crm-integration

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by ingesting data from external CRM platforms (Close, HubSpot, and Salesforce) which is then processed within the agent's context.
  • Ingestion points: External data enters through API search and retrieval functions defined in SKILL.md and reference/close-deep-dive.md.
  • Boundary markers: The provided code snippets do not implement specific delimiters or 'ignore' instructions to isolate untrusted data from the CRM.
  • Capability inventory: The skill includes capabilities for network operations via httpx and file system writes to the content/ workspace path.
  • Sanitization: No validation or sanitization logic is explicitly included in the integration patterns to mitigate potentially malicious content within CRM records.
  • [COMMAND_EXECUTION]: The skill includes Python implementation patterns for performing network-based CRUD operations and authentication routines against well-known CRM API providers.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 01:06 PM
Security Audit — agent-trust-hub — crm-integration