crm-duplicate-detector

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill instructions describe a legitimate data processing workflow for CRM maintenance. No malicious code, obfuscation, or unauthorized access patterns were detected.\n- [DATA_EXFILTRATION]: The skill handles sensitive CRM data (emails, names, phone numbers) which is the intended use case. Data is processed locally and outputs are directed to the standard output directory (/mnt/user-data/outputs/) or interactive React artifacts. There are no network operations to non-whitelisted or suspicious domains.\n- [PROMPT_INJECTION]: The skill was evaluated for potential indirect prompt injection vulnerabilities due to its ingestion of untrusted CRM data.\n
  • Ingestion points: CRM data fetched via HubSpot and Salesforce MCP tools, and user-provided CSV files.\n
  • Boundary markers: The skill defines expected column headers and schemas, which help delimit input data.\n
  • Capability inventory: Capabilities are restricted to read operations (MCP), local file system writes, and UI generation; there are no tools for arbitrary code execution or external network communication.\n
  • Sanitization: Robust normalization steps (stripping accents, case-folding, and removing punctuation) are implemented, which serve to sanitize input and reduce the risk of malicious payload execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 02:21 PM
Security Audit — agent-trust-hub — crm-duplicate-detector