hubspot

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFE
Full Analysis
  • Write-Safety Protocols: The skill implements a robust verification step that requires the agent to present all proposed CRM changes in a structured table for user approval before execution. This prevents automated or unintended modifications to customer records.
  • Preflight Access Verification: Instructions mandate a pre-answer gate that invokes a user-context check. This ensures that the agent operates with authoritative context and adheres to defined source mappings before retrieving or drafting data.
  • Operational Batch Limits: The skill restricts CRM object modifications to a maximum of ten items per batch. This limitation ensures that large-scale changes are broken down into manageable segments, facilitating better human review and oversight.
  • Data Ingestion Surface: The skill is designed to process external CRM data, which constitutes a standard surface for potential indirect prompt injection. The skill mitigates this through its requirement for clear read/write availability checks and the aforementioned manual approval for any data writes.
  • Tool Usage Guidance: Detailed instructions are provided for using specific HubSpot tools (e.g., search_properties, search_crm_objects), emphasizing the use of current methods over deprecated ones and promoting consistent data retrieval patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 02:40 AM
Security Audit — agent-trust-hub — hubspot