omh-buzz
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes system utilities such as
ffprobeto inspect media metadata and thehermesandbuzzCLI tools for gateway management. These are standard functional operations for a messaging operator and do not involve unsafe execution of arbitrary user input. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data including inbound messages and relay receipts. It mitigates injection risks by requiring specific boundary markers (e.g.,
accepted=trueand non-empty event IDs) and using structured parsing (parse_buzz_delivery_receipt) to validate relay responses. Capabilities are limited to specific CLI operations, and the skill cautions against automated retries of ambiguous external data. - Ingestion points: Inbound messages (referenced in
references/setup.md) and delivery receipts (referenced inreferences/media.md). - Boundary markers: The skill requires explicit verification of receipt status and event IDs before accepting external state changes.
- Capability inventory: Execution of
ffprobe,hermesCLI, andbuzzCLI. - Sanitization: Implementation of structured receipt parsing to validate external relay output.
Audit Metadata