By 1989v
Universal AI harness engineering: 6-layer architecture with SDD pipeline, 5-dimension review, lifecycle (GC/evolve/diet/audit), enforcement hooks, context routing, and doc generation
Compare current harness against external benchmarks — repos, posts, best practices
Generate CLAUDE.md and docs/ tree — default: fill gaps only, --full for regeneration
End-to-end feature development pipeline: shape → write → review → create-tasks → implement → validate. Trigger on: 새 기능, 기능 추가, 피처 개발, feature 만들어, 서비스 구현, new feature, add feature
Run garbage collection — detect dead code, doc drift, rule violations, stale harness
Initialize AI harness for any project — auto-scan + interactive setup + doc-gen + hooks + routing
Garbage collection agent — scans project for dead code, doc drift, rule violations, stale harness
Documentation specialist. Write specs and requirements only. No code.
Use this agent to generate CLAUDE.md and docs/ tree. Analyzes project structure, collects custom requirements, generates documentation, and validates against codebase.
Harness audit agent — compares current harness with external sources
Production code implementation specialist. No tests.
Create a task breakdown from a spec for implementation.
하네스 복잡도를 점검하고, CLAUDE.md를 지도 수준으로 압축하며, 불필요한 규칙을 제거한다.
docs/ 디렉터리를 단일 index.html 문서 사이트로 변환한다.
Evaluate whether implementation has drifted from the active spec.
실패 패턴을 하네스 규칙으로 인코딩하여 재발 방지.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
다른 서비스에 공통으로 활용할 Claude Code 플러그인 모노레포.
| 플러그인 | Prefix | 설명 |
|---|---|---|
| hns | /hns: | AI 하네스 엔지니어링 — SDD 파이프라인, 5차원 리뷰, Lifecycle(GC/evolve/diet/audit), 문서 생성 |
| ai-debugger | /ai-debugger: | API 디버깅 에이전트 — IO 캡처, curl 생성, 로그 분석 |
| private-repo | /private-repo: | git submodule로 디렉토리별 public/private 가시성 제어 |
| content-analyzer | /content-analyzer: | URL 콘텐츠 분석 (YouTube, LinkedIn, web post, Git repo) |
# marketplace 등록 (최초 1회)
claude marketplace add https://github.com/1989v/ai.git --name ai-common
# 플러그인 설치
claude plugins install hns@ai-common
claude plugins install ai-debugger@ai-common
claude plugins install private-repo@ai-common
claude plugins install content-analyzer@ai-common
# 세션 내 리로드
/reload-plugins
대상 프로젝트의 .claude/settings.json:
{
"enabledPlugins": {
"hns@ai-common": true,
"ai-debugger@ai-common": true,
"private-repo@ai-common": true,
"content-analyzer@ai-common": true
}
}
AI 하네스 엔지니어링 완전체 플러그인. 19개 커맨드.
/hns:init 하네스 스캐폴딩 (auto-scan → doc-gen → hooks → routing)
/hns:new-feature 신규 기능 파이프라인 (shape → write → review → tasks)
/hns:shape-spec 요구사항 수집 + 스펙 폴더 초기화
/hns:write-spec spec.md 작성
/hns:spec-review 5차원 스펙 리뷰 (arch/domain/impl/test/usecase)
/hns:create-tasks spec → task group 분해
/hns:implement-tasks task group 실행 (Ralph Loop)
/hns:orchestrate-tasks 순차/병렬 오케스트레이션
/hns:interview-capture 구현 전 게이트 인터뷰
/hns:drift-check 구현-스펙 불일치 감지
/hns:verify 표준 → 린트 → 빌드 → 테스트
/hns:verify-crosscheck 6레이어 교차 일관성 검증
/hns:doc-gen CLAUDE.md + docs/ 생성
/hns:doc-validate docs ↔ 코드 일치 검증
/hns:doc-html docs/ → HTML 사이트 생성
/hns:harness-gc 가비지 컬렉션 (dead code, doc drift, stale rules)
/hns:harness-evolve 실패 패턴 → 규칙 인코딩
/hns:harness-diet 불필요한 규칙 제거 (Bitter Lesson)
/hns:harness-audit 외부 벤치마크 비교
새 플러그인/커맨드/스킬을 만들 때: docs/plugin-development-guide.md
핵심 규칙:
commands/*.md는 frontmatter에 name 필드를 넣지 않는다 (prefix가 안 붙음).claude-plugin/marketplace.json에 반드시 등록ai/
├── .claude-plugin/marketplace.json # marketplace 인덱스
├── plugins/
│ ├── hns/ # 하네스 엔지니어링 (19 commands, 12 skills, 10 agents)
│ ├── ai-debugger/ # API 디버깅 (2 commands, 6 skills, 1 agent)
│ ├── private-repo/ # Private repo 분리 (1 command, 1 skill)
│ └── content-analyzer/ # 콘텐츠 분석 (1 command, 3 skills, 1 agent)
├── docs/
│ ├── plugin-development-guide.md # 플러그인 개발 가이드
│ ├── specs/ # 설계 스펙
│ └── study/ # 학습 자료
└── README.md
npx claudepluginhub 1989v/ai --plugin hnsControl per-directory public/private visibility via git submodules. Extracts directories into separate GitHub repos with visibility choice, history preservation, and submodule setup.
API debugging agent with full IO capture, log analysis, and codebase exploration
Analyze content from URLs (YouTube, LinkedIn, web posts, Git repos) and generate structured learning plans or analysis documents
Session harness plugin for Claude Code workflow automation
Harness for Claude Code — skills, /harness:* slash commands, persona subagents, lifecycle hooks, and MCP tools without per-repo `harness setup`. Sibling plugins exist for Cursor, Gemini CLI, and Codex.
Harness Engineering framework - skills, agents, and commands for safe, reviewable, incremental agent-driven development. Includes RPEQ workflow (Research, Plan, Execute, QA), ast-grep setup, and codebase analysis tools.
Makes a repo agent-ready: AGENTS.md, boundary tests, CI pipeline, GC scripts — based on OpenAI's harness engineering methodology
Agents specialized in prompt engineering and AI interaction. Focuses on effective prompt design and AI model optimization.
Specification-first AI harness: 11 structural gates, 11 Ouroboros commands, 11 agent personas, and 3-tier architecture enforcement. v2.1: Pair Mode (Navigator-Driver + independent test design + /review command).