waku-agent-assistant

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow ingests outsider-authored free text as the user’s input message in the main agent loop (e.g., via Gateway/chat/Telegram/Slack handlers) and passes it to the LLM for tool use and reply, creating indirect prompt-injection exposure.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's development/install instructions explicitly run "git clone https://github.com/ShenSeanChen/waku-agent", which fetches remote repository contents (including agent instructions like SOUL.md and Python code) that are then installed/executed and therefore can directly control agent prompts or execute code.