ai-control-audit
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as architecture documentation, incident history, and evaluation harness results. This creates a surface for indirect prompt injection where instructions hidden within these inputs could manipulate the final audit recommendation.
- Ingestion points: Identified in the 'Required Inputs' section, including architecture documents, telemetry summaries, and incident history.
- Boundary markers: Absent. The skill does not define specific markers or instructions to isolate the ingested data from the agent's task-level logic.
- Capability inventory: The skill makes critical recommendations (SHIP/DO NOT SHIP) based on analyzed data, though it does not explicitly call for file system or network operations in the provided text.
- Sanitization: No sanitization or validation of external data is specified in the audit methodology.
Audit Metadata