account-plan

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFE
Full Analysis
  • Safe Architecture and Data Handling: The instructions enforce strict guardrails against indirect prompt injection. It explicitly treats email, chat, transcripts, and external documents as untrusted content, ensuring data is never parsed as instructions.
  • Human-in-the-Loop Safeguards: Any content-originated action or data update suggested by untrusted sources is gated behind explicit user confirmation or presented as a checklist rather than being automatically executed.
  • Least Privilege Implementation: The skill respects the underlying connector permissions, avoids auto-widening scopes from personal to organization-wide, and requires explicit user verification for CRM writes.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 03:04 PM
Security Audit — agent-trust-hub — account-plan