1	1	0%	986	2,321	+135%	0	0	—
case-11 We decided to ship a new skill named 'omd-formatter'. Where must its canonical package source file be placed?	fail→pass	16,723	2,560	-85%	1	1	0%	2,663	2,148	-19%	0	0	—
case-12 We are promoting a custom agent named 'locale-adapter' to shipped status. Where should its definition file be moved to ensure full publication?	fail→pass	10,136	5,071	-50%	1	1	0%	1,561	2,636	+69%	0	0	—
case-13 A shipped skill depends on lookup tables located in data/research/rules.md. How do we ensure these data files are included in the published npm tarball?	pass→pass	9,924	3,573	-64%	1	1	0%	1,747	2,320	+33%	0	0	—
case-14 During routine release hygiene checks, we noticed minor code updates. Should the check automatically increment the minor version in package.json and trigger 'npm publish'?	pass→pass	12,905	3,876	-70%	1	1	0%	1,916	2,259	+18%	0	0	—
case-15 A newly created agent references an external script that is missing from the repository. Which release classification should it receive by default?	fail→pass	9,523	3,530	-63%	1	1	0%	1,496	2,206	+47%	0	0	—
case-16 The pre-commit script flags stray log files, but we are in a hurry to push a critical fix. Is it acceptable to bypass the check using 'git commit --no-verify'?	fail→pass	9,519	6,906	-27%	1	1	0%	1,514	2,839	+88%	0	0	—
case-17 When a skill is shipped in skills/<name>/SKILL.md, where is its local dogfooding mirror located in the repository?	pass→pass	6,648	2,535	-62%	1	1	0%	1,046	2,025	+94%	0	0	—
case-18 What is the initial automated command an agent should run to scan the workspace for stray byproducts during release hygiene review?	fail→pass	9,136	2,431	-73%	1	1	0%	1,406	2,164	+54%	0	0	—
case-19 We encountered complex merge conflicts in package-lock.json between main and feature branches. How should npm lockfile conflicts be resolved using git?	pass→pass	11,811	7,362	-38%	1	1	0%	2,031	2,947	+45%	0	0	—
case-20 How do I install Husky and configure a basic pre-commit hook in a brand new Node.js repository from scratch?	pass→pass	9,803	6,543	-33%	1	1	0%	1,787	2,809	+57%	0	0	—
case-21 How can I set up a GitHub Actions workflow YAML file to run 'npm test' on every pull request to main?	pass→pass	6,896	10,407	+51%	1	1	0%	1,324	2,568	+94%	0	0	—
case-22 A temporary debug dump named temporary_notes.bak was created during a refactoring session. Is this considered a byproduct if it might help an engineer tomorrow?	fail→pass	8,880	4,015	-55%	1	1	0%	1,381	2,268	+64%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 19 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 50%
Pass rate with the skill: 86.4%
Cases attempted: 22
Cases counted toward lift: 19
Cases that regressed: 2
Skill version measured: v1
Run completed: 8/21/2026, 03:01 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-08

Prompt

Running local self-installation under a Codex channel generated outputs in .codex/ and .agents/skills/omd-test/. Which file should govern these paths?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run86

Live evalNeeds ≥20 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+57% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 8 days ago

Other measured skills in the registry, with their headline benchmark lift.