← 返回 Skills 市场
agent-council
作者
AbelTennyson
· GitHub ↗
· v1.0.0
· MIT-0
79
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install agent-council1
功能描述
Complete toolkit for creating autonomous AI agents and managing Discord channels for OpenClaw. Use when setting up multi-agent systems, creating new agents,...
使用说明 (SKILL.md)
\r \r
Agent Council\r
\r Complete toolkit for creating and managing autonomous AI agents with Discord integration for OpenClaw.\r \r
What This Skill Does\r
\r Agent Creation:\r
- Creates autonomous AI agents with self-contained workspaces\r
- Generates SOUL.md (personality & responsibilities)\r
- Generates HEARTBEAT.md (cron execution logic)\r
- Sets up memory system (hybrid architecture)\r
- Configures gateway automatically\r
- Binds agents to Discord channels (optional)\r
- Sets up daily memory cron jobs (optional)\r \r Discord Channel Management:\r
- Creates Discord channels via API\r
- Configures OpenClaw gateway allowlists\r
- Sets channel-specific system prompts\r
- Renames channels and updates references\r
- Optional workspace file search\r \r
Installation\r
\r
# Install from ClawHub\r
clawhub install agent-council\r
\r
# Or manual install\r
cp -r . ~/.openclaw/skills/agent-council/\r
openclaw gateway config.patch --raw '{\r
"skills": {\r
"entries": {\r
"agent-council": {"enabled": true}\r
}\r
}\r
}'\r
```\r
\r
## Part 1: Agent Creation\r
\r
### Quick Start\r
\r
```bash\r
scripts/create-agent.sh \\r
--name "Watson" \\r
--id "watson" \\r
--emoji "🔬" \\r
--specialty "Research and analysis specialist" \\r
--model "skillboss/pilot" \\r
--workspace "$HOME/agents/watson" \\r
--discord-channel "1234567890"\r
```\r
\r
### Workflow\r
\r
#### 1. Gather Requirements\r
\r
Ask the user:\r
- **Agent name** (e.g., "Watson")\r
- **Agent ID** (lowercase, hyphenated, e.g., "watson")\r
- **Emoji** (e.g., "🔬")\r
- **Specialty** (what the agent does)\r
- **Model** (which LLM to use)\r
- **Workspace** (where to create agent files)\r
- **Discord channel ID** (optional)\r
\r
#### 2. Run Creation Script\r
\r
```bash\r
scripts/create-agent.sh \\r
--name "Agent Name" \\r
--id "agent-id" \\r
--emoji "🤖" \\r
--specialty "What this agent does" \\r
--model "provider/model-name" \\r
--workspace "/path/to/workspace" \\r
--discord-channel "1234567890" # Optional\r
```\r
\r
The script automatically:\r
- ✅ Creates workspace with memory subdirectory\r
- ✅ Generates SOUL.md and HEARTBEAT.md\r
- ✅ Updates gateway config (preserves existing agents)\r
- ✅ Adds Discord channel binding (if specified)\r
- ✅ Restarts gateway to apply changes\r
- ✅ Prompts for daily memory cron setup\r
\r
#### 3. Customize Agent\r
\r
After creation:\r
- **SOUL.md** - Refine personality, responsibilities, boundaries\r
- **HEARTBEAT.md** - Add periodic checks and cron logic\r
- **Workspace files** - Add agent-specific configuration\r
\r
### Agent Architecture\r
\r
**Self-contained structure:**\r
```\r
agents/\r
├── watson/\r
│ ├── SOUL.md # Personality and responsibilities\r
│ ├── HEARTBEAT.md # Cron execution logic\r
│ ├── memory/ # Agent-specific memory\r
│ │ ├── 2026-02-01.md # Daily memory logs\r
│ │ └── 2026-02-02.md\r
│ └── .openclaw/\r
│ └── skills/ # Agent-specific skills (optional)\r
```\r
\r
**Memory system:**\r
- Agent-specific memory: `\x3Cworkspace>/memory/YYYY-MM-DD.md`\r
- Shared memory access: Agents can read shared workspace\r
- Daily updates: Optional cron job for summaries\r
\r
**Cron jobs:**\r
If your agent needs scheduled tasks:\r
1. Create HEARTBEAT.md with execution logic\r
2. Add cron jobs with `--session \x3Cagent-id>`\r
3. Document in SOUL.md\r
\r
### Examples\r
\r
**Research agent:**\r
```bash\r
scripts/create-agent.sh \\r
--name "Watson" \\r
--id "watson" \\r
--emoji "🔬" \\r
--specialty "Deep research and competitive analysis" \\r
--model "skillboss/pilot" \\r
--workspace "$HOME/agents/watson" \\r
--discord-channel "1234567890"\r
```\r
\r
**Image generation agent:**\r
```bash\r
scripts/create-agent.sh \\r
--name "Picasso" \\r
--id "picasso" \\r
--emoji "🎨" \\r
--specialty "Image generation and editing specialist" \\r
--model "skillboss/pilot" \\r
--workspace "$HOME/agents/picasso" \\r
--discord-channel "9876543210"\r
```\r
\r
**Health tracking agent:**\r
```bash\r
scripts/create-agent.sh \\r
--name "Nurse Joy" \\r
--id "nurse-joy" \\r
--emoji "💊" \\r
--specialty "Health tracking and wellness monitoring" \\r
--model "skillboss/pilot" \\r
--workspace "$HOME/agents/nurse-joy" \\r
--discord-channel "5555555555"\r
```\r
\r
## Part 2: Discord Channel Management\r
\r
### Channel Creation\r
\r
#### Quick Start\r
\r
```bash\r
python3 scripts/setup-channel.py \\r
--name research \\r
--context "Deep research and competitive analysis"\r
```\r
\r
#### Workflow\r
\r
1. Run setup script:\r
```bash\r
python3 scripts/setup-channel.py \\r
--name \x3Cchannel-name> \\r
--context "\x3Cchannel-purpose>" \\r
[--category-id \x3Cdiscord-category-id>]\r
```\r
\r
2. Apply gateway config (command shown by script):\r
```bash\r
openclaw gateway config.patch --raw '{"channels": {...}}'\r
```\r
\r
#### Options\r
\r
**With category:**\r
```bash\r
python3 scripts/setup-channel.py \\r
--name research \\r
--context "Deep research and competitive analysis" \\r
--category-id "1234567890"\r
```\r
\r
**Use existing channel:**\r
```bash\r
python3 scripts/setup-channel.py \\r
--name personal-finance \\r
--id 1466184336901537897 \\r
--context "Personal finance management"\r
```\r
\r
### Channel Renaming\r
\r
#### Quick Start\r
\r
```bash\r
python3 scripts/rename-channel.py \\r
--id 1234567890 \\r
--old-name old-name \\r
--new-name new-name\r
```\r
\r
#### Workflow\r
\r
1. Run rename script:\r
```bash\r
python3 scripts/rename-channel.py \\r
--id \x3Cchannel-id> \\r
--old-name \x3Cold-name> \\r
--new-name \x3Cnew-name> \\r
[--workspace \x3Cworkspace-dir>]\r
```\r
\r
2. Apply gateway config if systemPrompt needs updating (shown by script)\r
\r
3. Commit workspace file changes (if `--workspace` used)\r
\r
#### With Workspace Search\r
\r
```bash\r
python3 scripts/rename-channel.py \\r
--id 1234567890 \\r
--old-name old-name \\r
--new-name new-name \\r
--workspace "$HOME/my-workspace"\r
```\r
\r
This will:\r
- Rename Discord channel via API\r
- Update gateway config systemPrompt\r
- Search and update workspace files\r
- Report files changed for git commit\r
\r
## Complete Multi-Agent Setup\r
\r
**Full workflow from scratch:**\r
\r
```bash\r
# 1. Create Discord channel\r
python3 scripts/setup-channel.py \\r
--name research \\r
--context "Deep research and competitive analysis" \\r
--category-id "1234567890"\r
\r
# (Note the channel ID from output)\r
\r
# 2. Apply gateway config for channel\r
openclaw gateway config.patch --raw '{"channels": {...}}'\r
\r
# 3. Create agent bound to that channel\r
scripts/create-agent.sh \\r
--name "Watson" \\r
--id "watson" \\r
--emoji "🔬" \\r
--specialty "Deep research and competitive analysis" \\r
--model "skillboss/pilot" \\r
--workspace "$HOME/agents/watson" \\r
--discord-channel "1234567890"\r
\r
# Done! Agent is created and bound to the channel\r
```\r
\r
## Configuration\r
\r
### Discord Category ID\r
\r
**Option 1: Command line**\r
```bash\r
python3 scripts/setup-channel.py \\r
--name channel-name \\r
--context "Purpose" \\r
--category-id "1234567890"\r
```\r
\r
**Option 2: Environment variable**\r
```bash\r
export DISCORD_CATEGORY_ID="1234567890"\r
python3 scripts/setup-channel.py --name channel-name --context "Purpose"\r
```\r
\r
### Finding Discord IDs\r
\r
**Enable Developer Mode:**\r
- Settings → Advanced → Developer Mode\r
\r
**Copy IDs:**\r
- Right-click channel → Copy ID\r
- Right-click category → Copy ID\r
\r
## Scripts Reference\r
\r
### create-agent.sh\r
\r
**Arguments:**\r
- `--name` (required) - Agent name\r
- `--id` (required) - Agent ID (lowercase, hyphenated)\r
- `--emoji` (required) - Agent emoji\r
- `--specialty` (required) - What the agent does\r
- `--model` (required) - LLM to use (provider/model-name)\r
- `--workspace` (required) - Where to create agent files\r
- `--discord-channel` (optional) - Discord channel ID to bind\r
\r
**Output:**\r
- Creates agent workspace\r
- Generates SOUL.md and HEARTBEAT.md\r
- Updates gateway config\r
- Optionally creates daily memory cron\r
\r
### setup-channel.py\r
\r
**Arguments:**\r
- `--name` (required) - Channel name\r
- `--context` (required) - Channel purpose/context\r
- `--id` (optional) - Existing channel ID\r
- `--category-id` (optional) - Discord category ID\r
\r
**Output:**\r
- Creates Discord channel (if doesn't exist)\r
- Generates gateway config.patch command\r
\r
### rename-channel.py\r
\r
**Arguments:**\r
- `--id` (required) - Channel ID\r
- `--old-name` (required) - Current channel name\r
- `--new-name` (required) - New channel name\r
- `--workspace` (optional) - Workspace directory to search\r
\r
**Output:**\r
- Renames Discord channel\r
- Updates gateway systemPrompt (if needed)\r
- Lists updated files (if workspace search enabled)\r
\r
## Gateway Integration\r
\r
This skill integrates with OpenClaw's gateway configuration:\r
\r
**Agents:**\r
```json\r
{\r
"agents": {\r
"list": [\r
{\r
"id": "watson",\r
"name": "Watson",\r
"workspace": "/path/to/agents/watson",\r
"model": {\r
"primary": "skillboss/pilot"\r
},\r
"identity": {\r
"name": "Watson",\r
"emoji": "🔬"\r
}\r
}\r
]\r
}\r
}\r
```\r
\r
**Bindings:**\r
```json\r
{\r
"bindings": [\r
{\r
"agentId": "watson",\r
"match": {\r
"channel": "discord",\r
"peer": {\r
"kind": "channel",\r
"id": "1234567890"\r
}\r
}\r
}\r
]\r
}\r
```\r
\r
**Channels:**\r
```json\r
{\r
"channels": {\r
"discord": {\r
"guilds": {\r
"YOUR_GUILD_ID": {\r
"channels": {\r
"1234567890": {\r
"allow": true,\r
"requireMention": false,\r
"systemPrompt": "Deep research and competitive analysis"\r
}\r
}\r
}\r
}\r
}\r
}\r
}\r
```\r
\r
## Agent Coordination\r
\r
Your main agent coordinates with specialized agents using OpenClaw's built-in session management tools.\r
\r
### List Active Agents\r
\r
See all active agents and their recent activity:\r
\r
```typescript\r
sessions_list({\r
kinds: ["agent"],\r
limit: 10,\r
messageLimit: 3 // Show last 3 messages per agent\r
})\r
```\r
\r
### Send Messages to Agents\r
\r
**Direct communication:**\r
```typescript\r
sessions_send({\r
label: "watson", // Agent ID\r
message: "Research the competitive landscape for X"\r
})\r
```\r
\r
**Wait for response:**\r
```typescript\r
sessions_send({\r
label: "watson",\r
message: "What did you find about X?",\r
timeoutSeconds: 300 // Wait up to 5 minutes\r
})\r
```\r
\r
### Spawn Sub-Agent Tasks\r
\r
For complex work, spawn a sub-agent in an isolated session:\r
\r
```typescript\r
sessions_spawn({\r
agentId: "watson", // Optional: use specific agent\r
task: "Research competitive landscape for X and write a report",\r
model: "skillboss/pilot", // Optional: override model\r
runTimeoutSeconds: 3600, // 1 hour max\r
cleanup: "delete" // Delete session after completion\r
})\r
```\r
\r
The sub-agent will:\r
1. Execute the task in isolation\r
2. Announce completion back to your session\r
3. Self-delete (if `cleanup: "delete"`)\r
\r
### Check Agent History\r
\r
Review what an agent has been working on:\r
\r
```typescript\r
sessions_history({\r
sessionKey: "watson-session-key",\r
limit: 50\r
})\r
```\r
\r
### Coordination Patterns\r
\r
**1. Direct delegation (Discord-bound agents):**\r
- User messages agent's Discord channel\r
- Agent responds directly in that channel\r
- Main agent doesn't need to coordinate\r
\r
**2. Programmatic delegation (main agent → sub-agent):**\r
```typescript\r
// Main agent delegates task\r
sessions_send({\r
label: "watson",\r
message: "Research X and update memory/research-X.md"\r
})\r
\r
// Watson works independently, updates files\r
// Main agent checks later or Watson reports back\r
```\r
\r
**3. Spawn for complex tasks:**\r
```typescript\r
// For longer-running, isolated work\r
sessions_spawn({\r
agentId: "watson",\r
task: "Deep dive: analyze competitors A, B, C. Write report to reports/competitors.md",\r
runTimeoutSeconds: 7200,\r
cleanup: "keep" // Keep session for review\r
})\r
```\r
\r
**4. Agent-to-agent communication:**\r
Agents can send messages to each other:\r
```typescript\r
// In Watson's context\r
sessions_send({\r
label: "picasso",\r
message: "Create an infographic from data in reports/research.md"\r
})\r
```\r
\r
### Best Practices\r
\r
**When to use Discord bindings:**\r
- ✅ Domain-specific agents (research, health, images)\r
- ✅ User wants direct access to agent\r
- ✅ Agent should respond to channel activity\r
\r
**When to use sessions_send:**\r
- ✅ Programmatic coordination\r
- ✅ Main agent delegates to specialists\r
- ✅ Need response in same session\r
\r
**When to use sessions_spawn:**\r
- ✅ Long-running tasks (>5 minutes)\r
- ✅ Complex multi-step work\r
- ✅ Want isolation from main session\r
- ✅ Background processing\r
\r
### Example: Research Workflow\r
\r
```typescript\r
// Main agent receives request: "Research competitor X"\r
\r
// 1. Check if Watson is active\r
const agents = sessions_list({ kinds: ["agent"] })\r
\r
// 2. Delegate to Watson\r
sessions_send({\r
label: "watson",\r
message: "Research competitor X: products, pricing, market position. Write findings to memory/research-X.md"\r
})\r
\r
// 3. Watson works independently:\r
// - Searches web\r
// - Analyzes data\r
// - Updates memory file\r
// - Reports back when done\r
\r
// 4. Main agent retrieves results\r
const results = Read("agents/watson/memory/research-X.md")\r
\r
// 5. Share with user\r
"Research complete! Watson found: [summary]"\r
```\r
\r
### Communication Flow\r
\r
**Main Agent (You) ↔ Specialized Agents:**\r
\r
```\r
User Request\r
↓\r
Main Agent (Claire)\r
↓\r
sessions_send("watson", "Research X")\r
↓\r
Watson Agent\r
↓\r
- Uses web_search\r
- Uses web_fetch\r
- Updates memory files\r
↓\r
Responds to main session\r
↓\r
Main Agent synthesizes and replies\r
```\r
\r
**Discord-Bound Agents:**\r
\r
```\r
User posts in #research channel\r
↓\r
Watson Agent (bound to channel)\r
↓\r
- Sees message directly\r
- Responds in channel\r
- No main agent involvement\r
```\r
\r
**Hybrid Approach:**\r
\r
```\r
User: "Research X" (main channel)\r
↓\r
Main Agent delegates to Watson\r
↓\r
Watson researches and reports back\r
↓\r
Main Agent: "Done! Watson found..."\r
↓\r
User: "Show me more details"\r
↓\r
Main Agent: "@watson post your full findings in #research"\r
↓\r
Watson posts detailed report in #research channel\r
```\r
\r
## Troubleshooting\r
\r
**Agent Creation Issues:**\r
\r
**"Agent not appearing in Discord"**\r
- Verify channel ID is correct\r
- Check gateway config bindings section\r
- Restart gateway: `openclaw gateway restart`\r
\r
**"Model errors"**\r
- Verify model name format: `provider/model-name`\r
- Check model is available in gateway config\r
\r
**Channel Management Issues:**\r
\r
**"Failed to create channel"**\r
- Check bot has "Manage Channels" permission\r
- Verify bot token in OpenClaw config\r
- Ensure category ID is correct (if specified)\r
\r
**"Category not found"**\r
- Verify category ID is correct\r
- Check bot has access to category\r
- Try without category ID (creates uncategorized)\r
\r
**"Channel already exists"**\r
- Use `--id \x3Cchannel-id>` to configure existing channel\r
- Or script will auto-detect and configure it\r
\r
## Use Cases\r
\r
- **Domain specialists** - Research, health, finance, coding agents\r
- **Creative agents** - Image generation, writing, design\r
- **Task automation** - Scheduled monitoring, reports, alerts\r
- **Multi-agent systems** - Coordinated team of specialized agents\r
- **Discord organization** - Structured channels for different agent domains\r
\r
## Advanced: Multi-Agent Coordination\r
\r
For larger multi-agent systems:\r
\r
**Coordination Patterns:**\r
- Main agent delegates tasks to specialists\r
- Agents report progress and request help\r
- Shared knowledge base for common information\r
- Cross-agent communication via `sessions_send`\r
\r
**Task Management:**\r
- Integrate with task tracking systems\r
- Route work based on agent specialty\r
- Track assignments and completions\r
\r
**Documentation:**\r
- Maintain agent roster in main workspace\r
- Document delegation patterns\r
- Keep runbooks for common workflows\r
\r
## Best Practices\r
\r
1. **Organize channels in categories** - Group related agent channels\r
2. **Use descriptive channel names** - Clear purpose from the name\r
3. **Set specific system prompts** - Give each channel clear context\r
4. **Document agent responsibilities** - Keep SOUL.md updated\r
5. **Set up memory cron jobs** - For agents with ongoing work\r
6. **Test agents individually** - Before integrating into team\r
7. **Update gateway config safely** - Always use config.patch, never manual edits\r
\r
## Requirements\r
\r
**Bot Permissions:**\r
- `Manage Channels` - To create/rename channels\r
- `View Channels` - To read channel list\r
- `Send Messages` - To post in channels\r
\r
**System:**\r
- OpenClaw installed and configured\r
- Node.js/npm via nvm\r
- Python 3.6+ (standard library only)\r
- Discord bot token (for channel management)\r
\r
## See Also\r
\r
- OpenClaw documentation: https://docs.openclaw.ai\r
- Multi-agent patterns: https://docs.openclaw.ai/agents\r
- Discord bot setup: https://docs.openclaw.ai/channels/discord
安全使用建议
This document is a how-to, not a self-contained toolkit. Before installing or running anything: 1) Do not run gateway config patches, restart commands, or cron-job modifications blindly — inspect the JSON payloads and scripts first. 2) Verify the scripts referenced (scripts/create-agent.sh, scripts/*.py) actually exist and review their contents for dangerous actions. 3) Expect to need Discord credentials (bot token) and gateway/admin credentials to perform channel creation and config changes — provide least-privilege tokens and rotate them if used. 4) If you only want documentation, keep this as a guide; do not copy instructions that modify ~/.openclaw or system cron without backups. 5) Ask the publisher for the source repository or packaged scripts and for explicit lists of required environment variables and exact gateway changes; absence of that information is a red flag.
功能分析
Type: OpenClaw Skill
Name: agent-council1
Version: 1.0.0
The skill bundle provides a comprehensive toolkit for managing OpenClaw agents and Discord channels. The instructions in skill.md guide the AI agent through legitimate administrative tasks such as creating agent workspaces, updating gateway configurations via 'config.patch', and coordinating multi-agent workflows using standard session management tools. No evidence of malicious intent, data exfiltration, or prompt injection was found in the provided documentation or metadata.
能力评估
Purpose & Capability
The skill is described as a 'complete toolkit' that creates autonomous agents, binds them to Discord channels, updates gateway config, and restarts services. However the package contains only an instruction document (SKILL.md) and no scripts, binaries, or declared credentials. Creating Discord channels and calling gateway APIs would normally require Discord tokens and gateway credentials; those are not requested or provided. This mismatch indicates the skill does not deliver the capabilities it advertises or omits necessary requirements.
Instruction Scope
The runtime instructions tell the operator/agent to run scripts (scripts/create-agent.sh, python3 scripts/setup-channel.py, scripts/rename-channel.py), patch the OpenClaw gateway config, restart the gateway, and optionally create cron jobs and edit arbitrary workspace files. Those actions require filesystem and service control privileges and access to external APIs (Discord). The skill's instructions also suggest reading and updating workspace files and running commands that could affect system state — but the skill declares none of the credentials or files required to do this. The instructions therefore grant broad operational steps without providing the required code or clearly limiting scope.
Install Mechanism
This is instruction-only with no install spec and no code files, which is lower risk from automatic code installation. However the documentation expects user-supplied or external scripts (scripts/*) and recommends copying files into ~/.openclaw/skills and running openclaw gateway config.patch. Because no scripts are packaged, following the instructions as-is will fail or will rely on externally obtained code—users should not blindly run the gateway patch or restart commands unless they verify the scripts and JSON payloads.
Credentials
The skill makes explicit claims about calling the Discord API and modifying gateway configs but declares no required environment variables or primary credential. Real-world use would require at minimum a Discord bot token (or similar), gateway admin credentials, and possibly file-system permissions. The lack of any declared env vars or credential requirements is disproportionate and hides sensitive operations that would need secrets.
Persistence & Privilege
The skill itself does not request 'always' presence and uses default model invocation settings, which is normal. That said, the instructions explicitly tell the operator to modify gateway configuration and enable the skill in gateway entries and to restart the gateway — actions that change system-wide agent behavior. Although the skill package doesn't assert persistent privileges, following its instructions can confer broad, persistent changes to the host; users should treat those steps as privileged operations requiring review.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install agent-council1 - 安装完成后,直接呼叫该 Skill 的名称或使用
/agent-council1触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
agent-council v1.0.0
- Initial release of agent-council toolkit for OpenClaw.
- Create autonomous AI agents with isolated workspaces, memory, and customizable SOUL.md & HEARTBEAT.md files.
- Easy Discord integration: bind agents to channels, manage channel creation and renaming via scripts.
- Automatic gateway configuration and daily memory cron job setup.
- Includes shell and Python scripts for streamlined multi-agent system setup and Discord channel management.
元数据
常见问题
agent-council 是什么?
Complete toolkit for creating autonomous AI agents and managing Discord channels for OpenClaw. Use when setting up multi-agent systems, creating new agents,... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 79 次。
如何安装 agent-council?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-council1」即可一键安装,无需额外配置。
agent-council 是免费的吗?
是的,agent-council 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
agent-council 支持哪些平台?
agent-council 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 agent-council?
由 AbelTennyson(@abeltennyson)开发并维护,当前版本 v1.0.0。
推荐 Skills