skills/qu-skills/skills/agent-tools/Gen Agent Trust Hub

agent-tools

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download a CLI binary from dist.inference.sh and cli.inference.sh (well-known platform domains). It includes SHA-256 verification and Sigstore signature checks in the manual installation steps to ensure integrity.
  • [REMOTE_CODE_EXECUTION]: The skill documents a curl | sh installation pattern. While this is a remote execution vector, it is the standard distribution method for this specific platform and is documented neutrally as intended functionality for the 'belt' CLI.
  • [COMMAND_EXECUTION]: The skill uses various bash commands to interact with the installed CLI, such as belt app run and belt login. These are standard CLI operations for AI model interaction and do not involve suspicious privilege escalation or persistence techniques.
  • [DATA_EXFILTRATION]: The CLI documentation notes that local files (e.g., images, audio) are automatically uploaded to the inference.sh cloud when provided as input paths. This is a core feature of the service and is clearly documented for the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 07:03 PM
Security Audit — agent-trust-hub — agent-tools