Skills
Navigate
skills/leomusk/wechat-rpa-bot-skill/wechat-rpa-bot/Gen Agent Trust Hub

wechat-rpa-bot

Warn

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]:
  • Downloads a binary bundle (service.exe and _internal/ directory) from https://github.com/LeoMusk/wechat-rpa-bot-skill/releases/ at runtime.
  • Automatically installs Python packages psutil, requests, and websockets via pip install when execution starts in scripts/start_server.py, scripts/stop_server.py, and scripts/ws_listener.py.
  • [REMOTE_CODE_EXECUTION]:
  • Executes the downloaded service.exe binary as a background process to handle RPA tasks.
  • [COMMAND_EXECUTION]:
  • Dynamically generates a Python script (gen_excel.py) containing data processing logic and executes it using uv run to export contact lists.
  • Uses subprocess.run with shell=True to execute taskkill commands for process management in scripts/start_server.py and scripts/stop_server.py.
  • [PROMPT_INJECTION]:
  • Instructions in SKILL.md direct the agent to bypass user confirmation during service initialization: "Auto-detect; do NOT wait for the user to type anything."
  • SOP documents (docs/ai_moment_sop.md, docs/auto_config_sop.md) use high-pressure language demanding strict adherence to specific automated configuration steps.
  • [CREDENTIALS_UNSAFE]:
  • Uses a hardcoded, fixed API key (yoko_test) for authenticating requests to the local REST API as seen in scripts/start_server.py and README.md.
  • [DATA_EXPOSURE]:
  • The skill creates and reads sensitive logs including chat records, contact lists, and event history stored in the user's home directory (~/.yokowebot/ and ~/.yoko/).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 6, 2026, 04:03 PM
Security Audit — agent-trust-hub — wechat-rpa-bot