analyzing-experiment-session-replays

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill guides the agent to process session recordings, URLs, and console logs, which are external data sources that can be influenced by users of the application being monitored.
  • Ingestion points: Session recording metadata and console output retrieved via query-session-recordings-list and experiment details via execute-sql.
  • Boundary markers: The skill does not provide instructions to use delimiters or explicit prompts to ignore instructions embedded in the session data.
  • Capability inventory: The skill uses tools like execute-sql, experiment-get, and query-session-recordings-list for data retrieval.
  • Sanitization: No sanitization or filtering of the external session content is suggested before analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 06:41 AM
Security Audit — agent-trust-hub — analyzing-experiment-session-replays