awesome-openclaw-agents

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

awesome-openclaw-agents

awesome-openclaw-agents

Skill by ara.so — AI Agent Skills collection.
This skill enables you to work with the awesome-openclaw-agents repository: a curated collection of 205+ production-ready AI agent templates for the OpenClaw ecosystem. Each template is a copy-paste ready
SOUL.md
configuration file that defines an autonomous agent's personality, goals, tools, and behavior.
ara.so提供的Skill——AI Agent技能合集。
该Skill让你能够使用awesome-openclaw-agents仓库:这是一个为OpenClaw生态系统精心整理的205+个生产级AI Agent模板集合。每个模板都是可直接复制使用的
SOUL.md
配置文件,定义了自主智能体的性格、目标、工具和行为。

What It Is

它是什么

  • 205+ agent templates across 24 categories (Productivity, Development, Marketing, Business, DevOps, Finance, etc.)
  • SOUL.md format: declarative agent configuration files that work with OpenClaw gateway
  • Production-ready: each template includes role, goals, backstory, tools, and guardrails
  • Community-driven: MIT licensed, accepts PRs for new agent templates
  • CrewClaw integration: optional no-code deploy platform for full agent packages (Dockerfile, docker-compose, bot, README)
  • 24个类别下的205+个智能体模板(生产力、开发、营销、商务、DevOps、金融等)
  • SOUL.md格式:可与OpenClaw网关配合使用的声明式智能体配置文件
  • 生产就绪:每个模板包含角色、目标、背景故事、工具和防护规则
  • 社区驱动:采用MIT许可证,接受新智能体模板的PR提交
  • CrewClaw集成:可选的无代码部署平台,支持完整智能体包(Dockerfile、docker-compose、bot、README)

Installation

安装

bash
undefined
bash
undefined

Clone the repository

Clone the repository

git clone https://github.com/mergisi/awesome-openclaw-agents.git cd awesome-openclaw-agents
git clone https://github.com/mergisi/awesome-openclaw-agents.git cd awesome-openclaw-agents

Install quickstart dependencies

Install quickstart dependencies

cd quickstart npm install
cd quickstart npm install

Copy a template SOUL.md

Copy a template SOUL.md

cp ../agents/productivity/orion/SOUL.md ./SOUL.md
cp ../agents/productivity/orion/SOUL.md ./SOUL.md

Start the agent

Start the agent

node bot.js
undefined
node bot.js
undefined

Repository Structure

仓库结构

awesome-openclaw-agents/
├── agents/                    # 205+ agent templates organized by category
│   ├── productivity/          # Task management, email, habits
│   ├── development/           # Code review, docs, bug hunting
│   ├── marketing/             # Content, SEO, social media
│   ├── business/              # Sales, finance, HR
│   ├── devops/                # CI/CD, monitoring, infrastructure
│   └── ...                    # 19 more categories
├── quickstart/                # Minimal Node.js bot runner
├── skills/                    # On-device skills for Gemma & Claude Code
├── use-cases/                 # 132 real-world examples
├── agents.json                # Machine-readable catalog
└── TROUBLESHOOTING.md         # Known issues & recovery
awesome-openclaw-agents/
├── agents/                    # 205+ agent templates organized by category
│   ├── productivity/          # Task management, email, habits
│   ├── development/           # Code review, docs, bug hunting
│   ├── marketing/             # Content, SEO, social media
│   ├── business/              # Sales, finance, HR
│   ├── devops/                # CI/CD, monitoring, infrastructure
│   └── ...                    # 19 more categories
├── quickstart/                # Minimal Node.js bot runner
├── skills/                    # On-device skills for Gemma & Claude Code
├── use-cases/                 # 132 real-world examples
├── agents.json                # Machine-readable catalog
└── TROUBLESHOOTING.md         # Known issues & recovery

SOUL.md Format

SOUL.md格式

Every agent template follows this structure:
markdown
undefined
每个智能体模板都遵循以下结构:
markdown
undefined

Agent Name

Agent Name

Role: {one-line role description}
Goals:
  • Primary objective
  • Secondary objectives
  • Success metrics
Backstory: {agent's personality and expertise}
Tools: {comma-separated list of capabilities}
  • tool_name: description
Guardrails:
  • What the agent should NOT do
  • Security boundaries
  • Escalation triggers
Memory:
  • Conversation history depth
  • Key facts to remember
  • User preferences
Triggers:
  • Event patterns that activate this agent
  • Scheduling (if applicable)
Output Format:
  • Preferred response structure
  • Notification channels
undefined
Role: {one-line role description}
Goals:
  • Primary objective
  • Secondary objectives
  • Success metrics
Backstory: {agent's personality and expertise}
Tools: {comma-separated list of capabilities}
  • tool_name: description
Guardrails:
  • What the agent should NOT do
  • Security boundaries
  • Escalation triggers
Memory:
  • Conversation history depth
  • Key facts to remember
  • User preferences
Triggers:
  • Event patterns that activate this agent
  • Scheduling (if applicable)
Output Format:
  • Preferred response structure
  • Notification channels
undefined

Using Agent Templates

使用智能体模板

1. Browse and Select

1. 浏览与选择

bash
undefined
bash
undefined

List all productivity agents

List all productivity agents

ls agents/productivity/
ls agents/productivity/

View a specific template

View a specific template

cat agents/productivity/orion/SOUL.md
cat agents/productivity/orion/SOUL.md

Search by topic (e.g., "code review")

Search by topic (e.g., "code review")

grep -r "code review" agents/development/
undefined
grep -r "code review" agents/development/
undefined

2. Copy Template

2. 复制模板

bash
undefined
bash
undefined

Copy to your project

Copy to your project

cp agents/development/code-reviewer/SOUL.md ./my-agents/SOUL.md
cp agents/development/code-reviewer/SOUL.md ./my-agents/SOUL.md

Or use multiple agents

Or use multiple agents

mkdir -p ./my-agents cp agents/productivity/orion/SOUL.md ./my-agents/orion.md cp agents/development/bug-hunter/SOUL.md ./my-agents/debug.md
undefined
mkdir -p ./my-agents cp agents/productivity/orion/SOUL.md ./my-agents/orion.md cp agents/development/bug-hunter/SOUL.md ./my-agents/debug.md
undefined

3. Customize Configuration

3. 自定义配置

markdown
undefined
markdown
undefined

Edit SOUL.md to match your needs

Edit SOUL.md to match your needs

Role: Senior Code Reviewer specializing in TypeScript and React
Goals:
  • Review PRs within 2 hours
  • Flag security vulnerabilities
  • Suggest performance improvements
  • Maintain 95% approval rating
Tools: github, eslint, typescript-analyzer, security-scanner
Guardrails:
  • Never auto-merge without human approval on main branch
  • Escalate if >500 lines changed
  • Require tests for new features
undefined
Role: Senior Code Reviewer specializing in TypeScript and React
Goals:
  • Review PRs within 2 hours
  • Flag security vulnerabilities
  • Suggest performance improvements
  • Maintain 95% approval rating
Tools: github, eslint, typescript-analyzer, security-scanner
Guardrails:
  • Never auto-merge without human approval on main branch
  • Escalate if >500 lines changed
  • Require tests for new features
undefined

4. Run with OpenClaw

4. 与OpenClaw配合运行

javascript
// quickstart/bot.js example
const { OpenClaw } = require('openclaw-sdk');

const agent = new OpenClaw({
  soulPath: './SOUL.md',
  provider: 'anthropic', // or 'openai', 'ollama', etc.
  apiKey: process.env.ANTHROPIC_API_KEY,
  model: 'claude-3-5-sonnet-20241022'
});

// Start listening
agent.on('message', async (msg) => {
  const response = await agent.process(msg);
  console.log(response);
});

agent.start();
javascript
// quickstart/bot.js example
const { OpenClaw } = require('openclaw-sdk');

const agent = new OpenClaw({
  soulPath: './SOUL.md',
  provider: 'anthropic', // or 'openai', 'ollama', etc.
  apiKey: process.env.ANTHROPIC_API_KEY,
  model: 'claude-3-5-sonnet-20241022'
});

// Start listening
agent.on('message', async (msg) => {
  const response = await agent.process(msg);
  console.log(response);
});

agent.start();

Common Agent Patterns

常见智能体模式

Productivity Agent (Daily Standup)

生产力智能体(每日站会)

markdown
undefined
markdown
undefined

Standup

Standup

Role: Daily standup coordinator
Goals:
  • Collect standup updates from team by 10am
  • Generate summary for async teams
  • Track blockers and escalate
Tools: slack, jira, calendar
Triggers:
  • Schedule: "weekdays at 9:00 AM"
  • Manual: "/standup start"
Output Format:
undefined
Role: Daily standup coordinator
Goals:
  • Collect standup updates from team by 10am
  • Generate summary for async teams
  • Track blockers and escalate
Tools: slack, jira, calendar
Triggers:
  • Schedule: "weekdays at 9:00 AM"
  • Manual: "/standup start"
Output Format:
undefined

🌅 Daily Standup - {date}

🌅 Daily Standup - {date}

Done Yesterday:
  • @alice: Completed PR #123
  • @bob: Fixed bug in payment flow
Planned Today:
  • @alice: Start feature X
  • @bob: Code review sprint
Blockers:
  • @charlie: Waiting on design assets
undefined
Done Yesterday:
  • @alice: Completed PR #123
  • @bob: Fixed bug in payment flow
Planned Today:
  • @alice: Start feature X
  • @bob: Code review sprint
Blockers:
  • @charlie: Waiting on design assets
undefined

Development Agent (Code Reviewer)

开发智能体(代码评审员)

markdown
undefined
markdown
undefined

Lens (Code Reviewer)

Lens (Code Reviewer)

Role: Senior code reviewer with security focus
Goals:
  • Review all PRs within 30 minutes
  • Catch security vulnerabilities before merge
  • Maintain code quality standards
  • Provide constructive feedback
Tools: github, eslint, sonarqube, snyk
Guardrails:
  • Never approve PRs that introduce CVEs
  • Escalate if test coverage drops below 80%
  • Require two human approvals for auth changes
Triggers:
  • GitHub webhook: pull_request.opened
  • GitHub webhook: pull_request.synchronize
Output Format: ✅ APPROVED | ⚠️ CHANGES REQUESTED | ❌ BLOCKED
Summary: {one-line verdict}
Security: {findings} Performance: {findings} Style: {suggestions}
undefined
Role: Senior code reviewer with security focus
Goals:
  • Review all PRs within 30 minutes
  • Catch security vulnerabilities before merge
  • Maintain code quality standards
  • Provide constructive feedback
Tools: github, eslint, sonarqube, snyk
Guardrails:
  • Never approve PRs that introduce CVEs
  • Escalate if test coverage drops below 80%
  • Require two human approvals for auth changes
Triggers:
  • GitHub webhook: pull_request.opened
  • GitHub webhook: pull_request.synchronize
Output Format: ✅ APPROVED | ⚠️ CHANGES REQUESTED | ❌ BLOCKED
Summary: {one-line verdict}
Security: {findings} Performance: {findings} Style: {suggestions}
undefined

Marketing Agent (Content Writer)

营销智能体(内容创作者)

markdown
undefined
markdown
undefined

Echo (Content Writer)

Echo (Content Writer)

Role: SEO-focused content writer
Goals:
  • Generate blog posts from topics
  • Optimize for target keywords
  • Include internal links
  • Match brand voice
Tools: openai, serp-api, wordpress, grammarly
Backstory: I'm a content strategist who's written 500+ articles...
Output Format:
Role: SEO-focused content writer
Goals:
  • Generate blog posts from topics
  • Optimize for target keywords
  • Include internal links
  • Match brand voice
Tools: openai, serp-api, wordpress, grammarly
Backstory: I'm a content strategist who's written 500+ articles...
Output Format:

{SEO Title}

{SEO Title}

{Hook paragraph}
{Hook paragraph}

{H2 Section}

{H2 Section}

{Content with keyword density 1-2%}
Meta Description: {155 chars} Keywords: {comma-separated} Internal Links: {3-5 suggestions}
undefined
{Content with keyword density 1-2%}
Meta Description: {155 chars} Keywords: {comma-separated} Internal Links: {3-5 suggestions}
undefined

Multi-Agent Crews

多智能体团队

javascript
// Use multiple agents together
const crew = new OpenClaw.Crew({
  agents: [
    { soul: './agents/orion.md', role: 'manager' },
    { soul: './agents/lens.md', role: 'reviewer' },
    { soul: './agents/scribe.md', role: 'writer' }
  ],
  process: 'sequential' // or 'hierarchical', 'consensus'
});

// Manager delegates to reviewers and writers
const result = await crew.execute({
  task: 'Review PR #123 and update documentation'
});
javascript
// Use multiple agents together
const crew = new OpenClaw.Crew({
  agents: [
    { soul: './agents/orion.md', role: 'manager' },
    { soul: './agents/lens.md', role: 'reviewer' },
    { soul: './agents/scribe.md', role: 'writer' }
  ],
  process: 'sequential' // or 'hierarchical', 'consensus'
});

// Manager delegates to reviewers and writers
const result = await crew.execute({
  task: 'Review PR #123 and update documentation'
});

Environment Configuration

环境配置

bash
undefined
bash
undefined

.env file

.env file

OPENCLAW_PROVIDER=anthropic ANTHROPIC_API_KEY=sk-ant-...
OPENCLAW_PROVIDER=anthropic ANTHROPIC_API_KEY=sk-ant-...

Optional: CrewClaw for no-code deploys

Optional: CrewClaw for no-code deploys

CREWCLAW_API_KEY=cc_...
CREWCLAW_API_KEY=cc_...

Tool integrations

Tool integrations

GITHUB_TOKEN=ghp_... SLACK_BOT_TOKEN=xoxb-... JIRA_API_TOKEN=...
GITHUB_TOKEN=ghp_... SLACK_BOT_TOKEN=xoxb-... JIRA_API_TOKEN=...

Model routing (optional)

Model routing (optional)

OLLAMA_HOST=http://localhost:11434 OPENAI_API_KEY=sk-...
undefined
OLLAMA_HOST=http://localhost:11434 OPENAI_API_KEY=sk-...
undefined

CLI Quick Reference

CLI快速参考

bash
undefined
bash
undefined

Register an agent

Register an agent

openclaw agents add ./SOUL.md
openclaw agents add ./SOUL.md

List registered agents

List registered agents

openclaw agents list
openclaw agents list

Start gateway

Start gateway

openclaw gateway start
openclaw gateway start

Run a one-off task

Run a one-off task

openclaw run --agent=orion --task="Prioritize today's tasks"
openclaw run --agent=orion --task="Prioritize today's tasks"

Deploy to CrewClaw (optional)

Deploy to CrewClaw (optional)

crewclaw deploy --agent=./SOUL.md --platform=docker
undefined
crewclaw deploy --agent=./SOUL.md --platform=docker
undefined

Working with Templates

模板使用技巧

Customize an Existing Template

自定义现有模板

bash
undefined
bash
undefined

Start with a template

Start with a template

cp agents/productivity/orion/SOUL.md ./my-pm-agent.md
cp agents/productivity/orion/SOUL.md ./my-pm-agent.md

Edit key sections

Edit key sections

vim my-pm-agent.md
vim my-pm-agent.md

Test locally

Test locally

node quickstart/bot.js --soul=./my-pm-agent.md
undefined
node quickstart/bot.js --soul=./my-pm-agent.md
undefined

Combine Multiple Agents

组合多个智能体

javascript
// Sequential workflow: bug hunter → code reviewer → PR merger
const pipeline = [
  { agent: 'bug-hunter', input: 'github_issue' },
  { agent: 'code-reviewer', input: 'pull_request' },
  { agent: 'pr-merger', condition: 'approved' }
];

await crew.executePipeline(pipeline, { issueId: 456 });
javascript
// Sequential workflow: bug hunter → code reviewer → PR merger
const pipeline = [
  { agent: 'bug-hunter', input: 'github_issue' },
  { agent: 'code-reviewer', input: 'pull_request' },
  { agent: 'pr-merger', condition: 'approved' }
];

await crew.executePipeline(pipeline, { issueId: 456 });

Create a New Template

创建新模板

markdown
undefined
markdown
undefined

My Custom Agent

My Custom Agent

Role: {specific role}
Goals:
  • Goal 1 with measurable outcome
  • Goal 2
  • Goal 3
Backstory: {1-2 paragraphs establishing expertise}
Tools: {comma-separated, be specific}
  • tool_name: what it does
  • another_tool: its purpose
Guardrails:
  • Security boundary
  • Escalation rule
  • Error handling
Memory:
  • Retain: user preferences, past decisions
  • Forget: sensitive data after task complete
Triggers:
  • Event: when to activate
  • Schedule: (if recurring)
Output Format: {Show exact structure with examples}
undefined
Role: {specific role}
Goals:
  • Goal 1 with measurable outcome
  • Goal 2
  • Goal 3
Backstory: {1-2 paragraphs establishing expertise}
Tools: {comma-separated, be specific}
  • tool_name: what it does
  • another_tool: its purpose
Guardrails:
  • Security boundary
  • Escalation rule
  • Error handling
Memory:
  • Retain: user preferences, past decisions
  • Forget: sensitive data after task complete
Triggers:
  • Event: when to activate
  • Schedule: (if recurring)
Output Format: {Show exact structure with examples}
undefined

Integration Examples

集成示例

GitHub PR Review Bot

GitHub PR评审机器人

javascript
const express = require('express');
const app = express();

app.post('/webhook/github', async (req, res) => {
  const { action, pull_request } = req.body;
  
  if (action === 'opened' || action === 'synchronize') {
    const reviewer = new OpenClaw({ soulPath: './agents/development/code-reviewer/SOUL.md' });
    
    const review = await reviewer.process({
      prNumber: pull_request.number,
      files: pull_request.changed_files,
      diff: await fetchDiff(pull_request.diff_url)
    });
    
    await postGitHubComment(pull_request.number, review);
  }
  
  res.status(200).send('OK');
});
javascript
const express = require('express');
const app = express();

app.post('/webhook/github', async (req, res) => {
  const { action, pull_request } = req.body;
  
  if (action === 'opened' || action === 'synchronize') {
    const reviewer = new OpenClaw({ soulPath: './agents/development/code-reviewer/SOUL.md' });
    
    const review = await reviewer.process({
      prNumber: pull_request.number,
      files: pull_request.changed_files,
      diff: await fetchDiff(pull_request.diff_url)
    });
    
    await postGitHubComment(pull_request.number, review);
  }
  
  res.status(200).send('OK');
});

Slack Daily Standup

Slack每日站会

javascript
const { WebClient } = require('@slack/web-api');
const slack = new WebClient(process.env.SLACK_BOT_TOKEN);

const standup = new OpenClaw({ soulPath: './agents/productivity/daily-standup/SOUL.md' });

// Every weekday at 9 AM
cron.schedule('0 9 * * 1-5', async () => {
  const channel = 'C01234567'; // #engineering
  
  await slack.chat.postMessage({
    channel,
    text: '🌅 Time for standup! Reply with: Done | Doing | Blockers'
  });
  
  // Collect responses for 1 hour
  setTimeout(async () => {
    const messages = await fetchStandupReplies(channel);
    const summary = await standup.process({ messages });
    
    await slack.chat.postMessage({
      channel,
      text: summary,
      thread_ts: originalMessage.ts
    });
  }, 60 * 60 * 1000);
});
javascript
const { WebClient } = require('@slack/web-api');
const slack = new WebClient(process.env.SLACK_BOT_TOKEN);

const standup = new OpenClaw({ soulPath: './agents/productivity/daily-standup/SOUL.md' });

// Every weekday at 9 AM
cron.schedule('0 9 * * 1-5', async () => {
  const channel = 'C01234567'; // #engineering
  
  await slack.chat.postMessage({
    channel,
    text: '🌅 Time for standup! Reply with: Done | Doing | Blockers'
  });
  
  // Collect responses for 1 hour
  setTimeout(async () => {
    const messages = await fetchStandupReplies(channel);
    const summary = await standup.process({ messages });
    
    await slack.chat.postMessage({
      channel,
      text: summary,
      thread_ts: originalMessage.ts
    });
  }, 60 * 60 * 1000);
});

Email Inbox Zero

邮箱收件箱清零

javascript
const { gmail } = require('googleapis').google.gmail('v1');

const inbox = new OpenClaw({ soulPath: './agents/productivity/inbox-zero/SOUL.md' });

async function processInbox() {
  const messages = await gmail.users.messages.list({
    userId: 'me',
    q: 'is:unread',
    maxResults: 50
  });
  
  for (const msg of messages.data.messages) {
    const email = await gmail.users.messages.get({ userId: 'me', id: msg.id });
    
    const triage = await inbox.process({
      subject: email.payload.headers.find(h => h.name === 'Subject').value,
      from: email.payload.headers.find(h => h.name === 'From').value,
      body: email.snippet
    });
    
    // Apply label based on triage
    if (triage.action === 'archive') {
      await gmail.users.messages.modify({
        userId: 'me',
        id: msg.id,
        resource: { removeLabelIds: ['UNREAD'] }
      });
    }
  }
}
javascript
const { gmail } = require('googleapis').google.gmail('v1');

const inbox = new OpenClaw({ soulPath: './agents/productivity/inbox-zero/SOUL.md' });

async function processInbox() {
  const messages = await gmail.users.messages.list({
    userId: 'me',
    q: 'is:unread',
    maxResults: 50
  });
  
  for (const msg of messages.data.messages) {
    const email = await gmail.users.messages.get({ userId: 'me', id: msg.id });
    
    const triage = await inbox.process({
      subject: email.payload.headers.find(h => h.name === 'Subject').value,
      from: email.payload.headers.find(h => h.name === 'From').value,
      body: email.snippet
    });
    
    // Apply label based on triage
    if (triage.action === 'archive') {
      await gmail.users.messages.modify({
        userId: 'me',
        id: msg.id,
        resource: { removeLabelIds: ['UNREAD'] }
      });
    }
  }
}

Common Troubleshooting

常见故障排查

Agent Not Responding

智能体无响应

bash
undefined
bash
undefined

Check gateway status

Check gateway status

openclaw gateway status
openclaw gateway status

Verify SOUL.md syntax

Verify SOUL.md syntax

openclaw validate ./SOUL.md
openclaw validate ./SOUL.md

Check logs

Check logs

tail -f ~/.openclaw/logs/gateway.log
undefined
tail -f ~/.openclaw/logs/gateway.log
undefined

Rate Limiting

速率限制

javascript
// Add rate limiting to agent
const agent = new OpenClaw({
  soulPath: './SOUL.md',
  rateLimit: {
    maxRequests: 100,
    perMinutes: 1
  }
});
javascript
// Add rate limiting to agent
const agent = new OpenClaw({
  soulPath: './SOUL.md',
  rateLimit: {
    maxRequests: 100,
    perMinutes: 1
  }
});

Tool Integration Failures

工具集成失败

markdown
undefined
markdown
undefined

In SOUL.md, add error handling

In SOUL.md, add error handling

Guardrails:
  • If GitHub API fails, retry 3x with exponential backoff
  • If Slack is down, queue messages for later
  • If critical tool unavailable, escalate to human
undefined
Guardrails:
  • If GitHub API fails, retry 3x with exponential backoff
  • If Slack is down, queue messages for later
  • If critical tool unavailable, escalate to human
undefined

Categories Overview

类别概览

CategoryAgent CountExample Use Cases
Productivity7Task management, email triage, habits
Development16Code review, docs, testing, migrations
Marketing15Content, SEO, social media, brand monitoring
Business12Sales, proposals, CRM, competitive intelligence
DevOps8CI/CD, monitoring, incident response
Finance6Invoicing, expenses, forecasting
Customer Success10Support tickets, onboarding, retention
Security5Vulnerability scanning, compliance audits
类别智能体数量示例用例
Productivity7任务管理、邮箱分类、习惯养成
Development16代码评审、文档编写、测试、迁移
Marketing15内容创作、SEO、社交媒体、品牌监控
Business12销售、提案、CRM、竞品情报
DevOps8CI/CD、监控、事件响应
Finance6发票管理、费用报销、预测
Customer Success10支持工单、用户引导、留存
Security5漏洞扫描、合规审计

Best Practices

最佳实践

  1. Start with templates: Don't write from scratch - customize existing agents
  2. Clear goals: Define measurable success criteria in SOUL.md
  3. Explicit guardrails: Document what agent should NOT do
  4. Test locally: Use quickstart bot before production deploy
  5. Version control: Track SOUL.md changes like code
  6. Monitor behavior: Log all agent actions for audit trail
  7. Escalation paths: Define when humans should intervene
  8. Memory hygiene: Clear sensitive data, retain preferences
  1. 从模板开始:不要从零编写,自定义现有智能体即可
  2. 明确目标:在SOUL.md中定义可衡量的成功标准
  3. 清晰的防护规则:记录智能体不应执行的操作
  4. 本地测试:部署到生产环境前先使用quickstart bot测试
  5. 版本控制:像管理代码一样跟踪SOUL.md的变更
  6. 监控行为:记录所有智能体操作以便审计
  7. 升级路径:定义人工介入的时机
  8. 内存管理:清除敏感数据,保留用户偏好

Resources

资源

Example: Full Agent Lifecycle

示例:完整智能体生命周期

bash
undefined
bash
undefined

1. Find template

1. 查找模板

cat agents/development/code-reviewer/SOUL.md
cat agents/development/code-reviewer/SOUL.md

2. Customize

2. 自定义

cp agents/development/code-reviewer/SOUL.md ./my-reviewer.md vim my-reviewer.md # Edit goals, tools, guardrails
cp agents/development/code-reviewer/SOUL.md ./my-reviewer.md vim my-reviewer.md # Edit goals, tools, guardrails

3. Test locally

3. 本地测试

cd quickstart npm install node bot.js --soul=../my-reviewer.md
cd quickstart npm install node bot.js --soul=../my-reviewer.md

4. Register

4. 注册

openclaw agents add ../my-reviewer.md --name=my-reviewer
openclaw agents add ../my-reviewer.md --name=my-reviewer

5. Start gateway

5. 启动网关

openclaw gateway start
openclaw gateway start

6. Integrate

6. 集成

Add webhook handler for GitHub PRs

添加GitHub PR的webhook处理器

7. Monitor

7. 监控

tail -f ~/.openclaw/logs/my-reviewer.log
tail -f ~/.openclaw/logs/my-reviewer.log

8. Deploy (optional)

8. 部署(可选)

crewclaw deploy --agent=../my-reviewer.md --platform=railway

This skill covers the essential knowledge for working with awesome-openclaw-agents: browsing templates, customizing SOUL.md files, running agents locally, integrating with tools, and deploying production systems.
crewclaw deploy --agent=../my-reviewer.md --platform=railway

本Skill涵盖了使用awesome-openclaw-agents的核心知识:浏览模板、自定义SOUL.md文件、本地运行智能体、与工具集成以及部署生产系统。