analyzing-experiment-session-replays

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill exclusively utilizes internal PostHog platform tools such as execute_sql, filter_session_recordings, and experiment_results_summary to perform its intended functions.\n- [SAFE]: SQL query templates for experiment metadata and feature flags are parameterized with placeholders for experiment and team identifiers, following standard platform integration patterns for data access.\n- [PROMPT_INJECTION]: The skill processes untrusted analytics data from session recordings and experiment metadata, which represents a potential surface for indirect prompt injection. This is a low-risk finding associated with the skill's primary purpose of behavioral analysis.\n
  • Ingestion points: Data is retrieved from system.experiments, system.feature_flags, and the output of filter_session_recordings (SKILL.md).\n
  • Boundary markers: No specific delimiters or instructions are provided to distinguish ingested session activity or console logs from the agent's operational logic.\n
  • Capability inventory: The skill is configured to use execute_sql for database queries and filter_session_recordings for data retrieval.\n
  • Sanitization: No explicit sanitization or filtering of the ingested analytics content is implemented in the workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 07:57 PM