prompt-engineering

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill demonstrates usage of the belt CLI tool for executing remote AI model tasks. Access to the system shell is restricted via the allowed-tools configuration, which limits the Bash tool to only commands starting with belt, effectively applying the principle of least privilege.
  • [EXTERNAL_DOWNLOADS]: The skill references installation instructions and related skill packages hosted on GitHub within the inference-sh organization. These resources are for documented setup and extension of the skill's capabilities and originate from the service provider's repository.
  • [DATA_EXFILTRATION]: No sensitive file access or unauthorized data transmission patterns were detected. Network operations are confined to the documented functionality of the CLI tool interacting with its primary platform (inference.sh).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 08:29 PM
Security Audit — agent-trust-hub — prompt-engineering