crm-maintenance

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill processes potentially untrusted content from email threads and calendar descriptions to update HubSpot records. This is a standard functional requirement for the skill's purpose. To mitigate risks, the skill uses a summarization approach rather than verbatim content insertion and requires explicit user confirmation before any data is written to the CRM.
  • Human-in-the-Loop Safeguards: All primary write operations, including contact creation and deal cleanup, are governed by approval gates. The instructions explicitly state that the agent must never apply updates without approval and must present side-by-side diffs for cleanup tasks.
  • Destructive Action Restrictions: The skill includes a 'Never delete records' rule, which prevents the agent from being manipulated into deleting contacts, deals, or activities, regardless of user input or external data prompts.
  • Data Integrity Patterns: The skill implements best practices for CRM management, such as case-insensitive email normalization to prevent duplicate contacts and deduplication logic for deal resolution.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 05:33 PM