An open Agent Execution Cards system - Production-ready workflow cards for AI agents
开放式的智能助理执行卡片 - 可直接用于生产的 AI Agent 工作流卡片
OpenExCard is an open Agent Execution Card system for orchestrating AI agent workflows.
OpenExCard 是一个开放式智能助理执行卡片系统,用于编排 AI Agent 工作流。
An Execution Card (EC) is a scenario-based work agreement and prompt template that:
- Defines "how to do" rather than "what to do" — it's an execution standard, not a task list
- Transforms agent-dependent "memory" into file-based "conventions"
- Provides step-by-step workflows with clear resource dependencies
- Ensures consistent, repeatable execution across sessions
执行卡片(EC)是场景化的工作约定和提示词模板:
- 定义**"怎么做"而非"做什么"** — 是执行标准,不是任务列表
- 将依赖 Agent 的"记忆"转化为依赖文件的"约定"
- 提供分步骤的工作流程和清晰的资源依赖
- 确保跨会话的一致、可重复执行
Unlike examples or tutorials that leave you guessing, each ExCard is production-ready:
与让你猜的示例或教程不同,每张 ExCard 都是生产就绪的:
-
✅ Ready to use — Just modify config paths and run, no development from scratch
开箱即用 — 只需修改配置路径即可运行,无需从零开发 -
✅ Well-documented — Dependencies, requirements, and caveats clearly listed
文档完善 — 清晰列出依赖、需求和注意事项 -
✅ Self-contained — Each card has its own folder with EC and documentation
独立完整 — 每张卡片都有独立的文件夹,包含 EC 和文档 -
✅ Composable — Combine multiple ECs with LongTask System for complex workflows
可组合 — 配合长程任务管理系统组合多张 EC 实现复杂工作流
"让工作流程成为人与AI平等遵守和传承的组织资产,而非绑定于某个人的个人技艺"
"Workflows as organizational assets, equally honored by humans and AI, not personal crafts bound to individuals"
Three Core Beliefs / 三大信仰:
-
Equal Partnership / 平等协作
The truth of workflows lies in files, not in anyone's memory. Both humans and AI follow the same documented conventions. 工作流的真相在文件里,不在任何人的记忆中。人类和 AI 都遵循同样的文件约定。 -
Transferable, Not Bound / 可传承,非绑定
Workflows exist independently of individuals (human or AI), becoming digital assets that persist and evolve. 工作流独立于个体(人或 AI)而存在,成为可持续演进的数字资产。 -
From Artisanal to Industrial / 从手工作坊到工业化
Transforming personal craftsmanship into reproducible, standardized methodologies that any agent can execute. 将个人技艺转化为可复制的标准化方法论,任何代理都能执行。
To make AI agent workflows:
- Simple as building blocks — snap together ECs to create complex automation
- Reusable across projects — write once, use anywhere
- Maintainable over time — clear documentation beats tribal knowledge
让 AI Agent 工作流:
- 像搭积木一样简单 — 组合 EC 创建复杂自动化
- 跨项目可复用 — 一次编写,到处使用
- 长期可维护 — 清晰文档胜过口头传承
openexcard/
├── README.md # This file / 本文件
├── LICENSE # MIT License / MIT 许可证
├── SPEC.md # (Coming soon) Full specification / (即将推出)完整规范
├── ExCard/ # Production-ready execution cards / 可直接用于生产的执行卡片
│ └── {category}/ # Category folder / 分类文件夹
│ ├── EC-XXX.md # The execution card / 执行卡片主体
│ └── README.md # Dependencies & usage guide / 依赖与使用指南
Browse the ExCard/ directory and pick a card that fits your needs.
浏览 ExCard/ 目录,选择适合你需求的卡片。
Each card has a README.md with:
每张卡片都包含 README.md,其中有:
- Required skills (with installation links) / 必需技能(含安装链接)
- Input data requirements / 输入数据要求
- Configuration parameters / 配置参数
- Usage caveats / 使用注意事项
Modify the configuration parameters (paths, API keys, etc.) to match your setup.
修改配置参数(路径、API 密钥等)以匹配你的环境。
Run the EC through your agent's execution system (e.g., OpenClaw heartbeat).
通过你的 Agent 执行系统运行 EC(例如 OpenClaw heartbeat)。
LongTask System / 长程任务管理系统
A state-driven task orchestration engine for managing complex workflows that span multiple sessions.
状态驱动的任务编排引擎,用于管理跨会话的复杂工作流。
- State Persistence / 状态持久化 - 任务状态保存到文件,可随时恢复
- Step Orchestration / 步骤编排 - 定义任务步骤和依赖关系
- Error Recovery / 错误恢复 - 支持重试和失败处理
- Perfect for / 适用于: Daily reports, multi-step prospecting, content pipelines / 每日报告、多步骤获客、内容流水线
EC Creator / 执行卡片创建工具
A validation and correction tool for creating standardized Execution Cards.
用于创建标准化执行卡片的校验和修正工具。
- Structure Validation / 结构校验 - Checks logical blocks completeness (resources, workflow, conventions) / 检查逻辑块完整性(资源、工作流、约定)
- Path Verification / 路径验证 - Validates all file paths and directories / 验证所有文件路径和目录
- Auto-Fix Suggestions / 自动修复建议 - Returns copy-pasteable Markdown snippets / 返回可复制的 Markdown 片段
- Real Skill Scanning / 真实技能扫描 - Checks
~/.openclaw/skills/for skill existence / 检查~/.openclaw/skills/目录中的技能是否存在
Use it when: / 使用场景:
- Creating a new EC from scratch / 从零创建新的 EC
- Validating an existing EC before publishing / 发布前验证现有 EC
- Converting draft workflows to standardized cards / 将草稿工作流转换为标准卡片
Workflow: / 工作流程:
Draft (temp_draft.md) → Lint → Fix → Validate → Standard EC (EC-XXX.md)
If you need a custom EC not available in the repository:
如果你需要仓库中没有的自定义 EC:
# 1. Write your workflow draft / 编写工作流草稿
cat > temp_draft.md << 'EOF'
## EC名称: My Custom Task
## EC描述: Brief description of what this card does
### 资源依赖
- Skill: my-skill
- File: input/data.json
### 工作流程
1. Read input data
2. Process with skill
3. Write output
EOF
# 2. Run EC Creator linter / 运行 EC Creator 校验
python3 ~/.openclaw/skills/ec_creator/ec_linter.py temp_draft.md
# 3. Apply fixes and validate / 应用修复并验证
# ... iterate until validation passes / ... 迭代直到验证通过
# 4. Save as standard EC / 保存为标准 EC
mv temp_draft.md ExCard/my-custom-task/EC-004.mdSee EC Creator SKILL.md for detailed usage.
详细用法参见 EC Creator SKILL.md。
- STRONGLY RECOMMEND: Set a specific agent for task management / 强烈推荐:设定专用的任务管理agent
- Update OpenExCard part in Soul.md / 在Soul.md文档中更新OpenExCard相关内容
- Install skills from Skills_Repo / 从技能仓库安装所需技能,如长程任务管理、EC创建等
# Navigate to the card / 进入卡片目录
cd ExCard/daily-report/
# Read the documentation / 阅读文档
cat README.md
# Check dependencies / 检查依赖
# - skill installed? / 技能已安装?
# - Input report available? / 输入报告可用?
# - API key configured? / API 密钥已配置?
# Modify EC-001.md with your paths / 根据你的路径修改 EC-001.md
# Then execute via your agent framework / 然后通过你的 Agent 框架执行| Problem / 问题 | Traditional / 传统方案 | OpenExCard |
|---|---|---|
| Not production-ready / 无法用于生产 | Examples lack error handling / 示例缺乏错误处理 | Battle-tested with real usage / 经过真实使用检验 |
| Unclear dependencies / 依赖不清晰 | Hidden assumptions / 隐含假设 | Explicit requirements in README / README 中明确要求 |
| Hard to customize / 难以定制 | One-size-fits-all / 一刀切 | Configurable parameters / 可配置参数 |
| No context / 缺乏上下文 | Isolated snippets / 孤立代码片段 | Full workflow with documentation / 完整工作流与文档 |
| general or customized / 通用还是定制 | longer promopts, more and more skill versions / 越来越长的上下文和越来越多的技能版本 | define specific conbined abilities/skills/resources to solve a certain single mission / 通过特定的组合能力/技能/资源解决特定的单点任务 |
| chaotic task management / 混乱的任务管理 | manually modified or just sufferring / 手动修改或忍着 | agents now can execute tasks card by card / 助理们现在可以逐张卡片执行任务 |
Want to contribute an ExCard?
想贡献一张 ExCard?
- Test in production for at least 1 week / 在生产环境测试至少 1 周
- Document all dependencies and caveats / 记录所有依赖和注意事项
- Submit a PR / 提交 PR
See CONTRIBUTING.md for detailed guidelines.
详细指南请参见 CONTRIBUTING.md。
| Card / 卡片 | Description / 描述 | Category / 分类 |
|---|---|---|
| EC-001-standard-article | 标准文章创作 (1500-2000字) - 包含调研、风格加载、配图预留、humanizer润色 | Writing / 写作 |
| EC-002-article-polish | 文章润色修改 - 支持humanizer/结构调整/语气转换/配图预留/全面润色 | Writing / 写作 |
| EC-005-style-fingerprint | 写作风格指纹提取 - 分析样本文本提取风格特征供后续写作使用 | Writing / 写作 |
| Card / 卡片 | Description / 描述 | Category / 分类 |
|---|---|---|
| EC-001-fetch-rss-24h | 24小时RSS数据采集 - 多源RSS抓取、数据库存储、HTML报告生成 | Data / 数据 |
| EC-002-sync-social-trends | 社交媒体趋势整合 - GitHub+知乎/微博/百度/头条/抖音/小红书多平台数据 | Data / 数据 |
| Card / 卡片 | Description / 描述 | Category / 分类 |
|---|---|---|
| EC-002-image-generation | 图像生成服务 - 基于Coze/Seedream生成概念图、海报、社媒配图 | Image / 图像 |
- Add more marketing cards (email outreach, content scheduling) / 添加更多营销卡片(邮件推广、内容排期)
- Add research cards (data collection, report generation) / 添加研究卡片(数据采集、报告生成)
- Create adapters for other agent frameworks (CrewAI, LangGraph) / 为其他 Agent 框架创建适配器(CrewAI、LangGraph)
- Build a registry/discovery system / 构建注册表/发现系统
MIT License - see LICENSE file
MIT 许可证 - 详见 LICENSE 文件
OpenExCard was developed by Noah while working with OpenClaw AI agents.
OpenExCard 由 Noah 与OpenClaw助理协作时开发。
Note / 注意: Each ExCard may have specific skill dependencies. Always check the card's README.md before use.
每张 ExCard 可能有特定的技能依赖。使用前请务必检查卡片的 README.md。
如果这个项目对你有帮助,可以请我喝杯咖啡 ☕️ If this project helps you, buy me a coffee ☕️
