clawdbot-config

📁 tumf/skills 📅 5 days ago

总安装量

周安装量

#51266

全站排名

安装命令

npx skills add https://github.com/tumf/skills --skill clawdbot-config

Agent 安装分布

amp 4

gemini-cli 4

claude-code 4

github-copilot 4

codex 4

kimi-cli 4

Skill 文档

Clawdbot Configuration Expert

Expert guidance for Clawdbot’s configuration system, skills management, and agent setup.

When to Use This Skill

Use this skill when you need to:

Configure Clawdbot: Set up ~/.clawdbot/clawdbot.json with agents, channels, gateway, or tools
Create Skills: Build AgentSkills-compatible SKILL.md with proper YAML frontmatter and metadata
Manage Skills: Configure skill loading, gating, environment injection, or installation
Troubleshoot: Diagnose configuration issues, skill loading problems, or agent behavior
Understand System: Learn about workspace structure, bootstrap files, or session management
Deep Dive: Access complete documentation archive (370+ files) for channels, automation, webhooks, cron jobs, gateway protocols, platform-specific guides, and advanced features

Quick Decision Tree

Need to configure Clawdbot?
ââ Setting up for first time? â See "Quick Start Guide"
ââ Managing skills? â See "Skills Management"
ââ Configuring channels? â See "Channel Configuration"
ââ Setting up remote access? â See "Remote Gateway Setup"
ââ Creating a skill? â See "Creating Skills"

Need examples?
ââ See references/config-examples.md for common patterns

Need deep documentation?
ââ Channel-specific setup? â See references/repomix-output.xml (channels/)
ââ Automation (cron, webhooks)? â See references/repomix-output.xml (automation/)
ââ Gateway protocols? â See references/repomix-output.xml (gateway/)
ââ Platform guides? â See references/repomix-output.xml (platforms/)
ââ CLI reference? â See references/repomix-output.xml (cli/)

Quick Start Guide

Initial Setup

Create minimal config (~/.clawdbot/clawdbot.json):

{
  "agents": {
    "defaults": {
      "workspace": "~/Projects/my-agent-workspace"
    }
  },
  "channels": {
    "whatsapp": {
      "allowFrom": ["+1555555YOUR_NUMBER"]
    }
  }
}

Initialize workspace:

clawdbot setup

This creates bootstrap files:

AGENTS.md – Operating instructions
SOUL.md – Persona and boundaries
TOOLS.md – Tool usage notes
IDENTITY.md – Agent identity
USER.md – User profile

Pair WhatsApp:

clawdbot channels login

Start gateway:

clawdbot gateway

Configuration File Location

Primary: ~/.clawdbot/clawdbot.json
Override: CLAWDBOT_CONFIG_PATH environment variable
Format: JSONC (JSON with Comments)

Core Configuration Areas

1. Agents Configuration

Controls the embedded agent runtime.

Minimal:

{
  "agents": {
    "defaults": {
      "workspace": "~/Projects/my-workspace",
      "model": "anthropic/claude-sonnet-4"
    }
  }
}

With sandbox:

{
  "agents": {
    "defaults": {
      "workspace": "~/Projects/my-workspace",
      "sandbox": {
        "enabled": true,
        "docker": {
          "image": "clawdbot/sandbox:latest",
          "setupCommand": "apt-get update && apt-get install -y python3-pip",
          "env": {
            "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}"
          }
        }
      }
    }
  }
}

Key fields:

workspace (required): Agent’s working directory
model: Primary model (format: provider/model)
sandbox: Docker isolation configuration
blockStreamingDefault: Streaming behavior ("on" | "off")

See references/config-schema.md for complete options.

2. Skills Configuration

Manage skill loading, filtering, and configuration.

Basic:

{
  "skills": {
    "allowBundled": ["gemini", "browser"],
    "entries": {
      "gemini": {
        "enabled": true,
        "apiKey": "${GEMINI_API_KEY}"
      }
    }
  }
}

Advanced:

{
  "skills": {
    "load": {
      "extraDirs": ["~/Projects/custom-skills/skills"],
      "watch": true,
      "watchDebounceMs": 250
    },
    "install": {
      "preferBrew": true,
      "nodeManager": "pnpm"
    },
    "entries": {
      "nano-banana-pro": {
        "enabled": true,
        "apiKey": "${GEMINI_API_KEY}",
        "config": {
          "model": "gemini-3-pro-image"
        }
      }
    }
  }
}

Skill precedence (highest to lowest):

<workspace>/skills
~/.clawdbot/skills
skills.load.extraDirs
Bundled skills

See references/skills-system.md for complete skills documentation.

3. Channels Configuration

Configure WhatsApp, Telegram, Discord, iMessage, and other channels.

WhatsApp:

{
  "channels": {
    "whatsapp": {
      "enabled": true,
      "allowFrom": ["+15555550123"],
      "groups": {
        "*": { "requireMention": true }
      }
    }
  }
}

Multi-channel:

{
  "channels": {
    "whatsapp": {
      "allowFrom": ["+15555550123"]
    },
    "telegram": {
      "enabled": true,
      "botToken": "${TELEGRAM_BOT_TOKEN}",
      "allowFrom": ["alice", "bob"]
    },
    "discord": {
      "enabled": true,
      "botToken": "${DISCORD_BOT_TOKEN}"
    }
  }
}

4. Gateway Configuration

WebSocket server and remote access.

Local (default):

{
  "gateway": {
    "port": 18789,
    "host": "127.0.0.1"
  }
}

Remote (Tailscale):

{
  "gateway": {
    "port": 18789,
    "bind": "tailnet",
    "auth": {
      "token": "${GATEWAY_AUTH_TOKEN}"
    }
  }
}

Start with:

clawdbot gateway --bind tailnet --token $GATEWAY_AUTH_TOKEN

5. Tools Configuration

Control tool behavior and permissions.

{
  "tools": {
    "exec": {
      "security": "ask",  // "allow" | "ask" | "deny"
      "applyPatch": true
    },
    "browser": {
      "enabled": true,
      "profilePath": "~/.clawdbot/browser-profile"
    }
  }
}

Skills Management

Creating a Skill

Skill structure:

my-skill/
âââ SKILL.md          (required: frontmatter + instructions)
âââ scripts/          (optional: executable code)
âââ references/       (optional: detailed docs)
âââ assets/           (optional: templates, files)

SKILL.md format:

---
name: my-skill
description: What the skill does and when to use it. Be specific about triggers.
metadata: {"clawdbot":{"requires":{"bins":["python3"],"env":["MY_API_KEY"]},"primaryEnv":"MY_API_KEY"}}
---

# My Skill

## Quick Start

Instructions for using this skill...

## Resources

Scripts: python3 "$SKILL_ROOT/scripts/process.py"

Metadata for gating:

{
  "clawdbot": {
    "emoji": "ð§",
    "requires": {
      "bins": ["python3", "pip"],
      "anyBins": ["npm", "pnpm"],
      "env": ["MY_API_KEY"],
      "config": ["browser.enabled"]
    },
    "primaryEnv": "MY_API_KEY",
    "os": ["darwin", "linux"],
    "install": [
      {
        "id": "brew",
        "kind": "brew",
        "formula": "my-tool",
        "bins": ["my-tool"],
        "label": "Install My Tool (brew)"
      }
    ]
  }
}

Key principles:

description is the primary trigger mechanism
Metadata enables load-time filtering
Keep SKILL.md concise; use references/ for details
Test scripts by actually running them

Skill Locations

Per-agent (highest precedence):

<workspace>/skills/my-skill/SKILL.md

Shared (all agents):

~/.clawdbot/skills/my-skill/SKILL.md

Custom folders:

{
  "skills": {
    "load": {
      "extraDirs": ["~/Projects/custom-skills/skills"]
    }
  }
}

Configuring Skills

Enable/disable and provide API keys:

{
  "skills": {
    "entries": {
      "my-skill": {
        "enabled": true,
        "apiKey": "${MY_API_KEY}",
        "env": {
          "MY_API_KEY": "${MY_API_KEY}",
          "MY_ENDPOINT": "https://api.example.com"
        }
      },
      "disabled-skill": {
        "enabled": false
      }
    }
  }
}

Bundled skills allowlist:

{
  "skills": {
    "allowBundled": ["gemini", "browser", "peekaboo"]
  }
}

Skills in Sandbox

When agent is sandboxed, skills run inside Docker.

Requirements:

Binaries checked on host at load time
Binaries must exist in container at runtime
Install via setupCommand:

{
  "agents": {
    "defaults": {
      "sandbox": {
        "docker": {
          "setupCommand": "apt-get update && apt-get install -y python3-pip && pip3 install ruff",
          "env": {
            "GEMINI_API_KEY": "${GEMINI_API_KEY}"
          }
        }
      }
    }
  }
}

Important: Global skills.entries.<skill>.env applies to host runs only. Use sandbox.docker.env for container.

Channel Configuration

Basic:

{
  "channels": {
    "whatsapp": {
      "enabled": true,
      "allowFrom": ["+15555550123"]
    }
  }
}

With groups:

{
  "channels": {
    "whatsapp": {
      "allowFrom": ["+15555550123", "+15555550124"],
      "groups": {
        "*": {
          "requireMention": true
        },
        // Specific group (use group JID)
        "120363012345678@g.us": {
          "requireMention": false
        }
      }
    }
  },
  "messages": {
    "groupChat": {
      "mentionPatterns": ["@clawd", "hey clawd"]
    }
  }
}

Pairing:

clawdbot channels login

{
  "channels": {
    "telegram": {
      "enabled": true,
      "botToken": "${TELEGRAM_BOT_TOKEN}",
      "allowFrom": ["username1", "username2"]
    }
  }
}

Get bot token: Create bot with @BotFather

Discord

{
  "channels": {
    "discord": {
      "enabled": true,
      "botToken": "${DISCORD_BOT_TOKEN}",
      "allowFrom": ["user_id_1", "user_id_2"]
    }
  }
}

Setup: Create application at Discord Developer Portal

iMessage (macOS)

{
  "channels": {
    "imessage": {
      "enabled": true,
      "cliPath": "/opt/homebrew/bin/imsg",
      "dbPath": "~/Library/Messages/chat.db"
    }
  }
}

Install: brew install steipete/tap/imsg

Remote Gateway Setup

Tailscale

Install Tailscale: https://tailscale.com/download
Generate token:

clawdbot doctor --generate-gateway-token

Update config:

{
  "gateway": {
    "bind": "tailnet",
    "auth": {
      "token": "YOUR_TOKEN_HERE"
    }
  }
}

Start gateway:

clawdbot gateway --bind tailnet --token $GATEWAY_AUTH_TOKEN

Connect from remote:

export CLAWDBOT_GATEWAY_URL="ws://gateway-hostname.tailnet.ts.net:18789"
export CLAWDBOT_GATEWAY_TOKEN="your-token"

clawdbot tui

SSH Tunnel

On remote machine (run gateway):

clawdbot gateway

On local machine (create tunnel):

ssh -L 18789:localhost:18789 user@remote-host

# In another terminal
clawdbot tui

Multi-Agent Routing

Route different users/channels to specialized agents:

{
  "agents": {
    "defaults": {
      "workspace": "~/Clawdbot/main",
      "model": "anthropic/claude-sonnet-4"
    },
    "list": [
      {
        "id": "reviewer",
        "workspace": "~/Clawdbot/reviewer",
        "routes": [
          { "channel": "telegram", "peer": "alice" }
        ],
        "tools": {
          "write": false,
          "exec": { "security": "deny" }
        }
      },
      {
        "id": "analyst",
        "workspace": "~/Clawdbot/analyst",
        "routes": [
          { "channel": "discord", "peer": "123456789012345678" }
        ]
      }
    ]
  }
}

Routing rules:

Match by channel + peer
Each agent has isolated workspace and sessions
Agent-specific tool policies

Troubleshooting

Check Status

# Gateway reachability and quick summary
clawdbot status

# Show effective configuration
clawdbot configure --show

# Validate config syntax
clawdbot configure --validate

# Health checks and quick fixes
clawdbot doctor

# Deep health probes
clawdbot health --verbose

View Logs

# Tail gateway logs
clawdbot logs --follow

# View specific number of lines
clawdbot logs --lines 100

Common Issues

Gateway won’t start:

Check if port is already in use: lsof -i :18789
Verify config syntax: clawdbot configure --validate
Check logs: clawdbot logs

Skills not loading:

Verify skill location: clawdbot skills list
Check metadata requirements (bins, env, config)
For sandboxed agents, ensure binaries exist in container

WhatsApp not responding:

Check allowFrom list includes your number
Verify QR code pairing: clawdbot channels login
Check channel status: clawdbot status

Remote connection fails:

Verify auth token is set correctly
Check gateway bind mode: clawdbot status
Test connectivity: curl http://gateway-host:18789/__health

References

Detailed Documentation

Config Schema: Complete configuration reference with all options
Skills System: In-depth skills documentation (metadata, gating, installation)
Config Examples: Practical examples for common use cases
Complete Documentation Archive: Full Clawdbot documentation (370+ files) including channels, automation, gateway, tools, platforms, and troubleshooting guides

Quick Reference

Config file: ~/.clawdbot/clawdbot.json

Commands:

clawdbot setup              # Initialize workspace
clawdbot gateway            # Start gateway
clawdbot configure --show   # View config
clawdbot status             # Check status
clawdbot doctor             # Health check
clawdbot skills list        # List skills
clawdbot channels login     # Pair WhatsApp

Model format: provider/model (e.g., anthropic/claude-sonnet-4)

Skill precedence: workspace â ~/.clawdbot/skills â extraDirs â bundled

Dashboard: http://127.0.0.1:18789/ (when gateway running)

Best Practices

Start minimal: Only configure what you need
Use allowFrom: Restrict channel access from the start
Separate concerns: Use config includes for modular setup
Version control: Track workspace bootstrap files (AGENTS.md, etc.)
Environment variables: Store API keys in env vars, not config
Test skills: Run scripts manually before deploying
Monitor logs: Use clawdbot logs --follow during development
Backup sessions: Sessions stored in ~/.clawdbot/agents/<agent>/sessions/

Using the Documentation Archive

The complete documentation archive (references/repomix-output.xml) contains 370+ markdown files covering:

Channels: WhatsApp, Telegram, Discord, iMessage (BlueBubbles), Signal, Slack, MS Teams, Matrix, and more
Automation: Cron jobs, webhooks, Gmail Pub/Sub integration, polling
Gateway: Remote access, Tailscale, authentication, health checks, sandboxing
Tools: Browser automation, elevated exec, Firecrawl, slash commands, subagents
Platforms: macOS, iOS, Android, Linux, Windows, Docker, Hetzner
Providers: Anthropic, OpenAI, GitHub Copilot, OpenRouter, Deepgram, and more
Concepts: Agent loop, sessions, memory, compaction, model failover, streaming
CLI: Complete command reference for all clawdbot commands

When to read the archive:

Setting up specific channels (e.g., BlueBubbles for iMessage, Telegram bots)
Configuring automation (cron jobs, webhooks, Gmail watch)
Troubleshooting channel-specific issues
Understanding gateway protocols and remote access
Platform-specific setup (macOS menu bar, iOS TestFlight, Android)
Advanced features (OAuth monitoring, voice wake, location commands)

How to use it:

The archive is XML-formatted with clear file paths
Search for specific topics (e.g., “telegram”, “webhook”, “sandbox”)
Each file section includes the full path and line numbers
Cross-reference with the directory structure at the top

Getting Help

Documentation: https://docs.clawd.bot
GitHub: https://github.com/clawdbot/clawdbot
Skills Registry: https://clawdhub.com

GitHub 仓库 ↗ ← 返回陌讯 Skills 聚合平台