quality-assurance-iteration-agent
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The
KimiSemanticAudit.auditfunction inmodules/kimi_semantic_audit.pyaccepts untrusted data (asset_contentandbrand_guidelines) and interpolates it directly into a prompt template for an external LLM. An adversary could include hidden instructions within a marketing asset to manipulate the audit's outcome, such as forcing a passing score regardless of actual quality or consistency. - Ingestion points:
asset_contentandbrand_guidelinesparameters inmodules/kimi_semantic_audit.py. - Boundary markers: No delimiters or safety instructions are used to distinguish untrusted content from the system prompt.
- Capability inventory: The skill controls the approval workflow for marketing assets, which are then passed to downstream agents for autonomous posting.
- Sanitization: No filtering or escaping of the input data is performed before it is sent to the LLM API.
Audit Metadata