brida-reflex-research-claim-verification

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external research claims and source summaries, which are untrusted data points.
  • Ingestion points: Data enters through the 'state' object (claim and sources) as defined in references/custom-reflex.json.
  • Boundary markers: The instructions in custom-reflex.json attempt to scope the agent with 'Given only the supplied claim and short source summaries'.
  • Capability inventory: The skill mentions using MCP, released SDKs, or REST for integration, but explicitly states in the 'Authority' section that it does not grant permissions for sensitive actions like payments, deployments, or destructive mutations.
  • Sanitization: No explicit sanitization of the input text is mentioned in the provided scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 08:21 PM
Security Audit — agent-trust-hub — brida-reflex-research-claim-verification