braintrust-discover-agent-failures
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes traces and logs which could contain malicious content. It mitigates this via explicit instructions to treat such data as evidence rather than instructions (documented in references/interaction-contract.md). Ingestion points: Traces, logs, model outputs, and dataset rows. Boundary markers: Explicit instructions in Section 9 of the interaction contract. Capability inventory: None. Sanitization: Instruction to maintain independent criteria.
- [PROMPT_INJECTION]: A heuristic detection of instruction override patterns was identified in references/interaction-contract.md. This is a false positive, as the text describes how to identify and ignore injection attempts within data traces rather than attempting to bypass the agent's own constraints.
Audit Metadata