benchmark-methodology
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The methodology directs the agent to analyze external competitor content, introducing a vulnerability surface for instructions embedded in that data.
- Ingestion points: The skill instructs the agent to examine
Competitor's own site,Case studies / work,Review directories,LinkedIn,Portfolio / craft platforms, andContent channelsas outlined in SKILL.md. - Boundary markers: There are no instructions to use delimiters or ignore potential commands found within the collected competitor evidence.
- Capability inventory: None. The skill is entirely instructional and does not provide custom scripts or tools with file-system or network access.
- Sanitization: The skill does not specify any validation or sanitization of the data retrieved from external URLs before processing.
Audit Metadata