Turn Claude Code into a full development environment

Quick Start • Workflows • What's Inside • Installation • Architecture • Customization • 한국어

Claude Forge is an open-source development environment for Claude Code that provides 11 specialized agents, 40 slash commands, 15 skill workflows, and 15 automation hooks. Often described as "oh-my-zsh for Claude Code", it transforms Claude Code from a basic CLI into a full-featured development environment. One install gives you agents, commands, skills, hooks, and 9 rule files -- all pre-wired and ready to go.

Think of it as oh-my-zsh for Claude Code: the same way oh-my-zsh enhances your terminal, Claude Forge supercharges your AI coding assistant.

Claude Forge is available on the Anthropic Official Plugin Marketplace and can be installed directly from Claude Code:

# Option A: Install from Official Marketplace (after approval)
/plugin install claude-forge@claude-plugins-official

# Option B: Install from Claude Forge Marketplace
/plugin marketplace add sangrokjung/claude-forge
/plugin install claude-forge@claude-forge

# Option C: Install directly from GitHub
claude plugin install github:sangrokjung/claude-forge

To update:

/plugin marketplace update claude-forge

Note: Claude Forge has been submitted to the Anthropic Official Plugin Directory and is pending review. In the meantime, use Option B or C above.

For development or customization, clone the repository:

# 1. Clone
git clone --recurse-submodules https://github.com/sangrokjung/claude-forge.git
cd claude-forge

# 2. Install (creates symlinks to ~/.claude)
./install.sh

# 3. Launch Claude Code
claude

install.sh symlinks everything to ~/.claude/, so git pull updates instantly.

If you find Claude Forge useful, please consider giving it a star -- it helps others discover this project.

If you're new to development or Claude Code, start with these:

Or just type /auto login page and let Claude Forge handle the entire plan-to-PR pipeline for you.

Real-world workflows that chain commands, agents, and skills together.

Build new features with a plan-first, test-first approach:

/plan → /tdd → /code-review → /handoff-verify → /commit-push-pr → /sync

graph LR
    P["/plan<br><small>Design & risk analysis</small>"] --> T["/tdd<br><small>Tests first, then code</small>"]
    T --> CR["/code-review<br><small>Quality & security check</small>"]
    CR --> HV["/handoff-verify<br><small>Fresh-context validation</small>"]
    HV --> CPR["/commit-push-pr<br><small>Commit, push, PR & merge</small>"]
    CPR --> S["/sync<br><small>Sync project docs</small>"]

    style P fill:#533483,stroke:#fff,color:#fff
    style T fill:#0f3460,stroke:#fff,color:#fff
    style CR fill:#0f3460,stroke:#fff,color:#fff
    style HV fill:#e94560,stroke:#fff,color:#fff
    style CPR fill:#1a1a2e,stroke:#fff,color:#fff
    style S fill:#16213e,stroke:#fff,color:#fff

Fast turnaround for bug fixes with automatic retry:

/explore → /tdd → /verify-loop → /quick-commit → /sync

Comprehensive security analysis combining CWE and STRIDE:

/security-review → /stride-analysis-patterns → /security-compliance

Parallel multi-agent execution for complex tasks:

/orchestrate → Agent Teams (parallel work) → /commit-push-pr

Most developers either use Claude Code with no customization or spend hours assembling individual configs. Claude Forge gives you a production-ready setup in 5 minutes.

git clone --recurse-submodules https://github.com/sangrokjung/claude-forge.git
cd claude-forge
./install.sh

The installer:

1. Checks dependencies (node, git, jq)
2. Initializes git submodules (CC CHIPS status bar)
3. Backs up existing ~/.claude/ if present
4. Creates symlinks for 7 directories + settings.json to ~/.claude/
5. Applies CC CHIPS custom overlay
6. Optionally installs MCP servers and external skills
7. Adds shell aliases (cc → claude, ccr → claude --resume)

Because it uses symlinks, git pull in the repo updates everything instantly.

# Run PowerShell as Administrator
.\install.ps1

Windows uses file copies instead of symlinks. Re-run install.ps1 after git pull to update.

Override settings without modifying tracked files:

cp setup/settings.local.template.json ~/.claude/settings.local.json
vim ~/.claude/settings.local.json

settings.local.json merges on top of settings.json automatically.

graph TB
    subgraph REPO["claude-forge (git repo)"]
        A["agents/ (11)"]
        C["commands/ (40)"]
        S["skills/ (15)"]
        H["hooks/ (15)"]
        R["rules/ (9)"]
        SC["scripts/"]
        CC["cc-chips/"]
        K["knowledge/"]
        REF["reference/"]
        SET["settings.json"]
    end

    INSTALL["./install.sh"]
    REPO --> INSTALL

    subgraph HOME["~/.claude/ (symlinked)"]
        HA["agents/"]
        HC["commands/"]
        HS["skills/"]
        HH["hooks/"]
        HR["rules/"]
        HSC["scripts/"]
        HCC["cc-chips/"]
        HSET["settings.json"]
    end

    INSTALL -->|symlink| HOME

    CLAUDE["claude (CLI)"]
    HOME --> CLAUDE

    style REPO fill:#1a1a2e,stroke:#e94560,color:#fff
    style HOME fill:#0f3460,stroke:#16213e,color:#fff
    style INSTALL fill:#e94560,stroke:#fff,color:#fff
    style CLAUDE fill:#533483,stroke:#fff,color:#fff

claude-forge/
  ├── agents/               Agent definitions (11 .md files)
  ├── cc-chips/             Status bar submodule
  ├── cc-chips-custom/      Custom status bar overlay
  ├── commands/             Slash commands (32 .md + 8 SKILL dirs)
  ├── docs/                 Screenshots, diagrams
  ├── hooks/                Event-driven shell scripts (15)
  ├── knowledge/            Knowledge base entries
  ├── reference/            Reference documentation
  ├── rules/                Auto-loaded rule files (9)
  ├── scripts/              Utility scripts
  ├── setup/                Installation guides + templates
  ├── skills/               Multi-step skill workflows (15)
  ├── install.sh            macOS/Linux installer (symlinks)
  ├── install.ps1           Windows installer (copies)
  ├── mcp-servers.json      MCP server configurations
  ├── settings.json         Claude Code settings
  ├── CONTRIBUTING.md       Contribution guide
  ├── SECURITY.md           Security policy
  └── LICENSE               MIT License

The Agent Router is a forced delegation system that ensures specialized agents handle their domain tasks instead of Claude doing everything directly. Without this, Claude defaults to handling all tasks itself -- even when a specialized agent would produce better results.

graph LR
    U["User Message"] --> SP["using-superpowers<br><small>1% rule: MUST check skills</small>"]
    SP --> AR["agent-router<br><small>Domain match check</small>"]
    AR -->|"Match found"| A["Agent Tool<br><small>Spawn specialist</small>"]
    AR -->|"No match"| D["Direct Response"]
    A --> R["Specialist Result"]

    style U fill:#1a1a2e,stroke:#fff,color:#fff
    style SP fill:#e94560,stroke:#fff,color:#fff
    style AR fill:#533483,stroke:#fff,color:#fff
    style A fill:#0f3460,stroke:#fff,color:#fff
    style D fill:#16213e,stroke:#fff,color:#fff
    style R fill:#0f3460,stroke:#fff,color:#fff

Team sub-agents (ad-compass, ad-scout-google, ad-scout-meta, action-architect, folder-hunter, mail-scout) are managed by their parent team agents and are not directly routed.

Both using-superpowers and agent-router include <SUBAGENT-STOP> guards to prevent infinite recursion when an agent is already running inside a subagent context.

Edit commands/agent-router.md to add your own agents to the routing table:

| my custom keyword | my-custom-agent |

The router will automatically delegate matching requests to your agent.

Each agent has a color in the UI for quick visual identification: