shipwright
A la carte toolkit for Claude Code. Best practices, persistent memory, and agent-team planning — pick what fits your project.
Open source Open in the app JSON README (API)
About
A la carte toolkit for Claude Code. Best practices, persistent memory, and agent-team planning — pick what fits your project.
Details
- Kind
- Plugins
- Topic
- AI, RAG & memory
- Publisher
- shipwright-ai
- Origin
- marketplace
- Category
- ferramentas
- Last push
- 2026-03-30T21:57:47Z
- Repository state
- ativo
- License
- MIT
- Added
- 2026-08-30 01:48:58
- Updated
- 2026-08-30 01:48:58
- Origin id
shipwright-ai/shipwright/shipwright
README
# Shipwright
A la carte toolkit for Claude Code. Best practices, persistent memory, and agent-team planning — pick what fits your project.
## Install
**Plugin (marketplace):**
```
/plugin install shipwright
/shipwright:setup
```
**Git clone:**
```bash
git clone https://github.com/shipwright-ai/shipwright.git
# In your project, in Claude Code:
Read ../shipwright/methodology/setup.md and onboard this project
```
## Three Layers
Setup guides you through each. Stop whenever you want.
### Layer 1: Methodology
Best practices for any project. No dependencies.
- CLAUDE.md with workflow rules and context gates
- Makefile with standard targets (`make test`, `make lint`, `make dev`)
- Developer profile for personalized workflow
- Quality gates (lint, test before commit)
- Context compaction for long sessions
### Layer 2: Brain (optional, recommended)
Persistent, searchable memory via MCP. Needs npm (for npx).
- **Custom kinds** — ideas, decisions, features, bugs, or whatever fits your project
- **Dynamic tags** — organize however makes sense, no predefined taxonomy
- **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`
- **Sections** — drop .md files next to memory.md, Brain parses them (checklists, progress, agent assignment)
- **Progress from checkboxes** — no status field. Brain derives not-started / in-progress / done automatically
- **Semantic + keyword search** — finds related work across sessions, even without exact matches
- **Screenshots** — capture UI at any breakpoint (375 mobile, 768 tablet, 1280 desktop)
- **Agent recall** — agents remember learnings across sessions, loaded at every spawn
- **Duplicate detection** — Brain blocks creation if 90%+ similar memory exists
- **Self-building graph** — refs are always bidirectional, connections auto-detected
- **MCP responses steer Claude** — every tool response includes next steps and reminders. Enforcement that doesn't get compacted away
Powered by [Shipwright Brain](https://github.com/shipwright-ai/shipwright-brain). Can also be installed standalone: `/plugin install shipwright-brain`.
Browse what Claude sees with [Shipwright UI](https://github.com/shipwright-ai/shipwright-ui): `make brain-ui` → http://localhost:3111
### Layer 3: Orchestration (optional, needs Brain)
Structured planning with agent teams.
- `/capture` — capture ideas into Brain
- `/plan` — plan with team review (product + critic), present for approval
- `/execute` — execute work-items with TDD, two-stage review, docs update
**Agent team** (discovery recommends based on your project):
| Agent | Role | When |
|---|---|---|
| Product | Coordinates planning, validates against vision/personas | Planning |
| Critic | Challenges assumptions, flags overbuilding | Planning |
| Developer | Implements with TDD, follows work-item spec | Execution |
| Reviewer | Spec compliance + code quality review | After execution |
| Doc-writer | Updates feature docs, captures screenshots | After review |
Additional agents recommended per project: UX (frontend), Tech Lead (complex architecture), Security (auth/APIs).
**Gate modes** — same system, different autonomy:
- Manual: review everything
- Semi-auto: review plans only
- Full-auto: hard stops only (blocked, security, destructive)
**Data model** — everything is Brain memories with tags and nesting:
```
docs/ideas/feature-x/memory.md idea
└── plan/memory.md plan (tag: plan)
└── task-1/ work-item (tag: work-item)
├── memory.md what & why
├── 1_execution_developer.md developer checklist
├── 2_review_reviewer.md reviewer checklist
└── 3_documentary_documenter.md doc-writer checklist
```
## Structure
```
shipwright/
├── skills/setup/SKILL.md /shipwright:setup entry point
├── methodology/ Layer 1 + 2
│ ├── setup.md onboarding conversation
│ ├── phases/ 5 progressive phases
│ ├── skills/ 12 teaching skills
│ ├── hooks/ quality gates
│ └── community-insights.md learnings from developers
├── orchestration/ Layer 3
│ ├── setup.md orchestration setup
│ ├── skills/ teaching files (ideation, planning, execution)
│ ├── agents/ 5W archetypes (product, critic, developer, reviewer, doc-writer)
│ └── docs/ design decisions and architecture
├── local/ developer-local (gitignored)
├── CLAUDE.md meta-guidance for shipwright itself
└── .claude-plugin/plugin.json marketplace manifest
```
## Three Scopes
| Scope | Where | Who benefits |
|-------|-------|-------------|
| **Team** | committed to shipwright | all developers via git pull |
| **Developer** | shipwright/local/ (gitignored) | just you, across all projects |
| **Project** | your project's docs/ | whoever works on that project |
## Key Principles
- **Files, not databases** — everything is markdown on disk
- **Checklists with exact `brain.tool()` calls** — not prose essays
- **Brain MCP responses steer behavior** — enforcement that doesn't get compacted away
- **Skills teach, not template** — Claude generates project-specific output
- **Ideas are living, work-items are immutable** — new requirements = new work-items
- **Ideology stays, patterns evolve** — WHY is stable, HOW changes with models
## Inspirations
- [superpowers](https://github.com/obra/superpowers) — brainstorming, TDD, subagent execution
- [oh-my-claudecode](https://github.com/yeachan-heo/oh-my-claudecode) — deep-interview, trace, slop-cleaner
## License
MIT