Openclaw Work Protocol
PassAudited by ClawScan on May 1, 2026.
Overview
This is an instruction-only workflow skill with no code or credentials, but it steers agents toward autonomous progress reporting, tool use, and persistent work notes.
Install this if you want agents to follow a disciplined, proactive workflow. Before using it on sensitive or high-impact tasks, specify approval boundaries, restrict tool access as needed, and review any persistent memory or learning-log files for private information.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may act more independently and provide frequent progress updates instead of waiting for step-by-step direction.
These rules strongly shape how an agent plans, reports, and handles problems. They are consistent with the workflow purpose and do not show deception, but they can affect the agent’s behavior across tasks.
用户的要求（最高优先级） ... 自我驱动 → 自己生成 todo-list，按顺序推进 ... 自主解决问题 → 遇到问题自己搞定，别卡住 ... 主动汇报 → 定期告诉进度，不等你问
Use this skill when you want that autonomous workflow, and give explicit limits for tasks where you want the agent to pause or ask before proceeding.
For tasks where tools are available, the agent may choose command-line, network, scripting, or browser-based actions to solve problems.
The skill recommends shell, scripting, and browser automation as preferred tools. This is disclosed and broadly purpose-aligned, but those tools can have real effects if the hosting agent already has access to them.
工具选择优先级 1. **Shell 命令** → curl, wget, grep, awk, sed 2. **脚本处理** → Python, Node.js 3. **浏览器自动化** → 最后手段
Set clear approval rules for destructive, account-changing, or sensitive operations, and restrict tool access to the directories and services needed for the task.
Information from one task may be written down and reused later, which can be helpful but may preserve mistakes, outdated instructions, or private details.
The protocol explicitly asks for persistent memory and learning logs. This supports the workflow purpose, but persistent notes can retain sensitive details or influence later tasks if not reviewed.
✅ **状态持久化** - MEMORY.md（长期记忆） - learning-log.md（学习记录） - memory/YYYY-MM-DD.md（每日日志）
Review memory files periodically, exclude secrets and confidential data, and define what the agent is allowed to record or reuse.