agent-evals
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes an evaluation loop that processes traces (logs) from agent execution. This creates a conceptual surface for indirect prompt injection if the traces contain malicious data that influences the evaluating agent.
- Ingestion points: The evaluation process ingests agent traces, including tool calls and message logs (mentioned in SKILL.md).
- Boundary markers: No specific boundary markers or instruction-ignoring delimiters are specified for the trace analysis process.
- Capability inventory: The skill itself does not contain executable code or provide tools for file writing, network operations, or command execution.
- Sanitization: No sanitization or filtering logic is described for the ingestion of trace data.
- [EXTERNAL_DOWNLOADS]: The skill references the Harbor framework for standardized agent evaluation tasks and environments.
- Evidence: SKILL.md points to github.com/harbor-framework/harbor.
Audit Metadata