Production-ready Claude Code plugins for ShipWithAI projects.
Website · Report Bug · Request Feature
Each plugin in this repo ships as a Claude Code skill set — install once, use via slash command or natural language. No build step, no runtime dependencies.
| Plugin | Skills | What it does |
|---|---|---|
shipwithai-starter |
/starter:init · /starter:review · /starter:update-ssot · and 6 more |
Sets up a standardized Claude Code harness for any project. One guided interview configures CLAUDE.md, permissions, hooks, MCP servers, agents, and SSOT docs. Three tiers: Essential (5 min), Standard (15 min), Full (30 min). |
shipwithai-auth |
/auth:setup · /auth:doctor |
Full auth stack for Next.js: Better Auth or Firebase, Google OAuth, UI pages, middleware, DB schema. Doctor scans 22 checks and reports health. |
Register the ShipWithAI marketplace once, then install any plugin by name:
# Register the ShipWithAI marketplace (one-time setup)
/plugin marketplace add ShipWithAI/shipwithai-plugins
# Install any plugin by name
/plugin install shipwithai-auth@shipwithaiClone this repo and point Claude Code at the plugin directory:
git clone https://github.com/ShipWithAI/shipwithai-plugins.git
claude --plugin-dir ./shipwithai-plugins/plugins/starter # or /auth, etc.Once installed, trigger a skill by slash command or describe what you need in chat:
/shipwithai-starter:init
# or
> Set up Claude Code harness for my project
/shipwithai-auth:setup
# or
> Set up Firebase Auth with Google login for my Next.js appSee each plugin's README.md for full usage, supported stack, and what gets generated.
shipwithai-plugins/
├── plugins/
│ ├── starter/ # shipwithai-starter plugin
│ │ ├── skills/
│ │ │ ├── init/ # Guided setup interview (SKILL.md + evals)
│ │ │ ├── new-project/ # Greenfield 3-phase interview → scaffold + harness
│ │ │ ├── setup-memory/ # CLAUDE.md + .claude/memory/
│ │ │ ├── setup-permissions/
│ │ │ ├── setup-hooks/ # hooks-catalog.json
│ │ │ ├── setup-mcp/ # mcp-registry.json
│ │ │ ├── setup-agents/ # agents-catalog.json
│ │ │ ├── setup-ssot/ # architecture.md + ADR + CODEMAPS
│ │ │ ├── setup-observability/ # observe.py hook + .claude/logs/ logging
│ │ │ ├── review/ # Harness health audit
│ │ │ └── update-ssot/ # Sync docs with codebase
│ │ ├── assets/ # Shared hook scripts (observe.py)
│ │ ├── agents/ # drift-monitor sub-agent
│ │ ├── commands/ # add-mcp, add-hook, add-agent, add-adr
│ │ ├── .claude-plugin/ # Plugin metadata (plugin.json)
│ │ ├── manifest.json # Skill registry
│ │ └── README.md
│ └── auth/ # shipwithai-auth plugin
│ ├── skills/
│ │ ├── auth-setup/ # Setup wizard (SKILL.md + assets + references)
│ │ └── auth-doctor/ # Health scanner (SKILL.md)
│ ├── commands/ # Slash command definitions
│ ├── manifest.json # Skill registry
│ └── README.md
├── scripts/
│ └── publish-plugin.sh # Validate and package a plugin for release
├── docs/
│ ├── ARCHITECTURE.md
│ ├── CODEMAPS/ # Codebase navigation guides
│ ├── adr/ # Architecture Decision Records
│ └── starter/ # Plugin PRDs and design specs
└── CLAUDE.md # Authoring conventions for Claude Code
No compiled code. All plugin content is Markdown (SKILL.md, references) and JSON (manifest, evals, settings).
- Read
CLAUDE.md— authoring conventions and blueprint references are required reading before any work. - Create a directory under
plugins/following the structure above. - Add
manifest.json(skill registry) andplugin.json(metadata). - Each skill needs a
SKILL.md(< 500 lines) andevals/evals.json(5+ test prompts). - Test on two real projects before opening a PR.
Open issues for bugs and feature requests. For new skills or providers, check open issues first to avoid duplicate work.
# Validate and dry-run a plugin before publishing
./scripts/publish-plugin.sh auth --dry-runMIT — see LICENSE.