Travel Swarm

Integrated travel planner combining FlyAI ticket prices, Gaode and Tencent map POI, Meituan food recommendations, and fallback McDonald's options.

This package appears to be a real multi-MCP travel planner, but there are multiple inconsistencies and operational risks you should check before installing: - Verify which environment variables the code actually reads. SKILL.md lists GAODE_API_KEY but code uses AMAP_API_KEY; ensure you map your keys correctly and avoid pasting unrelated secrets. Search the code for os.getenv(...) to find all expected names (e.g., AMAP_API_KEY, MINIMAX_API_KEY, FLYAI keys). - Expect the skill to make outbound network calls to FlyAI, Amap/Gaode, Tencent, Meituan and to model endpoints (minimax). Only supply API keys you trust and run the skill in an isolated environment if possible. - The repo includes scripts and docs to run a web service on port 7860 and watchdog/resurrection scripts and even a hardcoded external IP in docs. Do not blindly run startup/watchdog scripts as root; review them first and consider running under a container or sandbox. - The codebase contains unrelated-looking CNC/industrial logic and many test/dev files — this suggests code reuse. Review the code paths that handle user input and external calls (flyai_client, multi_mcp_client, minimax_client) for any unexpected data exfiltration or unknown endpoints. - If you are not comfortable auditing the code, run it in an isolated VM/container with no sensitive credentials mounted. Prefer providing per-service API keys with limited privileges and rotate them after testing. If you want, I can list the exact files that read environment variables and the exact env-var names the code expects, or scan the codebase for hardcoded external hosts and URLs to help you decide.

Type: OpenClaw Skill Name: travel-swarm Version: 1.0.0 The skill bundle contains multiple critical Remote Code Execution (RCE) vulnerabilities due to the unsafe use of `subprocess.run(shell=True)` with unsanitized user input in `backend/utils/flyai_client.py` and `shrimp_system/main_shrimp.py`. Furthermore, numerous sensitive API keys for services like Gaode Maps, Tencent Maps, and MiniMax are hardcoded throughout the codebase (e.g., in `food_nearby_api.py`, `multi_mcp_client.py`, and `real_api_client.py`). While these appear to be severe security oversights and poor coding practices rather than intentional malware, the combination of RCE vectors and leaked credentials makes the bundle highly risky for deployment.