happyhorse

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a risk of indirect prompt injection.
  • Ingestion points: The skill accepts a variety of external inputs including text prompt, first_frame URLs, video URLs, and reference_images URL arrays.
  • Boundary markers: No explicit delimiters or instructions are provided to distinguish user-supplied data from system instructions.
  • Capability inventory: The skill is authorized to use the Bash tool with the belt CLI, allowing it to perform network operations and execute remote applications.
  • Sanitization: There is no evidence of sanitization or safety checks for content fetched from external URLs.
  • [EXTERNAL_DOWNLOADS]: The skill references and depends on resources outside of its direct control.
  • Package Installation: Directs the installation of the belt-sh/cli package using npx.
  • Remote Documentation: Links to installation scripts and guides hosted in the inference-sh/skills GitHub repository.
  • Platform Dependency: Relies on the inference.sh platform for all inference and generation tasks.
  • [COMMAND_EXECUTION]: The skill uses a command-line interface to interact with external services.
  • Tool Usage: Employs the belt CLI for running models (belt app run) and searching for applications (belt app search).
  • Access Control: Tool access is correctly limited to the belt binary in the allowed-tools configuration, which follows the principle of least privilege.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 12:07 PM