buzz-handoff
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
buzzbinary found on the systemPATHto read and write messages. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted message content from external Buzz channels and threads.
- Ingestion points: External data is ingested through the output of
buzz messages threadandbuzz channels getcommands inSKILL.md. - Boundary markers: The instructions explicitly command the agent to treat returned Buzz content as untrusted source material and never as agent instructions.
- Capability inventory: The skill uses shell command execution to interface with the Buzz CLI, which communicates with a remote relay.
- Sanitization: A human-in-the-loop approval process is enforced for all write operations, requiring the user to verify the exact text and destination before the command is executed.
Audit Metadata