Skip to content

EpicHigh/claude-project-architect

project-architect

Scan any codebase. Generate tailored Claude Code configuration — automatically.

License: MIT

What is this?

project-architect is a Claude Code plugin that analyzes your project's tech stack and generates a complete .claude/ configuration — CLAUDE.md, skills, agents, hooks, and MCP servers — all tailored to how your project actually works. No manual setup, no guessing.

Demo

> /project-architect

Scanning project...
✓ Detected: TypeScript, Next.js 14, Tailwind CSS, Prisma, Jest, ESLint
✓ Found: app router, PostgreSQL, GitHub Actions

Generating configuration...
✓ CLAUDE.md — project overview, build/test/lint commands, conventions
✓ Methodology skills — implement-feature, fix-bug, tdd, design-system, schema-patterns, api-patterns
✓ Workflow skills — feature-development, bug-fix-lifecycle, code-review-fix (dispatch developer + reviewer agents)
✓ Invocable skills — /commit, /review
✓ Agents — developer (Next.js+Prisma+Tailwind), code-reviewer (React+security), architect, db-specialist (Prisma+PostgreSQL)
✓ Hook — lint + test pre-commit (ESLint + Jest confirmed installed)
✓ MCP — Context7 for Next.js docs

Install

# Add the marketplace and install
/plugin marketplace add EpicHigh/claude-project-architect
/plugin install project-architect@EpicHigh-claude-project-architect

Or install locally for development:

git clone https://github.com/EpicHigh/claude-project-architect.git
claude --plugin-dir ./claude-project-architect

Usage

Open Claude Code in your project and run:

/project-architect

The plugin runs autonomously in 3 phases:

  1. Scan — reads config files, directory structure, git history (read-only, never executes code)
  2. Generate — composes configuration tailored to your project's specific stack intersection, then self-reviews and refines until no gaps remain
  3. Present — shows what was generated, why, and what the self-review improved

What Gets Generated

Skills (3 types)

Type Output When
Methodology implement-feature, fix-bug, improve-architecture Always
Methodology tdd Test framework detected
Methodology design-system Styling framework detected
Methodology api-patterns Backend framework detected
Methodology schema-patterns Database/ORM detected
Workflow feature-development, bug-fix-lifecycle, code-review-fix, etc. By project judgment
Workflow security-audit-workflow, db-optimization-workflow, etc. By detection + judgment
Invocable commit Always
Invocable review Git detected

Methodology skills auto-activate and teach HOW to approach tasks. Workflow skills orchestrate multi-step work by dispatching named agents through phases with verification gates. Invocable skills are user-triggered via /skill-name.

All skills follow Anthropic's official skill best practices. Workflow patterns inspired by mattpocock/skills and superpowers.

Agents

Output When
architect (stack-aware) Complex architecture detected
code-reviewer (stack-specific checklists) Active development detected
developer (framework-specific patterns) Test framework + linter detected
database-specialist (ORM + engine expertise) Database/ORM detected
security-engineer (threat modeling) Backend framework detected
devops-automator (Docker + CI/CD) Docker or CI/CD detected
Additional agents from agency-agents catalog By detection + judgment

No agents are mandatory — selection is by detection + judgment.

Other Layers

Layer Output When
CLAUDE.md Project overview, build/test/lint commands, conventions Always
INSTRUCTION.md Quick-start onboarding guide Always
Hooks lint pre-commit Linter detected
Hooks lint + test pre-commit Linter + fast tests detected
MCP Context7 docs server Framework with docs detected

Supported Stacks

project-architect detects and generates configuration for a wide range of technologies.

Languages

TypeScript, JavaScript, Python, Go, Rust, Ruby, Java, Kotlin, C#, Swift, Dart, PHP, Elixir, Scala, Zig

Frontend Frameworks

React, Vue, Angular, Svelte, Next.js, Nuxt, Remix, Astro, SolidJS, SvelteKit, Gatsby, Preact, Qwik, Lit, Ember

Backend Frameworks

Express, NestJS, Fastify, Hono, Koa, Gin, Chi, Echo, Fiber, ConnectRPC, gRPC, FastAPI, Django, Flask, Rails, Spring Boot, Phoenix, Actix Web

Databases & ORMs

PostgreSQL, MySQL, MongoDB, Redis, SQLite, Elasticsearch, SQL Server, DynamoDB, CockroachDB Prisma, Drizzle, TypeORM, Sequelize, Mongoose, Ent, GORM, SQLAlchemy, Django ORM, ActiveRecord

Build Tools & Package Managers

Vite, Webpack, Rollup, Parcel, esbuild, SWC, tsup, Turbopack, Gradle, Maven npm, Yarn, pnpm, Bun, Go Modules, Cargo, Poetry, uv, Bundler, Composer

Testing

Jest, Vitest, Mocha, pytest, Go testing, RSpec, JUnit, Playwright, Cypress, Puppeteer

Linters & Formatters

ESLint, Biome, Prettier, golangci-lint, Ruff, Black, RuboCop, Clippy, rustfmt

CI/CD & Cloud

GitHub Actions, GitLab CI, CircleCI, Jenkins, Vercel, Netlify, Fly.io AWS, GCP, Azure, Docker, Kubernetes, Terraform, Pulumi

Monorepo Tools

pnpm workspaces, npm workspaces, Nx, Lerna, Turborepo, Go workspaces, Cargo workspaces, Bazel

How It Works

project-architect is a Claude Code plugin — it's pure markdown and JSON, no executable code. The slash command tells Claude what to scan (detection-guide.md) and what to generate (generation-guide.md). Claude does the rest using its built-in tools (Read, Glob, Grep, Bash, Write).

Architecture:

.claude-plugin/plugin.json        → Plugin manifest
commands/project-architect.md      → Main command (3 autonomous phases)
references/detection-guide.md      → What to look for during scan
references/generation-guide.md     → Guidelines and examples for output composition

Examples

See complete generated output for different stacks:

  • React + Next.js — skills, agents tailored to Next.js + Prisma + Tailwind + Jest
  • Go API Server — skills, agents tailored to Go + Gin + Ent + golangci-lint
  • Python + FastAPI — skills, agents tailored to FastAPI + SQLAlchemy + pytest + Ruff

Customization

After generation, everything is yours to edit:

  • Add skills — create SKILL.md in .claude/skills/<name>/
  • Tune hooks — edit .claude/settings.json
  • Re-run — run /project-architect again to regenerate (it merges with existing config, never overwrites)

Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

The short version: it's all markdown. Adding support for a new stack means adding detection entries to detection-guide.md and examples to generation-guide.md.

License

MIT

About

Scan any codebase. Generate tailored Claude Code configuration — automatically.

Resources

License

Code of conduct

Contributing

Stars

Watchers

Forks

Releases

No releases published

Sponsor this project

 

Packages

 
 
 

Contributors

Languages