xiaomi-speaker

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The xiaomi-speaker.py script executes the mijiaAPI module as a subprocess to perform device discovery and basic control operations.
  • [EXTERNAL_DOWNLOADS]: The skill requires the third-party mijiaAPI Python package to be installed from the Python Package Index (PyPI).
  • [CREDENTIALS_UNSAFE]: The script reads sensitive session data stored in ~/.config/mijia-api/auth.json to authenticate with Xiaomi's cloud API.
  • [PROMPT_INJECTION]: Documentation of Indirect Prompt Injection surface: * Ingestion points: The say command accepts arbitrary text input from the agent for voice broadcasting in xiaomi-speaker.py. * Boundary markers: None; the text is passed directly to the speaker's TTS engine without delimiters. * Capability inventory: The skill can perform network operations via the Xiaomi API and execute local commands through the subprocess module. * Sanitization: The input text is not validated or filtered before being sent to the remote device.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 05:56 AM