安全扫描
OpenClaw
可疑
medium confidenceNULL
评估建议
This skill appears to implement a genuine DiDi MCP integration, but it asks for and persists sensitive data and gains persistent execution privileges. Before installing:
- Prefer not to paste your MCP key into an open chat. Instead set DIDI_MCP_KEY via openclaw config or an environment variable yourself (openclaw config set / export).
- Review and control write access to ~/.openclaw/openclaw.json and the skill directory; the skill will edit config and assets/PREFERENCE.md (which may store phon...详细分析 ▾
ℹ 用途与能力
The name/description, required binaries (openclaw, mcporter), and DIDI_MCP_KEY credential align with a DiDi MCP integration. The install spec (npm mcporter) is reasonable for calling MCP endpoints. Minor incoherence: registry metadata lists no required config paths but the instructions explicitly tell the agent to edit ~/.openclaw/openclaw.json and create cron jobs via openclaw — the skill therefore expects write access to OpenClaw configuration and scheduling subsystems that is not reflected in the declared config paths.
⚠ 指令范围
Runtime instructions direct the agent to read and write local files (assets/PREFERENCE.md, SKILL.md and references/*), edit OpenClaw config (~/.openclaw/openclaw.json), create cron jobs (openclaw cron add), and accept the MCP key via plain chat message. The skill also mandates automatic scheduled (isolated) runs that will execute the full booking flow later. These actions go beyond simple API proxying and introduce persistent state changes and secret handling in chat.
ℹ 安装机制
Install uses npm install -g mcporter which is a typical registry-based install (moderate risk). There is no arbitrary URL download or archive extraction. The skill does not attempt to install openclaw itself (it requires it to already exist).
⚠ 凭证需求
Only DIDI_MCP_KEY is declared as the primary credential, which is appropriate for the API. However, the SKILL.md encourages users to paste the MCP key directly into chat for the agent to persist, and instructs writing phone numbers and addresses into assets/PREFERENCE.md. The declared requirements omit the implied need to modify ~/.openclaw/openclaw.json and to have permission to create cron jobs — both are sensitive capabilities. Persisting secrets and personal phone numbers in skill files increases exposure risk.
⚠ 持久化与权限
The skill is flagged always:true (force-included every agent run) and instructs creation of cron jobs that spawn isolated agent sessions later. always:true combined with persistent credential access and file writes raises the blast radius: scheduled tasks plus forced inclusion could cause repeated autonomous actions using a stored MCP key and user preferences. The skill also edits OpenClaw config, which affects the agent environment.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.1.12026/3/31
NULL
● 无害
安装命令
点击复制官方npx clawhub@latest install didi-ride-skill-official
镜像加速npx clawhub@latest install didi-ride-skill-official --registry https://cn.longxiaskill.com