suggest-companies

Pass

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [SAFE]: The skill operates entirely on local files within the user's environment to provide company recommendations based on past interactions.\n- [COMMAND_EXECUTION]: The shell environment is limited to a small whitelist of commands (mkdir, test, date) via the allowed-tools configuration, preventing arbitrary code execution.\n- [DATA_EXFILTRATION]: No network access is requested or used, and the skill includes no instructions for transferring sensitive career or contact data externally.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external files, presenting a potential surface for indirect injection, though the risk is mitigated by the lack of powerful tools.\n
  • Ingestion points: ground-truth.md, crm/outreach.md, crm/companies.md, and domain-note files mentioned in SKILL.md.\n
  • Boundary markers: None explicitly defined in the procedures.\n
  • Capability inventory: Read, Write, and restricted Bash (mkdir, test, date) in SKILL.md.\n
  • Sanitization: Content from the files is used for scoring and ranking logic without specific escaping or validation steps described.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 29, 2026, 05:16 AM
Security Audit — agent-trust-hub — suggest-companies