NavClaw

Smart driving — exhaustive route search, may outperform default navigation. 导航/自驾/极限避堵, dozens of routes. One-tap iOS/Android deep link. Supports 高德/Amap. 智能...

This package appears to do what it says: it queries Amap, generates route candidates, and can post results to Mattermost. Before installing: 1) Verify the source (the skill claims a GitHub repo but the registry metadata lists source/homepage as unknown — confirm the upstream repo and review it yourself). 2) Only provide an Amap API key and Mattermost bot token if you trust the code and host; prefer creating a bot account with minimal privileges and a dedicated Amap key. 3) Be aware the skill writes credentials into config.py (a plain file) and may add an entry to OpenClaw long-term memory; avoid storing secrets in long-term memory or broad-scope locations. 4) If you want minimal blast radius, run the tool locally (not letting the agent autonomously write config files), or inspect and run wrapper.py manually to control when messages/logs are uploaded. 5) If you need higher assurance, ask the publisher for an official homepage or signed release and/or perform a line-by-line audit of the bundled files (wrapper.py and navclaw.py) to confirm there are no unexpected network calls beyond Amap and configured Mattermost endpoints.

Type: OpenClaw Skill Name: navclaw Version: 1.0.3 The NavClaw skill bundle provides advanced navigation features but introduces a significant security risk through its configuration process. The instructions in SKILL.md and README.md direct the AI agent to write user-supplied sensitive data (Amap API keys and Mattermost tokens) directly into 'config.py', a Python file that is imported by the core logic in 'navclaw.py' and 'wrapper.py'. This creates a high risk of Remote Code Execution (RCE) via prompt injection, as a malicious user could provide an 'API key' containing Python code that would be executed upon the next run. While the behavior appears intended for legitimate configuration, the lack of input sanitization for a file that is later executed makes it highly vulnerable.