{
  "markdown": "# Shipwright\n\nA la carte toolkit for Claude Code. Best practices, persistent memory, and agent-team planning — pick what fits your project.\n\n## Install\n\n**Plugin (marketplace):**\n```\n/plugin install shipwright\n/shipwright:setup\n```\n\n**Git clone:**\n```bash\ngit clone https://github.com/shipwright-ai/shipwright.git\n# In your project, in Claude Code:\nRead ../shipwright/methodology/setup.md and onboard this project\n```\n\n## Three Layers\n\nSetup guides you through each. Stop whenever you want.\n\n### Layer 1: Methodology\n\nBest practices for any project. No dependencies.\n\n- CLAUDE.md with workflow rules and context gates\n- Makefile with standard targets (`make test`, `make lint`, `make dev`)\n- Developer profile for personalized workflow\n- Quality gates (lint, test before commit)\n- Context compaction for long sessions\n\n### Layer 2: Brain (optional, recommended)\n\nPersistent, searchable memory via MCP. Needs npm (for npx).\n\n- **Custom kinds** — ideas, decisions, features, bugs, or whatever fits your project\n- **Dynamic tags** — organize however makes sense, no predefined taxonomy\n- **Format guides per kind** — each kind has its own template that Brain uses on creation. Tag-aware: `ideas` tagged `plan` get a different format than plain `ideas`\n- **Sections** — drop .md files next to memory.md, Brain parses them (checklists, progress, agent assignment)\n- **Progress from checkboxes** — no status field. Brain derives not-started / in-progress / done automatically\n- **Semantic + keyword search** — finds related work across sessions, even without exact matches\n- **Screenshots** — capture UI at any breakpoint (375 mobile, 768 tablet, 1280 desktop)\n- **Agent recall** — agents remember learnings across sessions, loaded at every spawn\n- **Duplicate detection** — Brain blocks creation if 90%+ similar memory exists\n- **Self-building graph** — refs are always bidirectional, connections auto-detected\n- **MCP responses steer Claude** — every tool response includes next steps and reminders. Enforcement that doesn't get compacted away\n\nPowered by [Shipwright Brain](https://github.com/shipwright-ai/shipwright-brain). Can also be installed standalone: `/plugin install shipwright-brain`.\n\nBrowse what Claude sees with [Shipwright UI](https://github.com/shipwright-ai/shipwright-ui): `make brain-ui` → http://localhost:3111\n\n### Layer 3: Orchestration (optional, needs Brain)\n\nStructured planning with agent teams.\n\n- `/capture` — capture ideas into Brain\n- `/plan` — plan with team review (product + critic), present for approval\n- `/execute` — execute work-items with TDD, two-stage review, docs update\n\n**Agent team** (discovery recommends based on your project):\n\n| Agent | Role | When |\n|---|---|---|\n| Product | Coordinates planning, validates against vision/personas | Planning |\n| Critic | Challenges assumptions, flags overbuilding | Planning |\n| Developer | Implements with TDD, follows work-item spec | Execution |\n| Reviewer | Spec compliance + code quality review | After execution |\n| Doc-writer | Updates feature docs, captures screenshots | After review |\n\nAdditional agents recommended per project: UX (frontend), Tech Lead (complex architecture), Security (auth/APIs).\n\n**Gate modes** — same system, different autonomy:\n- Manual: review everything\n- Semi-auto: review plans only\n- Full-auto: hard stops only (blocked, security, destructive)\n\n**Data model** — everything is Brain memories with tags and nesting:\n```\ndocs/ideas/feature-x/memory.md                             idea\n  └── plan/memory.md                                        plan (tag: plan)\n       └── task-1/                                          work-item (tag: work-item)\n            ├── memory.md                                   what & why\n            ├── 1_execution_developer.md                    developer checklist\n            ├── 2_review_reviewer.md                        reviewer checklist\n            └── 3_documentary_documenter.md                 doc-writer checklist\n```\n\n## Structure\n\n```\nshipwright/\n├── skills/setup/SKILL.md        /shipwright:setup entry point\n├── methodology/                 Layer 1 + 2\n│   ├── setup.md                 onboarding conversation\n│   ├── phases/                  5 progressive phases\n│   ├── skills/                  12 teaching skills\n│   ├── hooks/                   quality gates\n│   └── community-insights.md   learnings from developers\n├── orchestration/               Layer 3\n│   ├── setup.md                 orchestration setup\n│   ├── skills/                  teaching files (ideation, planning, execution)\n│   ├── agents/                  5W archetypes (product, critic, developer, reviewer, doc-writer)\n│   └── docs/                    design decisions and architecture\n├── local/                       developer-local (gitignored)\n├── CLAUDE.md                    meta-guidance for shipwright itself\n└── .claude-plugin/plugin.json   marketplace manifest\n```\n\n## Three Scopes\n\n| Scope | Where | Who benefits |\n|-------|-------|-------------|\n| **Team** | committed to shipwright | all developers via git pull |\n| **Developer** | shipwright/local/ (gitignored) | just you, across all projects |\n| **Project** | your project's docs/ | whoever works on that project |\n\n## Key Principles\n\n- **Files, not databases** — everything is markdown on disk\n- **Checklists with exact `brain.tool()` calls** — not prose essays\n- **Brain MCP responses steer behavior** — enforcement that doesn't get compacted away\n- **Skills teach, not template** — Claude generates project-specific output\n- **Ideas are living, work-items are immutable** — new requirements = new work-items\n- **Ideology stays, patterns evolve** — WHY is stable, HOW changes with models\n\n## Inspirations\n\n- [superpowers](https://github.com/obra/superpowers) — brainstorming, TDD, subagent execution\n- [oh-my-claudecode](https://github.com/yeachan-heo/oh-my-claudecode) — deep-interview, trace, slop-cleaner\n\n## License\n\nMIT\n",
  "bytes": 5933,
  "sha": "2b10e25163cd25dc6619c6542752c62a349d29ad5e3fcab31c8c220d76330701",
  "repo_slug": "shipwright-ai/shipwright",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_shipwright_ai_shipwright_shipwright_4991ff6c/readme"
}