deep-research

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes results from Mixpanel queries to generate interpretations, hypotheses, and dashboards, creating a surface for potential injection from the data layer.
  • Ingestion points: Data returned from Mixpanel event and property queries during Phase 2 (Validation) and Phase 3 (Investigation).
  • Boundary markers: The instructions provide a logical workflow, but do not specify technical delimiters to separate raw database content from the agent's internal instructions.
  • Capability inventory: The agent is authorized to run Mixpanel queries and create dashboards via the Mixpanel MCP server.
  • Sanitization: No explicit instructions for sanitizing or filtering database results are included in the skill documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 10:16 PM