`). - **Persistent memory layer** at `.claude/memory/` — facts about the project, user preferences, corrections, and references survive across sessions. Inherit memories from another workspace via `/memory-import`. Use a fully-built application as a reference via `/reference-app`. End-of-session summaries via `/handoff` so tomorrow's session picks up cleanly. - **2 safety hooks** that block dangerous operations before they execute. - **45+ MCP servers** wired up via `.mcp.json` (with `${ENV}` placeholders — no secrets committed). - **Cross-platform Node.js installer** that asks one friendly question and sets everything up. What the installer can fetch on demand (open-source, gitignored, never published with your repo): - **~150 VoltAgent agents** in 10 domains. - **~185 wshobson agents** grouped by plugin. - **~38 gstack skills** (Garry Tan's stack: `/office-hours`, `/ship`, `/review`, `/qa`, `/canary`, `/design-review`, etc.). - **14 superpowers skills** — TDD, brainstorming, parallel subagents, code review. - **24 Three.js skills** — WebGL, WebGPU, R3F, drei, physics. - **8 GSAP skills** — core, timeline, ScrollTrigger, plugins, React, performance. - **23 impeccable commands** — design fluency: polish, audit, critique, typeset, animate. - **WebGPU + TSL skill**, **design-motion-principles**, **ui-ux-pro-max**. - **Claude Code plugins**: frontend-design, superpowers, ui-ux-pro-max, ruflo (10 plugins), claude-mem (opt-in). [![Slash commands](https://img.shields.io/badge/slash--commands-200%2B-blue?style=flat-square)](CHEATSHEET.md) [![Agents](https://img.shields.io/badge/agents-12_original_%2B_~370_optional-purple?style=flat-square)](docs/ARCHETYPES.md) [![Vendored](https://img.shields.io/badge/vendored--in--repo-yes-green?style=flat-square)](docs/LICENSES.md) [![MCP Servers](https://img.shields.io/badge/MCP_servers-45%2B-orange?style=flat-square)](.mcp.json) [![Plugins](https://img.shields.io/badge/native_plugins-13-red?style=flat-square)](.claude/plugins.json) [![Rules](https://img.shields.io/badge/rules-9-green?style=flat-square)](CATALOG.md) [![Stack](https://img.shields.io/badge/stack-agnostic-brightgreen?style=flat-square)](#design-philosophy) [![Platform](https://img.shields.io/badge/platform-win%20%7C%20mac%20%7C%20linux-lightgrey?style=flat-square)](#quick-start) [![License](https://img.shields.io/badge/license-MIT-yellow?style=flat-square)](LICENSE) --- ## Why This Exists Every time you start a Claude Code session, it starts from zero — no memory of your conventions, no safety guardrails, no reusable workflows. This repository fixes that by providing a shared configuration layer that gives every project: - **30 slash-command skills** for common workflows (code review, testing, security, deployment, and more) - **8 specialized agents** that Claude can delegate complex tasks to (architecture, TDD, security auditing, mentoring) - **9 auto-loading rule files** that enforce coding standards without you having to repeat yourself - **Safety hooks** that block dangerous operations before they execute - **30 MCP servers** for web fetching, browser automation, design (Figma), code editing (Serena), GitHub/GitLab, databases, Slack, Stripe, Kubernetes, Cloudflare, AI/ML, and more All of it is **stack-agnostic** — skills auto-detect your project's language, framework, and tools at runtime. --- ## Table of Contents - [Quick Start](#quick-start) - [What's Included](#whats-included) - [Skills (30)](#skills-30) - [Agents (8)](#agents-8) - [Rules (9)](#rules-9) - [MCP Servers (30)](#mcp-servers-30) - [Hooks (2)](#hooks-2) - [Customization](#customization) - [Adding Your Own](#adding-your-own) - [Upgrading](#upgrading) - [Using as a Git Submodule](#using-as-a-git-submodule) - [Design Philosophy](#design-philosophy) - [Troubleshooting](#troubleshooting) - [Contributing](#contributing) - [License](#license) --- ## Quick Start — one command ```bash cd your-project npx github:viknesh20-20/claude-code-tool-kit ``` That's it. The installer asks one question — *what are you building?* — and sets up the matching bundle (plugins, skills, agents, MCPs, secrets) inside the current directory. To install everything without picking: ```bash npx github:viknesh20-20/claude-code-tool-kit --all ``` **Windows PowerShell** works the same way: ```powershell cd your-project npx github:viknesh20-20/claude-code-tool-kit ``` That's the recommended path. Everything else below is alternatives for specific situations. --- ### Alternative: clone + run If you'd rather have the toolkit on disk (so you can update it yourself): ```bash git clone https://github.com/viknesh20-20/claude-code-tool-kit.git ~/claude-code-toolkit cd /path/to/your/project node ~/claude-code-toolkit/install.mjs ``` ### Alternative: file-only quick install No archetypes, no prompts, no plugins — just the core skills, agents, rules, hooks: ```bash npx github:viknesh20-20/claude-code-tool-kit --archetype files-only ``` ### Alternative: git submodule (for teams) ```bash cd your-project git submodule add https://github.com/viknesh20-20/claude-code-tool-kit.git claude-code-toolkit node claude-code-toolkit/install.mjs --target . git add . && git commit -m "chore: add claude-code-toolkit" ``` ### What the installer does 1. Detects your toolchain (Node, git, npx, Claude Code, uvx) and tells you what's missing with install links. 2. Asks for an **archetype** (web · saas · ai · all · custom · files-only) — friendly question with examples. 3. Copies skills, agents, rules, hooks, `.mcp.json`, `settings.json` into your project. 4. Adds plugin marketplaces and installs Claude Code plugins matching the archetype. 5. Clones agent collections (VoltAgent, wshobson, optionally contains-studio with consent) into gitignored paths. 6. Adds per-archetype MCP servers and prompts for any API keys you have (skip any to add later). 7. Initializes the persistent memory layer at `.claude/memory/`. 8. Generates a license register (`docs/LICENSES.md`) reflecting what was installed. Re-running the installer is safe — already-installed items are skipped. After setup, customize `CLAUDE.md` with your project's stack, then open [CHEATSHEET.md](CHEATSHEET.md) for the intent → command lookup. --- ## Archetypes The installer asks one question: *what are you building?* Six answers, six bundles. | Archetype | For | Key adds | |---|---|---| | **web** | $10K-tier sites, agencies, marketing | impeccable · ui-ux-pro-max · 21st-dev/magic · frontend-design · gsap-skills · motion-principles · figma + playwright MCP | | **saas** | $100K-tier B2B / SaaS | everything from `web` + superpowers + ruflo (core, swarm, testgen, security-audit, observability) + gstack + claude-context + DB/payments/comms/monitoring MCPs | | **ai** | Autonomous agents, RAG, LLM apps | full ruflo (core, swarm, autopilot, federation, agentdb, rag-memory, knowledge-graph) + superpowers + qdrant + chroma + claude-context + huggingface + perplexity + firecrawl | | **all** | Solo founders, "kitchen sink" | Union of `web` + `saas` + `ai` | | **custom** | Power users | Granular per-service prompts | | **files-only** | Skeptics, offline | Just the toolkit's skills/agents/rules/hooks | See [docs/ARCHETYPES.md](docs/ARCHETYPES.md) for what each one installs in detail. --- ## What this can build | Tier | Use the toolkit to ship | Provided by | |---|---|---| | **Premium websites ($10K+)** | Animated landing pages, design-forward marketing sites, portfolios | impeccable, ui-ux-pro-max, 21st-dev/magic, gsap-skills, motion-principles, frontend-design, figma + playwright MCP | | **Enterprise SaaS ($100K+)** | Multi-tenant dashboards, B2B products, billing flows, complex backends | superpowers, ruflo, gstack, claude-mem (opt-in), claude-context (RAG over codebase), full DB/payments/monitoring stack, 190+ specialized agents | | **AI agents / RAG / LLM apps** | Autonomous swarms, vector-search apps, research bots, agentic pipelines | full ruflo plugin set, qdrant + chroma MCPs, claude-context, huggingface, perplexity, tavily, exa, browserbase, e2b, replicate, elevenlabs, sequential-thinking, 150+ AI/data agents | For the comprehensive register of every tool the toolkit can drive — design references, brainstorming whiteboards (15 ranked free-first), Three.js / WebGPU / WebGL stack, testing, DevOps, AI MCPs — see [docs/TOOLBOX.md](docs/TOOLBOX.md). --- ## What's Included ``` claude-code-toolkit/ ├── CLAUDE.md # Universal project instructions (<100 lines) ├── .mcp.json # 30 MCP servers (Figma, Serena, Playwright, Stripe, K8s, etc.) ├── LICENSE # MIT ├── CONTRIBUTING.md # How to contribute ├── CATALOG.md # Full index of everything ├── install.sh # Interactive installer with secret config ├── setup.sh # Quick setup (no prompts) ├── update.sh # Pull latest config │ └── .claude/ ├── settings.json # Tool allow/deny permissions │ ├── agents/ # 8 specialized agents │ ├── architect.md # System design & ADRs │ ├── code-reviewer.md # Autonomous code review │ ├── devops.md # CI/CD & infrastructure │ ├── documentation.md # Docs generation │ ├── mentor.md # Teaching & explaining │ ├── performance.md # Optimization specialist │ ├── security-auditor.md # Threat modeling & OWASP │ └── tdd.md # Test-driven development │ ├── rules/ # 9 auto-loading rule files │ ├── api-design.md # REST conventions & status codes │ ├── code-quality.md # Naming, complexity, functions │ ├── database.md # Migrations, queries, indexes │ ├── documentation.md # Doc standards & maintenance │ ├── error-handling.md # Errors, logging, retry │ ├── git-workflow.md # Commits, branching, PRs │ ├── performance.md # Algorithms, caching, I/O │ ├── security.md # Secrets, validation, auth │ └── testing.md # Coverage, structure, mocking │ ├── hooks/ # 2 lifecycle hooks │ ├── pre-tool-use.sh # Safety gate (blocks dangerous ops) │ └── post-tool-use.sh # Post-execution (extensible) │ └── skills/ # 30 slash-command skills ├── api-design/ ├── db-migration/ ├── architecture-review/ ├── debug/ ├── changelog/ ├── dependency-audit/ ├── ci-pipeline/ ├── deploy-checklist/ ├── code-coverage/ ├── docker-setup/ ├── convert-code/ ├── env-setup/ ├── create-component/ ├── error-monitor/ ├── estimate/ ├── explain-code/ ├── fix-issue/ ├── generate-docs/ ├── git-cleanup/ ├── grill-me/ ├── migrate-framework/ ├── onboard/ ├── optimize/ ├── pr-summary/ ├── refactor/ ├── review-pr/ ├── security-scan/ ├── tdd/ └── write-tests/ └── write-skill/ ``` --- ## Skills (30) All skills are **stack-agnostic** — they auto-detect your project's language and framework at runtime. ### Code Quality | Command | What It Does | |---------|-------------| | `/review-pr` | Full PR review — correctness, security, performance, tests, conventions. Outputs severity-rated findings + GO/NO-GO verdict. | | `/refactor` | Detects code smells (long methods, deep nesting, duplication), plans safe refactoring, executes with test verification. | | `/explain-code` | Multi-level explanation — one-sentence summary, high-level overview, detailed walkthrough, ASCII data flow diagram. | ### Testing | Command | What It Does | |---------|-------------| | `/write-tests` | Generates tests auto-detecting Jest/pytest/Go/Rust/etc. Covers happy path, edge cases, errors, async. Ready-to-commit output. | | `/code-coverage` | Analyzes coverage gaps, identifies untested critical paths, generates tests for the most important uncovered areas. | | `/tdd` | Strict red-green-refactor. Writes failing test first, minimal implementation, then refactors. Never writes code before tests. | ### Security | Command | What It Does | |---------|-------------| | `/security-scan` | Scans for injection flaws, hardcoded secrets, auth issues, XSS, CSRF, SSRF. Outputs severity-rated findings + security score (0-10). | | `/dependency-audit` | Audits packages for vulnerabilities, outdated versions, unused deps, and license compliance across any ecosystem. | ### Performance | Command | What It Does | |---------|-------------| | `/optimize` | Finds algorithmic bottlenecks, N+1 queries, memory leaks, missed parallelization. Before/after code with impact ratings. | ### Architecture & Design | Command | What It Does | |---------|-------------| | `/architecture-review` | Maps dependencies, identifies circular deps and layer violations, assesses scalability. ASCII architecture diagram. | | `/api-design` | Designs REST/GraphQL APIs — endpoints, schemas, error codes, pagination. Outputs OpenAPI spec. | | `/create-component` | Scaffolds modules following existing project patterns — main file, tests, types, index. | ### DevOps & Infrastructure | Command | What It Does | |---------|-------------| | `/deploy-checklist` | Pre-deployment verification — tests, security, migrations, env vars, rollback plan. GO/NO-GO verdict. | | `/db-migration` | Generates up/down migrations with safety analysis, index recommendations, and rollback instructions. | | `/git-cleanup` | Identifies merged branches, stale refs, large files in history. Cleans up with confirmation. | | `/ci-pipeline` | Generates CI config (GitHub Actions, GitLab CI) with lint, test, build, deploy stages + caching. | | `/docker-setup` | Creates optimized multi-stage Dockerfile, docker-compose.yml, and .dockerignore. | | `/env-setup` | Generates .env.example, checks prerequisites, documents setup steps, verifies environment. | ### Project Management | Command | What It Does | |---------|-------------| | `/fix-issue` | Fetches GitHub issue, implements fix, creates branch and PR. References issue in commits. | | `/estimate` | Breaks tasks into subtasks with T-shirt sizing (S/M/L/XL), risk assessment, and dependency mapping. | | `/changelog` | Generates changelog from git history in Keep a Changelog format. Auto-categorizes commits. | | `/pr-summary` | Generates PR title and description from branch diff — summary, changes, test plan, breaking changes. | ### Documentation & Utilities | Command | What It Does | |---------|-------------| | `/generate-docs` | Generates README, API docs, or inline documentation matching existing project style. | | `/debug` | Structured debugging — gather info, read error, isolate problem, common patterns, fix, prevent recurrence. | | `/convert-code` | Converts code between languages preserving logic and using idiomatic patterns in the target language. | | `/onboard` | Scans entire codebase and generates an onboarding guide — stack, architecture, entry points, conventions, gotchas. | | `/migrate-framework` | Framework migration assistant — maps APIs, creates migration plan, provides file-by-file order. | | `/error-monitor` | Sets up error tracking (Sentry/Bugsnag) with error boundaries, middleware, and custom error classes. | ### Meta | Command | What It Does | |---------|-------------| | `/write-skill` | Meta-skill that creates new skills — generates SKILL.md with proper frontmatter and structure. | | `/grill-me` | Interrogates you about your plan until every edge case, failure mode, and assumption is resolved. Questions only — no answers. | --- ## Agents (8) Agents are specialized personas Claude can delegate to. They provide deeper expertise than skills by maintaining a focused role throughout the task. | Agent | Specialty | Delegates When | |-------|-----------|---------------| | **code-reviewer** | Autonomous code review | Reviewing any code changes for quality, security, performance | | **architect** | System design, ADRs, Mermaid diagrams | Evaluating architecture, planning features, making design decisions | | **tdd** | Test-driven development | Building features with strict red-green-refactor discipline | | **security-auditor** | Threat modeling, OWASP Top 10 | Assessing security, pre-deployment audits, compliance reviews | | **devops** | CI/CD, Docker, deployment | Setting up pipelines, containerization, infrastructure | | **performance** | Profiling, bottleneck analysis | Finding and fixing performance issues | | **documentation** | Docs matching project style | Writing READMEs, API docs, onboarding guides | | **mentor** | Teaching with analogies and examples | Explaining concepts, learning new technologies, understanding code | Agents live in `.claude/agents/` and can be invoked with the `/agents` command or auto-delegated by Claude. --- ## Rules (9) Rules are modular instruction files that **auto-load into every Claude Code session**. They enforce coding standards without you having to repeat yourself. | Rule | What It Enforces | |------|-----------------| | **code-quality** | Naming conventions, function length (<40 lines), nesting depth (<3), no dead code, DRY | | **security** | No hardcoded secrets, input validation, parameterized queries, auth checks, dependency updates | | **testing** | >80% coverage on business logic, arrange-act-assert, descriptive test names, deterministic tests | | **git-workflow** | Conventional commits, atomic commits, branch naming, PR size (<400 lines), no force push to main | | **error-handling** | Explicit handling, structured logging, meaningful messages, retry with backoff, circuit breakers | | **performance** | No O(n^2) in production, no N+1 queries, connection pooling, caching strategy, lazy loading | | **documentation** | Document "why" not "what", JSDoc/docstrings for public APIs, keep docs updated | | **api-design** | Plural nouns, proper HTTP status codes, consistent error format, pagination, versioning | | **database** | Migration files for all changes, index foreign keys, parameterized queries, connection pooling | Rules live in `.claude/rules/` and require no manual invocation — they are always active. --- ## MCP Servers (30) 30 pre-configured MCP servers across 12 categories. Servers using `${ENV_VAR}` references are safe to keep in `.mcp.json` — they resolve at runtime. Set vars in your shell profile or `.env.local`. ### Core (zero config) | Server | Purpose | |--------|---------| | **fetch** | Make HTTP requests from Claude | | **filesystem** | Scoped file read/write access | | **memory** | Persistent knowledge across sessions | | **sequential-thinking** | Structured multi-step reasoning for complex problems | | **context7** | Up-to-date library documentation (replaces stale training data) | ### Code & Semantic Editing | Server | Purpose | Setup | |--------|---------|-------| | **serena** | Semantic code retrieval, symbol-level editing, multi-language support | Requires `uvx` (Python) | ### Frontend & Design | Server | Purpose | Env Vars | |--------|---------|----------| | **figma** | Design-to-code — pull variables, components, layout from Figma | `FIGMA_ACCESS_TOKEN` | ### Browser Automation & Web Scraping | Server | Purpose | Env Vars | |--------|---------|----------| | **playwright** | Full browser automation — testing, screenshots, form filling | None | | **puppeteer** | Headless Chrome control | None | | **firecrawl** | Web scraping, crawling, structured data extraction | `FIRECRAWL_API_KEY` | ### Search & Research | Server | Purpose | Env Vars | |--------|---------|----------| | **brave-search** | Web search | `BRAVE_API_KEY` | | **perplexity** | AI-powered research with citations (Sonar models) | `PERPLEXITY_API_KEY` | ### Version Control & Code Platforms | Server | Purpose | Env Vars | |--------|---------|----------| | **github** | PR reviews, issues, repo operations | `GITHUB_PERSONAL_ACCESS_TOKEN` | | **gitlab** | Merge requests, issues, CI pipelines | `GITLAB_PERSONAL_ACCESS_TOKEN`, `GITLAB_API_URL` | ### Databases | Server | Purpose | Env Vars | |--------|---------|----------| | **postgres** | PostgreSQL operations | `POSTGRES_CONNECTION_STRING` | | **sqlite** | SQLite operations | None | | **redis** | Redis cache & data store | `REDIS_URL` | | **mongodb** | MongoDB queries & Atlas cluster management | `MONGODB_URI` | | **supabase** | Supabase Postgres, auth, edge functions | `SUPABASE_URL`, `SUPABASE_API_KEY` | ### Infrastructure & Deployment | Server | Purpose | Env Vars | |--------|---------|----------| | **kubernetes** | K8s cluster management, pods, Helm charts | Uses local kubeconfig | | **cloudflare** | DNS, Workers, R2, Zero Trust (2,500+ API endpoints) | `CLOUDFLARE_API_TOKEN`, `CLOUDFLARE_ACCOUNT_ID` | ### Payments & Commerce | Server | Purpose | Env Vars | |--------|---------|----------| | **stripe** | Payments, subscriptions, customers | `STRIPE_SECRET_KEY` | ### Communication | Server | Purpose | Env Vars | |--------|---------|----------| | **slack** | Team messaging & channel operations | `SLACK_BOT_TOKEN`, `SLACK_TEAM_ID` | | **twilio** | SMS, voice, communications | `TWILIO_ACCOUNT_SID`, `TWILIO_AUTH_TOKEN` | ### Project Management | Server | Purpose | Env Vars | |--------|---------|----------| | **notion** | Knowledge base & documentation | `NOTION_TOKEN` | | **linear** | Issue tracking & sprint management | `LINEAR_API_KEY` | ### Monitoring & Error Tracking | Server | Purpose | Env Vars | |--------|---------|----------| | **sentry** | Error tracking & crash monitoring | `SENTRY_ACCESS_TOKEN` | ### AI & ML | Server | Purpose | Env Vars | |--------|---------|----------| | **huggingface** | Models, datasets, Spaces, papers from HF Hub | `HF_TOKEN` | --- ## Hooks (2) ### pre-tool-use.sh — Safety Gate Blocks dangerous operations before they execute: | Blocked | Why | |---------|-----| | `rm -rf /` (system paths) | Prevents accidental system destruction | | `curl ... \| bash` | Prevents remote code execution | | `git push --force` | Requires explicit approval | | `sudo ...` | Prevents privilege escalation | | Write to `.env*` | Prevents secret file corruption | | Write to lock files | Lock files managed by package managers | | Write to `.git/` | Git internals managed by git commands | ### post-tool-use.sh — Extensible Currently a passthrough. Uncomment sections to enable auto-linting, audit logging, or time tracking. --- ## Customization ### Shared vs Personal | File | Scope | Committed? | |------|-------|-----------| | `CLAUDE.md` | Team | Yes — project-specific instructions | | `.claude/rules/*` | Team | Yes — coding standards | | `.claude/agents/*` | Team | Yes — specialized agents | | `.claude/skills/*` | Team | Yes — workflow skills | | `CLAUDE.local.md` | You only | No (gitignored) — personal notes | | `.claude/settings.local.json` | You only | No (gitignored) — permission overrides | ### Override a Skill for One Project ```bash # Copy the shared skill to project level cp -r claude-code-toolkit/.claude/skills/review-pr .claude/skills/review-pr # Edit .claude/skills/review-pr/SKILL.md with your customizations # Project-level skills take precedence ``` ### Add Project-Specific Rules Create `.claude/rules/my-project-rules.md` in your project — it will auto-load alongside the shared rules. --- ## Adding Your Own See [CONTRIBUTING.md](CONTRIBUTING.md) for detailed templates and guidelines. Quick overview: ### New Skill ```bash mkdir -p .claude/skills/my-skill # Create .claude/skills/my-skill/SKILL.md with frontmatter # Or use: /write-skill to generate it automatically ``` ### New Agent ```bash # Create .claude/agents/my-agent.md with name, description, tools frontmatter ``` ### New Rule ```bash # Create .claude/rules/my-rule.md — plain markdown, no frontmatter # Auto-loads into every session ``` --- ## Upgrading ### Standalone clone ```bash cd ~/claude-code-toolkit && git pull origin main bash ~/claude-code-toolkit/setup.sh /path/to/your/project ``` ### Git submodule ```bash bash claude-code-toolkit/update.sh git add claude-code-toolkit && git commit -m "chore: update claude-code-toolkit" ``` New skills, agents, and rules are added automatically. Your project-level customizations are never overwritten. --- ## Using as a Git Submodule ```bash # Add to project git submodule add https://github.com/viknesh20-20/claude-code-tool-kit.git claude-code-toolkit bash claude-code-toolkit/setup.sh . git add .gitmodules claude-code-toolkit .claude/ CLAUDE.md .mcp.json .gitignore git commit -m "chore: add claude-code-toolkit submodule" # Clone a project that uses it git clone --recurse-submodules https://github.com/YOUR_USERNAME/your-project.git bash claude-code-toolkit/setup.sh . # Forgot --recurse-submodules? git submodule update --init --recursive bash claude-code-toolkit/setup.sh . ``` --- ## Design Philosophy **Stack-agnostic by default.** Every skill auto-detects the project's language, framework, and tools using shell injection. The same `/write-tests` generates Jest tests for React and pytest tests for FastAPI. **Detect, don't prescribe.** CLAUDE.md defines universal principles. Rules define specific standards. Your project adds the stack-specific details. Upgrading the shared config never conflicts with your project. **Layered architecture.** Rules (always-active standards) + Skills (on-demand workflows) + Agents (specialized personas) + Hooks (safety gates). Each layer has a clear purpose and doesn't overlap. **Safety first.** The pre-tool-use hook blocks destructive operations. The settings.json deny-list prevents writes to sensitive files. You get guardrails without manual review of every tool call. **Composable and upgradable.** Override any single skill, agent, or rule at the project level without affecting others. `git pull` to get new capabilities. Setup is idempotent — safe to re-run. --- ## Troubleshooting | Problem | Fix | |---------|-----| | Skills not in `/` menu | Re-run `setup.sh` — skills may not be in `.claude/skills/` | | Agents not available | Check `.claude/agents/` has `.md` files | | Rules not loading | Verify `.claude/rules/*.md` exists and has content | | MCP server connection fails | Check `npx` is in PATH (install Node.js 18+) | | GitHub MCP 401 error | Set `GITHUB_PERSONAL_ACCESS_TOKEN` env var | | MCP server "env var not set" | Re-run `install.sh` or set vars manually in `.env.local` / shell profile | | Hook not triggering | Run `chmod +x .claude/hooks/*.sh` | | setup.sh fails on Windows | Use Git Bash (installed with Git for Windows) | | Submodule files missing | Run `git submodule update --init --recursive` | | Settings not taking effect | Check `.claude/settings.json` format with `python3 -m json.tool` | --- ## Contributing See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines on adding skills, agents, rules, and hooks. --- ## License [MIT](LICENSE) — this toolkit (the 12 agents, 30 skills, 9 rules, 2 hooks, installer, docs) is MIT-licensed and is yours to push, fork, modify, and ship. **Third-party content is never committed to this repo.** The installer fetches it on demand into gitignored paths. Each fetched folder gets a `NOTICE.md` with attribution. You can publish your repo without redistributing third-party code. The full per-source license register is at [docs/LICENSES.md](docs/LICENSES.md). One source (`contains-studio`) has no LICENSE upstream — the installer requires explicit consent before fetching it and defaults to "no."