support-with-evidence

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user input which creates a surface for indirect prompt injection. However, internal rules like the 'Honesty Rule' and 'Falsifiability Gate' act as logical mitigations.
  • Ingestion points: Input is accepted from the user in the form of arguments or claims.
  • Boundary markers: The input is placed in a markdown block.
  • Capability inventory: The agent uses web search for research.
  • Sanitization: The skill uses a filtering process to extract only falsifiable claims, reducing the impact of non-testable or malicious input.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:57 AM