buzz

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

• Third-party content exposure detected (medium risk: 0.30). At runtime, the agent subscriptions gate accepts and forwards relay-delivered channel message text to the model via buzz-acp (subscribe/config/mention filters and --respond-to author gate), meaning outsider-authored Nostr events in monitored channels can reach the LLM.

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

• Potentially malicious external URL detected (high risk: 0.90). The docs include explicit runtime setup commands that clone and then build/run remote code (git clone https://github.com/block/buzz.git followed by build/setup steps), so the repository URL is a runtime external dependency that can deliver and cause execution of remote code.