flux-image

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the belt-sh/cli tool via npx and references installation guidelines from a public GitHub repository hosted by the inference.sh organization.
  • Evidence: https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md and npx skills add belt-sh/cli.
  • [INDIRECT_PROMPT_INJECTION]: The skill takes user-provided text prompts and incorporates them into shell commands executed via the belt CLI. While this is the intended functionality, it represents an attack surface where maliciously crafted inputs could attempt to influence command execution.
  • Ingestion points: User prompts are used in the --input JSON argument of belt app run commands (e.g., in SKILL.md).
  • Boundary markers: None; instructions show direct interpolation into command strings.
  • Capability inventory: The skill is restricted to the Bash(belt *) toolset.
  • Sanitization: No specific sanitization or escaping instructions for user input are provided in the markdown documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 02:18 PM
Security Audit — agent-trust-hub — flux-image