Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Construction Mediation Kg
v1.0.0建设工程商事调解知识图谱专家技能。当用户涉及建设工程纠纷分析、商事调解策略、 调解流程设计、利益方识别、争议焦点梳理、BATNA/WATNA评估、调解协议起草等 场景时,应使用此技能。触发词包括:工程纠纷调解、商事调解、调解知识图谱、 调解策略、争议解决、工程款纠纷、工期争议、质量争议、调解协议、调解流程、 利益...
⭐ 0· 23·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
技能名与描述一致地面向建设工程调解;包含案件管理、证据管理、类案检索和与外部法律AI(如通义法睿 / dashscope、智谱)集成的代码模块,这与“增强的法律AI能力”声明匹配。值得注意的轻微不一致:registry/metadata 未声明任何必需环境变量或凭据,但 SKILL.md 和代码(farui_integration.py)明确期望 DASHSCOPE_API_KEY / ZHIPU_API_KEY 等 API key。
Instruction Scope
SKILL.md 明确指示读取包内参考文档 (references/*.md) —— 与技能目的直接相关 —— 并建议通过 pip install dashscope 并设置 API keys 来调用外部模型。关键风险点是:当调用通义法睿/智谱等服务时,案件描述和证据摘录可能被发送到第三方服务端;虽然 SKILL.md 提醒脱敏,但实际运行并不会强制脱敏,存在数据泄露或合规风险。没有发现指令要求访问与本域无关的系统路径或凭据。
Install Mechanism
包清单中没有 install spec(instruction-only / code-included),SKILL.md 推荐用户执行 pip install dashscope。缺乏正式 install 字段意味着运行时依赖需由用户手动安装;pip/ PyPI 依赖有中等风险(第三方包可能执行任意代码)。没有发现从可疑外部 URL 直接下载或执行不明二进制文件;网络交互通过已命名的服务端点(dashscope 阿里云 API)进行。建议在受控环境中安装并核验 dashscope 包来源与版本。
Credentials
SKILL.md 和 farui_integration.py 使用并建议配置 DASHSCOPE_API_KEY 和可选 ZHIPU_API_KEY,但 registry metadata 未声明这些为要求环境变量(required env vars = none)。这构成声明不一致:技能在运行时会依赖敏感 API keys(用于外部模型调用),这些凭据会影响隐私并授予远程服务接收案件数据的能力。技能不要求其它系统凭据或访问权限。
Persistence & Privilege
技能没有设置 always:true,也不会自动修改其他技能或全局 agent 配置。代码会在本地创建/写入目录(./case_data, ./evidence_data 等)并保存案件/证据 JSON,这是与功能预期一致的持久化行为;建议用户确认运行目录与权限并了解数据存储在本地磁盘的含义(备份、加密、清除策略)。
What to consider before installing
What to consider before installing:
- API keys & external services: This skill optionally integrates with third‑party legal AI providers (dashscope / 通义法睿, 智谱). You will need to provide DASHSCOPE_API_KEY / ZHIPU_API_KEY for those features. Any case text or evidence you send to those APIs will leave your environment — redact or avoid sending sensitive personal/company data unless the provider and contract meet your privacy/compliance needs.
- Manifest inconsistency: The registry metadata does not declare required env vars, but the SKILL.md and code expect them. Ask the publisher to update manifest to list optional/required credentials so you can assess permissions before enabling the skill.
- Third‑party packages: The skill recommends pip installing dashscope. Treat PyPI dependencies as a moderate risk — verify the dashscope package's provenance, pin specific versions, and install in an isolated/ephemeral environment (virtualenv / container) if possible.
- Local data storage: The skill writes case and evidence files to local directories (./case_data, ./evidence_data). Ensure these directories are placed where you want persistent data stored, consider disk encryption, access controls, and purge policies for sensitive case files.
- Code review & trust: Source files are included and largely human-readable. If you plan to use the integration features, review farui_integration.py (and other modules) and test calls with non-sensitive sample data. Prefer disabling external provider integration if you need strictly local-only operation.
- Recommended mitigations: run in isolated environment, require explicit consent before sending real case data to external APIs, confirm dashscope provider's data retention policy, and request the publisher to declare env vars in the registry and to provide a minimal reproducible install spec (including pinned dependency versions).Like a lobster shell, security has layers — review code before you run it.
latestvk972qq5rykd0waj00pge7e67m5848a8n
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
