benchmark-methodology
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes a process that involves ingesting untrusted data from external sources, creating a potential surface for indirect prompt injection attacks.
- Ingestion points: According to
SKILL.md, the agent is instructed to gather data from external sources including 'Competitor's own site / social / signage' and 'Reviews, directories, trade/sector associations'. - Boundary markers: There are no specific instructions or boundary markers provided to ensure the agent ignores or isolates potential instructions embedded in the external research materials.
- Capability inventory: The skill requires 'read' and 'search' capabilities to perform research and OSINT tasks, as stated in the 'Capability and Permission Boundaries' section of
SKILL.md. - Sanitization: The skill does not define any sanitization or validation logic for external data before it is processed by the agent to generate dimension scores.
Audit Metadata