trace
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including session recordings, logs, and event streams, which provides a potential surface for indirect prompt injection attacks.
- Ingestion points: The skill reads session replay analysis, click/scroll/navigation patterns, and real user operation logs (SKILL.md).
- Boundary markers: The skill specifies strict boundary markers including PII masking, whitelisting interactive elements, and requiring DPA coverage for third-party data (SKILL.md, reference/session-analysis.md).
- Capability inventory: Capabilities include behavior pattern extraction, frustration detection, and journey reconstruction, which are qualitative analysis tasks and do not involve high-privilege system operations (SKILL.md).
- Sanitization: Instructions mandate client-side redaction of all HTML text nodes and images before data leaves the browser, providing a strong privacy-by-default control (reference/session-analysis.md).
Audit Metadata