요원 스킬은 클로드 코드로 로드됩니다.

레이아웃

기술은 skills/ 아래에 위치하며, 다음과 같은 버킷으로 묶어집니다:

• engineering/ — 일일 코드 작업
• productivity/ — 일일 비코드 워크플로우 도구
• misc/ — 주변에 남아 있었지만 거의 사용되지 않았습니다
• personal/ — 제 세팅에 연동되어 승진하지 않음
• in-progress/ — 아직 출하 준비가 되지 않은 초안들
• deprecated/ — 더 이상 사용되지 않음

각 스킬은 SKILL.md(YAML 프론트매터 — name 및 [[T9])와 번들된 스크립트 또는 참조 파일을 포함하는 별도의 디렉터리입니다.

설치

npx skills (추천 - 모든 요원에게 효과)

대안 - 배시 스크립트

제 개발 루프에서는 — 모든 커플 가능한 스킬을 ~/.claude/skills/에 심칭 링크하세요:

저장소에 있는 모든 SKILL.md를 나열하세요:

참고문헌

공학

• debug-mantra — 네 가지 만트라 디버깅 규율: 실패 경로를 재현→ 추적→ 가설을 반박→ 모든 빵 부스러기를 교차 참조. 세션 시작 시 원문 그대로 암기한 후 수정 전에 순서대로 적용.
• 사후 분석 — 고정된 버그의 정식 엔지니어링 기록 — 근본 원인, 메커니즘, 수정, 검증, 그리고 어떻게 빠져나갔는지.

9arm-skills

Agent skills loaded by Claude Code.

Layout

Skills live under skills/, grouped into buckets:

• engineering/ — daily code work
• productivity/ — daily non-code workflow tools
• misc/ — kept around but rarely used
• personal/ — tied to my own setup, not promoted
• in-progress/ — drafts not yet ready to ship
• deprecated/ — no longer used

Each skill is its own directory containing a SKILL.md (with YAML frontmatter — name and description) and any bundled scripts or reference files.

Install

With npx skills (Recommended - works for every agent)

npx skills add thananon/9arm-skills

Alternative - Bash script

For my own dev loop — symlink every shippable skill into ~/.claude/skills/:

./scripts/link-skills.sh

List every SKILL.md in the repo:

./scripts/list-skills.sh

Reference

Engineering

• debug-mantra — Four-mantra debugging discipline: reproduce → trace the fail path → falsify the hypothesis → cross-reference every breadcrumb. Recites verbatim at session start, then applies in order before any fix.
• post-mortem — Write the canonical engineering record of a fixed bug — root cause, mechanism, fix, validation, how it slipped through. Engineer-audience; refuses to draft without a reliable repro, known cause, and validated fix.
• scrutinize — Outsider-perspective end-to-end review of a plan, PR, or code change. Questions intent (is there a simpler way?), traces the actual code path, and verifies the change does what it claims. Output is concise, actionable, with rationale.
• qwen-agent — Delegate menial, well-scoped coding tasks to a cheap Qwen-backed subagent via the claude-9arm command. Use for bulk renames, formatting, boilerplate, grep-style summarization, scaffolding, and build/test reporting.

Productivity

• management-talk — Rewrite engineer-to-engineer content for engineering-org leadership and shape it for the channel it’s going to (JIRA, Slack, async standup, email, meeting talking-points).
• qwenchance — Keeps a long Claude Code task on-track — breaks loops, bounds internal reasoning, watches the context budget, and triggers a clean handoff before the window fills.

Misc

(none yet)