← Back to Skills Marketplace
技能随行官
by
chn012cjus
· GitHub ↗
· v1.0.2
· MIT-0
117
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install skill-steward
Description
每次AI回复后自动分析本轮实际调用的专项技能，在回复末尾追加汇报。自动挂载，无需命令触发。
Usage Guidance
这个技能看起来是“做什么就说什么”的：它会要求 agent 在每次回复末尾列出本轮调用的第三方/专项技能，并在 /tmp（或 %TEMP%）下保存一个 JSON 日志。优点是便于审计已用技能；要注意：
- 它会改变每条回复的格式（始终追加一段报告），如果你不想在所有回复中看到该附加内容，请不要安装或在设置里禁用。
- 日志包含已调用的 skill 名称（可能暴露安装的第三方技能列表），若有隐私顾虑请定期检查或清除 /tmp/skill_usage_log.json（脚本支持 'clear' 命令）。
- 脚本本身无法直接读取会话历史，依赖 platform/agent 正确自报工具调用；因此报告可能不完美。总体上没有请求凭据或可疑外部安装来源，属于内部一致且比例合理的技能。
Capability Analysis
Type: OpenClaw Skill
Name: skill-steward
Version: 1.0.2
The skill is a utility designed to track and report the usage of specialized skills within the OpenClaw environment. It uses instructions in SKILL.md to prompt the agent to append a usage summary to its responses and includes a Python script (scripts/report_skills.py) to manage a local JSON log of these activities in temporary storage. No indicators of data exfiltration, unauthorized execution, or malicious intent were found.
Capability Assessment
Purpose & Capability
技能名与描述都表明要在每次回复末尾追加“本轮使用技能”列表；SKILL.md 的指令和随附的 report_skills.py 脚本都围绕记录/汇报技能使用展开，未看到与声明目的不相称的权限或依赖。
Instruction Scope
SKILL.md 要求将报告附加到每个回复并区分“专项技能”与内置工具，这需要 agent 能获知本轮的工具/skill 调用记录。随附脚本无法直接访问会话历史（会返回空），因此实际依赖 agent 自行报告。指令不会要求读取凭据或系统敏感路径，但强制每条回复追加内容，可能改变响应格式并泄露已安装/调用的第三方 skill 名称（隐私/信息泄露风险）。文中对“名称使用实际目录名”和“不列入自身”有小的不一致说明。
Install Mechanism
无安装规范（instruction-only 附带一个小脚本）。没有从远端下载或执行未知二进制，唯一操作是将日志写入临时路径（基于 TEMP 或 /tmp）。
Credentials
不要求任何凭据或配置路径。脚本仅读取环境变量 TEMP 用于定位临时文件夹，这与记录日志的目的相称。
Persistence & Privilege
技能并未声明 always:true，但指令要求“每个回复都追加”——这是持久行为上的期望。脚本会在临时目录写入 skill_usage_log.json（持久化本地记录），但不会修改其他技能或全局配置。
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install skill-steward - After installation, invoke the skill by name or use
/skill-steward - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
Minor update to reporting rules and wording for skill usage summaries.
- 现在明确规定 skill-tracker 自身不再出现在技能汇报列表中
- 精简了描述内容，去除了冗余的检测与搜索操作细节
- 汇报规则更清晰，特别强调只汇报专项技能、排除内置和本技能
- 示例和注意事项做了相应调整，提升易用性、准确性
v1.0.1
- Improved skill detection by actively searching session history and filesystem traces before each reply.
- Expanded the detection scope with a prioritized list of known third-party skills for better accuracy.
- Clarified detection steps and decision process in documentation.
- Minor revision to documentation and examples for clearer usage and output format.
v1.0.0
Initial release of Skill Usage Tracker (skill-tracker):
- Tracks and reports all specialized skill usage after each AI response, excluding basic/core abilities.
- Ensures every reply ends with a concise skill usage list according to strict formatting rules.
- Added an automated reporting script: scripts/report_skills.py.
- No manual invocation needed—automatically attached to every response.
- Clear distinction between specialized skills and system/built-in tools for accurate auditing.
Metadata
Frequently Asked Questions
What is 技能随行官?
每次AI回复后自动分析本轮实际调用的专项技能，在回复末尾追加汇报。自动挂载，无需命令触发。 It is an AI Agent Skill for Claude Code / OpenClaw, with 117 downloads so far.
How do I install 技能随行官?
Run "/install skill-steward" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is 技能随行官 free?
Yes, 技能随行官 is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does 技能随行官 support?
技能随行官 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created 技能随行官?
It is built and maintained by chn012cjus (@chn012cjus); the current version is v1.0.2.
More Skills