pokeclaw-android-ai-agent

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The PokeClaw Android AI agent monitors and consumes active window UI text and incoming notification text via Android Accessibility Services during task execution, allowing outsider-authored content displayed on screen or received in messages to directly reach the LLM.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The Hugging Face model URL downloads artifact weights at runtime via explicit code, and Hugging Face repositories are third-party unofficial hosting lacking strong first-party provenance, presenting a high risk dependency.