self-eval

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to instructions embedded in the evaluated work or conversation history which could manipulate the scoring process.\n
  • Ingestion points: Processes user-supplied arguments ($ARGUMENTS), the full session conversation history, and the .self-eval-scores.jsonl file.\n
  • Boundary markers: Lacks explicit delimiters or specific instructions to the agent to ignore potentially malicious directives within the content being evaluated.\n
  • Capability inventory: The skill performs read and append operations on a local file (.self-eval-scores.jsonl) to track session history.\n
  • Sanitization: No validation or filtering is applied to the incoming data or the existing history file before it is used to influence the agent's logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 06:57 PM
Security Audit — agent-trust-hub — self-eval