message-queues
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill audits and reviews repository content related to message queues, which constitutes an ingestion surface for untrusted data.
- Ingestion points: Auditing of queue-related repository content (identified in SKILL.md).
- Boundary markers: The skill does not explicitly define delimiters for untrusted content, though it focuses on logical auditing.
- Capability inventory: Writing audit reports to local file paths (SKILL.md).
- Sanitization: No specific sanitization or escaping of ingested content is described.
- Context: The risk is mitigated by instructions requiring explicit user approval for any state-changing actions and the recommendation to use read-only evidence for diagnosis.
- [COMMAND_EXECUTION]: The skill discusses administrative operations for message brokers (e.g., offset resets, queue purging). However, Rule 6 explicitly prohibits altering live broker state during diagnosis and requires formal authorization for any destructive recovery actions, ensuring human-in-the-loop control.
Audit Metadata