setup

name	setup
description	Auto-detect project tech stack and install matching codekit skills. Use when setting up a new project, onboarding to an existing codebase, or when the user says "setup skills", "install skills for this project", "detect my stack", or "what skills do I need".

Setup — Auto-Install Compatible Skills

Scan the current project, detect its tech stack, and install the matching codekit skills automatically.

Workflow

1. Detect — Scan project files to identify technologies, language, and project type
2. Match — Map detected technologies to codekit skills
3. Confirm — Show the user what will be installed
4. Install — Run codekit skills add <name> for each match
5. Update — Run codekit learn --all -y to regenerate platform rules with installed skills

---

Step 1: Detect Project Tech Stack

Scan the project root for these signals. Check them in order — stop early when the language ecosystem is clear.

JavaScript / TypeScript Ecosystem

Signal	What it tells you
package.json	Read dependencies, devDependencies, scripts, bin, main, exports
tsconfig.json	TypeScript project
bun.lockb / bunfig.toml	Bun runtime
next.config.*	Next.js fullstack
nuxt.config.*	Nuxt fullstack
svelte.config.*	SvelteKit fullstack
vite.config.*	Vite frontend
app.json / app.config.js with expo	Expo mobile
Podfile / ios/	iOS native
android/	Android native

Python Ecosystem

Signal	What it tells you
uv.lock / pyproject.toml with [tool.uv]	Python + uv package manager
pyproject.toml	Python project (check [project.dependencies] or [tool.poetry.dependencies])
requirements.txt	Python project (read for package names)
Pipfile	Python + pipenv
setup.py / setup.cfg	Python package

Other Languages

Signal	What it tells you
Cargo.toml	Rust project
go.mod	Go project
*Gemfile / .gemspec	Ruby project
composer.json	PHP project
**.csproj / .sln	.NET / C# project
pom.xml / build.gradle	Java / Kotlin project
Package.swift	Swift project
mix.exs	Elixir project
Makefile + .c/.cpp/*.h	C/C++ project

Cross-Language Signals

Signal	What it tells you
.github/workflows/	GitHub Actions CI/CD
.git/	Git repository
docker-compose.yml / Dockerfile	Containerized project
prisma/schema.prisma	Prisma ORM (PostgreSQL/MySQL/MongoDB)
drizzle.config.*	Drizzle ORM
mongod.conf / .mongodb	MongoDB

---

Step 2: Detect Project Type

Determine the project type — this decides the Cook workflow phases:

Signal	Project Type	Base Template	Phases
API routes, controllers, ORM, no UI — NestJS, ElysiaJS, FastAPI, Express, Hono	Backend	cook-backend	Plan → Code → Review → Test
React/Vue/Svelte SPA, CSS, components, no API routes	Frontend	cook-frontend	Plan → Design → Code → Review → Test
React Native, Expo, iOS/Android	Mobile	cook-mobile	Plan → Design → Code → Review → Test
Next.js, Nuxt, SvelteKit (API + UI in one project)	Fullstack	cook-fullstack	Plan → Design → Code API → Code UI → Review → Test
bin field, commander/yargs/meow/oclif, CLI entry	CLI	cook-cli	Plan → Code → Review → Test
main/exports in package.json, no app; or Python package with setup.py/pyproject.toml	Library	cook-library	Plan → Design → Code → Review → Test

If mixed or unclear, use Fullstack.

---

Step 3: Match Technologies to Skills

Use these lookup tables. Install a skill when any of its triggers match.

Always Install

These skills benefit every project regardless of language:

Skill	Reason
memory-bank	Persistent context across sessions
git-expert	Every project uses git
agents	Universal AI agent configuration (agents.md standard)

Python Detection

Trigger	Skills to Install
fastapi in Python deps (requirements.txt, pyproject.toml, Pipfile)	fastapi, python-fastapi
sqlalchemy, alembic in Python deps	database-expert
pymongo, motor, mongoengine in Python deps	database-expert
psycopg2, asyncpg in Python deps	database-expert
pgvector in Python deps	pgvector, database-expert
pytest, unittest in Python deps	testing-expert
ruff, flake8, pylint, black in Python deps	linting-expert

Note: Python with uv uses pyproject.toml ([project.dependencies]) or uv.lock. Read these files to detect packages.

TypeScript / JavaScript Detection

Trigger	Skills to Install
typescript in devDeps OR tsconfig.json exists	typescript-expert
react in deps (without react-native)	react
next in deps	nextjs
react OR next in deps	uiux-design-expert, web-design-guidelines, vercel-react-best-practices, vercel-composition-patterns, chrome-devtools
@nestjs/core in deps	typescript-expert, database-expert, testing-expert
elysia in deps AND bun runtime	elysiajs-ddd
elysia + mongoose in deps	elysiajs-ddd-mongoose (instead of elysiajs-ddd)
elysia + (prisma OR drizzle OR pg) in deps	elysiajs-ddd
react-native in deps (without expo)	react-native, react-native-best-practices, vercel-react-native-skills, mobile-app-distribution
expo in deps OR app.json with expo config	react-native-expo, react-native-best-practices, vercel-react-native-skills, mobile-app-distribution

React Recommended Stack

When react or next is detected, also check for these recommended frontend libraries. If they are not present in deps, suggest them as the recommended stack:

Library	Purpose	Package Name
TanStack Query	Server state & data fetching	@tanstack/react-query
TanStack Table	Headless table/datagrid	@tanstack/react-table
Zod	Schema validation	zod
Zustand	Client state management	zustand
Axios	HTTP client	axios
React Hook Form	Form handling	react-hook-form
Recharts	Charts & data visualization	recharts

Behavior:

• If the project already uses these libraries, no action needed — they're part of the stack.
• If the project is new or missing these, mention them as the recommended React stack when confirming with the user.
• recharts is optional — only recommend if the project has chart/dashboard features or the user asks for charts.

Database Detection

Trigger	Skills to Install
pg, postgres, @neondatabase/serverless, knex, prisma, drizzle, sequelize, typeorm in JS/TS deps	database-expert
mongoose, mongodb, @typegoose/typegoose in JS/TS deps	database-expert
mysql2 in JS/TS deps	database-expert
pgvector in deps OR SQL files with vector type	pgvector, database-expert
prisma/schema.prisma exists	database-expert

Architecture Detection

Trigger	Skills to Install
@nestjs/core OR elysia in deps with DDD-style folder structure (domain/, application/, infrastructure/)	clean-architecture-ddd, domain-driven-hexagon
@nestjs/cqrs in deps	domain-driven-hexagon
FastAPI with layered structure (domain/, services/, repositories/)	clean-architecture-ddd

Tooling Detection

Trigger	Skills to Install
jest, vitest, mocha, or @testing-library/* in JS/TS deps	testing-expert
eslint or prettier in deps/devDeps	linting-expert
.github/ directory exists	github, github-actions
commander, yargs, meow, or oclif in deps	cli-expert

Other Languages (Baseline Skills)

For languages without specialized skills, install the baseline:

Language	Skills to Install
Rust (Cargo.toml)	memory-bank, git-expert, testing-expert + cook (auto-generated for backend or library)
Go (go.mod)	memory-bank, git-expert, testing-expert + cook (auto-generated for backend or library)
Ruby (Gemfile)	memory-bank, git-expert, database-expert, testing-expert + cook (auto-generated for backend)
PHP (composer.json)	memory-bank, git-expert, database-expert, testing-expert + cook (auto-generated for backend)
.NET (*.csproj)	memory-bank, git-expert, database-expert, testing-expert + cook (auto-generated for backend or fullstack)
Java/Kotlin (pom.xml/build.gradle)	memory-bank, git-expert, database-expert, testing-expert + cook (auto-generated for backend)
Swift (Package.swift)	memory-bank, git-expert + cook (auto-generated for mobile or library)
Elixir (mix.exs)	memory-bank, git-expert, database-expert, testing-expert + cook (auto-generated for backend)
C/C++ (Makefile + .c/.h)	memory-bank, git-expert, testing-expert + cook (auto-generated for library)

For these languages, add github + github-actions if .github/ exists, and linting-expert if linter config files exist.

---

Step 4: Skip Rules

Do NOT install skills that conflict:

• react and nextjs are alternatives — install nextjs if both react and next are in deps
• react-native and react-native-expo are alternatives — install based on whether expo is present
• elysiajs-ddd and elysiajs-ddd-mongoose are alternatives — install based on whether mongoose is present
• python-fastapi and fastapi — install both (they complement each other with different depth)
• Do NOT install pre-built cook variants — always auto-generate cook (see Step 7)

---

Step 5: Confirm with User

Before installing, show the detection results and proposed skills:

Detected: TypeScript, Bun, ElysiaJS, Mongoose, GitHub Actions
Project type: Backend

Skills to install (11):
  Always:    memory-bank, agents, git-expert
  Auto-gen:  cook (customized for backend — ElysiaJS + Mongoose)
  Framework: elysiajs-ddd-mongoose, typescript-expert
  Database:  database-expert
  Arch:      clean-architecture-ddd, domain-driven-hexagon
  Tooling:   testing-expert, github, github-actions

Proceed? (Y/n)

If the user declines, ask which skills to skip.

---

Step 6: Install Skills

Run each install command:

codekit skills add <name>

Install sequentially to avoid conflicts. If a skill is already installed, skip it — codekit skills add will report it.

After installing, check for related skills the user might also want:

codekit skills related <name>

---

Step 7: Generate Cook Skill

Do NOT install a pre-built cook variant. Instead, generate a custom cook skill tailored to this project.

How to Generate

1. Read the matching base cook template: codekit skills add <cook-variant> to get the base, OR read it from .claude/skills/cook-<type>/SKILL.md
2. Generate .claude/skills/cook/SKILL.md by customizing the base template with project-specific details
3. Remove the base variant after generating (only keep the customized cook)

What to Customize

Replace generic terms with actual project details:

Generic	Replace With
"API routes, controllers"	Actual framework (e.g., "NestJS controllers", "FastAPI routers", "ElysiaJS handlers")
"models, migrations, schemas"	Actual ORM (e.g., "Prisma models", "Mongoose schemas", "SQLAlchemy models + Alembic migrations")
"Run the full test suite"	Actual command (e.g., "bun run test", "pytest", "vitest run")
"Run type checking"	Actual command (e.g., "tsc --noEmit", "mypy .")
"Run linting"	Actual command (e.g., "eslint .", "ruff check .")
"components, pages, state"	Actual stack (e.g., "React components with Zustand store", "Next.js pages with TanStack Query")
"data fetching"	Actual library (e.g., "TanStack Query", "SWR", "Axios")
"form handling"	Actual library (e.g., "React Hook Form + Zod validation")
"package manager"	Actual PM (e.g., "bun", "uv", "pnpm")

Generated Cook Structure

The generated .claude/skills/cook/SKILL.md must always include:

---
name: cook
description: [Project-specific description with actual tech stack]
---

# Cook — [Project Type] Workflow for [Project Name]

[Phase diagram]

---

## Memory Bank — Required for Every Session
[Always include this section — copy from any cook variant]

---

## [PLAN] Plan
[Customized with actual project context]

## [DESIGN] Design (if applicable)
[Customized with actual component/API patterns]

## [CODE] Code
[Customized with actual framework, ORM, patterns]

## [REVIEW] Review
[Customized checklist with actual tech-specific checks]

## [TEST] Test
[Customized with actual test runner, commands, patterns]

---

## Iteration
[Standard iteration rules]

## Phase Markers
[Standard phase markers]

Example: Generated Cook for a Next.js + Prisma + TanStack Project

---
name: cook
description: Workflow orchestration for the [project-name] fullstack project — Next.js 15, Prisma, TanStack Query, Zustand. Phases: Plan, Design, Code API, Code UI, Review, Test.
---

# Cook — Fullstack Workflow for [project-name]

Plan → Design → Code (API first) → Code (UI) → Review → Test

## Memory Bank — Required for Every Session
[... standard memory-bank section ...]

## [PLAN] Plan
1. Read requirements. Identify both API routes and page/component changes.
2. Break into backend subtasks (Prisma models, API routes) and frontend subtasks (components, pages).
3. Define the API contract — Next.js API route request/response shapes.
4. Define acceptance criteria for the full user flow.

## [DESIGN] Design
1. Design API contract first — Next.js API routes are the bridge.
2. Define component hierarchy and TanStack Query data fetching strategy.
3. Plan Zustand store shape for client state.
4. Plan loading states, optimistic UI, and error handling.

## [CODE] Code — API First
1. Implement Prisma schema changes and run `bunx prisma db push`.
2. Implement Next.js API routes / Server Actions.
3. Validate API works with a test or manual check.
4. Then implement React components with TanStack Query for data fetching.
5. Wire up React Hook Form + Zod for form validation.
6. Use Zustand for client-side state where needed.

## [REVIEW] Review
1. **Prisma** — Schema correct? Migrations clean? No N+1 queries?
2. **API routes** — Correct HTTP methods? Input validated with Zod? Proper error responses?
3. **Components** — TanStack Query keys correct? Loading/error states handled?
4. **Forms** — React Hook Form + Zod validation working? Error messages shown?
5. **Types** — No `any`? Prisma types flowing through to components?
6. **Security** — Auth checks on protected routes? No exposed secrets?

## [TEST] Test
1. `bun run test` — Run Vitest suite.
2. `tsc --noEmit` — Type check passes.
3. `eslint .` — No lint errors.
4. Test API routes with edge case inputs.
5. Visual check — run `bun run dev` and verify the feature.

---

Step 8: Update Platform Rules

After all skills are installed and cook is generated, regenerate platform rules:

codekit learn --all -y

This updates CLAUDE.md, .cursorrules, GEMINI.md, and AGENTS.md with the installed skills section.

---

Example Detection

Python FastAPI + uv Project (Backend)

# pyproject.toml
[project]
dependencies = [
    "fastapi>=0.115.0",
    "sqlalchemy[asyncio]>=2.0",
    "asyncpg>=0.30.0",
    "alembic>=1.14.0",
]

[tool.uv]
dev-dependencies = [
    "pytest>=8.0",
    "ruff>=0.8.0",
]

Detected: Python, uv, FastAPI, SQLAlchemy, asyncpg (PostgreSQL), pytest, ruff Project type: Backend (FastAPI API, no UI)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for backend)
• fastapi, python-fastapi (fastapi in deps)
• database-expert (sqlalchemy + asyncpg in deps)
• testing-expert (pytest in dev-deps)
• linting-expert (ruff in dev-deps)
• github, github-actions (if .github/ exists)

NestJS + PostgreSQL Project (Backend)

{
  "dependencies": {
    "@nestjs/core": "^10.0.0",
    "@nestjs/typeorm": "^10.0.0",
    "typeorm": "^0.3.0",
    "pg": "^8.0.0"
  },
  "devDependencies": {
    "typescript": "5.x",
    "jest": "29.x",
    "eslint": "9.x"
  }
}

Detected: TypeScript, NestJS, TypeORM, PostgreSQL, Jest, ESLint Project type: Backend (NestJS, no UI)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for backend)
• typescript-expert (typescript in devDeps)
• database-expert (typeorm + pg in deps)
• clean-architecture-ddd, domain-driven-hexagon (NestJS architecture)
• testing-expert (jest in devDeps)
• linting-expert (eslint in devDeps)
• github, github-actions (if .github/ exists)

Bun ElysiaJS + MongoDB Project (Backend)

{
  "dependencies": {
    "elysia": "^1.2.0",
    "mongoose": "^8.0.0",
    "@elysiajs/cors": "^1.0.0"
  },
  "devDependencies": {
    "typescript": "5.x",
    "bun-types": "latest"
  }
}

Detected: TypeScript, Bun, ElysiaJS, Mongoose (MongoDB) Project type: Backend (ElysiaJS API, no UI)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for backend)
• elysiajs-ddd-mongoose (elysia + mongoose in deps)
• typescript-expert (typescript in devDeps)
• database-expert (mongoose in deps)
• clean-architecture-ddd, domain-driven-hexagon (DDD architecture)
• testing-expert

Bun ElysiaJS + PostgreSQL Project (Backend)

{
  "dependencies": {
    "elysia": "^1.2.0",
    "prisma": "^6.0.0",
    "@prisma/client": "^6.0.0"
  },
  "devDependencies": {
    "typescript": "5.x",
    "bun-types": "latest"
  }
}

Detected: TypeScript, Bun, ElysiaJS, Prisma (PostgreSQL) Project type: Backend (ElysiaJS API, no UI)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for backend)
• elysiajs-ddd (elysia + prisma in deps, no mongoose)
• typescript-expert (typescript in devDeps)
• database-expert (prisma in deps)
• clean-architecture-ddd, domain-driven-hexagon (DDD architecture)
• testing-expert

Next.js + Prisma Project (Fullstack)

{
  "dependencies": {
    "next": "15.x",
    "react": "19.x",
    "@prisma/client": "6.x",
    "@tanstack/react-query": "^5.0.0",
    "@tanstack/react-table": "^8.0.0",
    "zustand": "^5.0.0",
    "axios": "^1.7.0",
    "react-hook-form": "^7.0.0",
    "zod": "^3.23.0"
  },
  "devDependencies": {
    "typescript": "5.x",
    "eslint": "9.x",
    "vitest": "3.x"
  }
}

Detected: TypeScript, Next.js, React, Prisma, TanStack Query, TanStack Table, Zustand, Axios, React Hook Form, Zod, ESLint, Vitest Project type: Fullstack (Next.js = API + UI)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for fullstack)
• nextjs (next in deps, prefer over react since both present)
• typescript-expert (typescript in devDeps)
• uiux-design-expert, web-design-guidelines, vercel-react-best-practices, vercel-composition-patterns, chrome-devtools (frontend stack)
• database-expert (prisma in deps)
• testing-expert (vitest in devDeps)
• linting-expert (eslint in devDeps)
• github, github-actions (if .github/ exists)

Recommended stack detected: TanStack Query, TanStack Table, Zustand, Axios, React Hook Form, Zod

React SPA Project (Frontend)

{
  "dependencies": {
    "react": "19.x",
    "react-dom": "19.x",
    "@tanstack/react-query": "^5.0.0",
    "@tanstack/react-table": "^8.0.0",
    "zustand": "^5.0.0",
    "axios": "^1.7.0",
    "react-hook-form": "^7.0.0",
    "zod": "^3.23.0",
    "recharts": "^2.0.0"
  },
  "devDependencies": {
    "typescript": "5.x",
    "vite": "6.x",
    "eslint": "9.x",
    "vitest": "3.x"
  }
}

Detected: TypeScript, React, Vite, TanStack Query, TanStack Table, Zustand, Axios, React Hook Form, Zod, Recharts, ESLint, Vitest Project type: Frontend (React SPA, no API layer)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for frontend)
• react (react in deps, no next)
• typescript-expert (typescript in devDeps)
• uiux-design-expert, web-design-guidelines, vercel-react-best-practices, vercel-composition-patterns, chrome-devtools (frontend stack)
• testing-expert (vitest in devDeps)
• linting-expert (eslint in devDeps)

Recommended stack detected: TanStack Query, TanStack Table, Zustand, Axios, React Hook Form, Zod, Recharts

React Native Expo Project (Mobile)

{
  "dependencies": {
    "expo": "52.x",
    "react-native": "0.76.x",
    "expo-router": "4.x"
  },
  "devDependencies": {
    "typescript": "5.x",
    "jest": "29.x"
  }
}

Detected: TypeScript, Expo, React Native, Jest Project type: Mobile (React Native + Expo)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for mobile)
• react-native-expo (expo in deps)
• react-native-best-practices, vercel-react-native-skills (mobile enhancers)
• mobile-app-distribution (mobile project)
• typescript-expert (typescript in devDeps)
• testing-expert (jest in devDeps)

CLI Tool Project (CLI)

{
  "bin": { "mytool": "./dist/index.js" },
  "dependencies": {
    "commander": "^12.0.0"
  },
  "devDependencies": {
    "typescript": "5.x",
    "vitest": "3.x"
  }
}

Detected: TypeScript, Commander, Vitest Project type: CLI (bin field + commander)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for CLI)
• cli-expert (commander in deps)
• typescript-expert (typescript in devDeps)
• testing-expert (vitest in devDeps)

Go Backend Project (Backend)

// go.mod
module github.com/user/api
go 1.23

require (
    github.com/gin-gonic/gin v1.10.0
    github.com/jackc/pgx/v5 v5.7.0
)

Detected: Go, Gin, pgx (PostgreSQL) Project type: Backend (API with database)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for backend)
• database-expert (pgx = PostgreSQL)
• testing-expert (Go has built-in testing)
• github, github-actions (if .github/ exists)

Rust Library Project (Library)

# Cargo.toml
[package]
name = "my-lib"
version = "0.1.0"

[lib]
name = "my_lib"

[dev-dependencies]
tokio-test = "0.4"

Detected: Rust, library crate Project type: Library ([lib] section in Cargo.toml)

Installed skills:

• memory-bank, agents, git-expert (always) + cook (auto-generated for library)
• testing-expert
• github, github-actions (if .github/ exists)