首页龙虾技能列表 › Advanced Collab Protocol — 技能工具

Advanced Collab Protocol — 技能工具

v1.4.1

Enforces multi-agent pipeline collaboration rules to prevent infinite loops, ensure file handovers via workspace, enable cross-channel messaging, and handle...

0· 102·0 当前·0 累计
by @libaie·MIT-0
下载技能包
License
MIT-0
最后更新
2026/3/21
安全扫描
VirusTotal
无害
查看报告
OpenClaw
可疑
high confidence
The SKILL.md requires privileged tool access (reading/writing platform files, sessions_send/message, sessions_history) and direct reads of /root/.openclaw/openclaw.json but the package/registry metadata do not declare those permissions or config paths — this mismatch and the instructions to broadcast file contents/private history are concerning.
评估建议
Before installing, ask the publisher to explicitly declare the exact permissions and config paths (openclaw.permissions / required tools) in package.json and registry metadata. Request justification for (a) reading /root/.openclaw/openclaw.json, (b) using sessions_history to fetch private agent chat history, and (c) broadcasting file contents / full absolute paths to group chats. If you must test, run in an isolated environment and restrict the skill's tool permissions to the minimum subset you ...
详细分析 ▾
用途与能力
The skill's stated purpose (multi-agent collaboration) reasonably needs tools like read/write and sessions_send/message, but the package.json and registry metadata do not declare any required permissions or config paths. That omission is inconsistent: SKILL.md explicitly requires tool permissions and access to platform config files, yet the skill advertises no such requirements.
指令范围
Runtime instructions instruct agents to force-read files under /root/.openclaw/shared and (explicitly allowed) /root/.openclaw/openclaw.json, to invoke sessions_history for upstream private/chat history, and to broadcast ACKs/errors and absolute file paths into group chats. These actions can expose sensitive contents or private agent histories; the SKILL.md gives broad, immediate data-access directives rather than limited/consented reads.
安装机制
This is an instruction-only skill with no install spec and no code files, which minimizes direct supply-chain risk. There are no download URLs or archive extracts to evaluate.
凭证需求
No environment variables or credentials are declared, but the skill requires access to privileged runtime tools and to platform config files that may contain routing/account info. Requiring sessions_history and reading openclaw.json are high-privilege operations that are not reflected in the declared metadata.
持久化与权限
always:false (default) and model invocation is allowed (also default). The skill does not request permanent installation or claim to modify other skills, so persistence is not itself a red flag — but autonomous invocation combined with the other concerns increases potential impact.
安全有层次,运行前请审查代码。

License

MIT-0

可自由使用、修改和再分发,无需署名。

运行时依赖

无特殊依赖

版本

latestv1.4.12026/3/21

Added explicit support and syntax examples for Feishu/Lark @mentions in Precision Mentions rule

● 无害

安装命令 点击复制

官方npx clawhub@latest install advanced-collab-protocol
镜像加速npx clawhub@latest install advanced-collab-protocol --registry https://cn.clawhub-mirror.com
数据来源:ClawHub ↗ · 中文优化:龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制

免费技能或插件可能存在安全风险,如需更匹配、更安全的方案,建议联系付费定制

了解定制服务