OpenExCard - Open Agent Execution Cards: A workflow framework for AI agents
Overview
- Production-ready workflow cards for AI agents\ - 可直接用于生产的 AI Agent 工作流卡片 An is a that: - Defines rather than — 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 : - 定义**而非** — 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: - ✅ — Just modify config paths and run, no development from scratch — 只需修改配置路径即可运行,无需从零开发 - ✅ — Dependencies, requirements, and caveats clearly listed — 清晰列出依赖、需求和注意事项 - ✅ — Each card has its own folder with EC and documentation — 每张卡片都有独立的文件夹,包含 EC 和文档 - ✅ — Combine multiple ECs with LongTask System for complex workflows — 配合长程任务管理系统组合多张 EC 实现复杂工作流 1. The truth of workflows lies in files, not in anyone's memory.
README
OpenExCard
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
What is OpenExCard? / 什么是 OpenExCard?
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。
What is an Execution Card? / 什么是执行卡片?
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
Why OpenExCard? / 为什么选择 OpenExCard?
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
Design Philosophy / 设计哲学
“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。
Project Goals / 项目目标
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
Repository Structure / 仓库结构
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
Quick Start / 快速开始
1. Choose an ExCard / 选择一张 ExCard
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。
2. Check Dependencies / 检查依赖
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
3. Configure / 配置
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。
4. Execute / 执行
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)。
5. Skills recommended / 建议与这些技能配合使用
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)
6. Create Your Own EC (Optional) / 创建自己的执行卡片(可选)
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.md
See EC Creator SKILL.md for detailed usage.
详细用法参见 EC Creator SKILL.md。
7. If you are working with OpenClaw / 如果你正在使用OpenClaw
- 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
Example: Daily Report Publishing / 示例:每日报告发布
# 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
Why OpenExCard? / 为什么选择 OpenExCard?
| 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 |
Contributing / 贡献
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。
Current Cards / 现有卡片
Content Creation / 内容创作
| 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 |
Data Collection / 数据采集
| 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 |
Image Generation / 图像生成
| 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 |
Roadmap / 路线图
- [ ] 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
License / 许可证
MIT License - see LICENSE file
MIT 许可证 - 详见 LICENSE 文件
Acknowledgments / 致谢
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。
支持我 / Support Me
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 ☕️
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add noah-1106/openexcard