customer-persona

Warn

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to install external packages from non-standard third-party sources.
  • npx skills add belt-sh/cli is recommended for installation.
  • The skill suggests adding further dependencies from inference-sh/skills@web-search, inference-sh/skills@ai-image-generation, and inference-sh/skills@prompt-engineering.
  • [EXTERNAL_DOWNLOADS]: The skill refers users to an external installation script hosted on GitHub for CLI setup.
  • Reference: https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input within search queries and image generation prompts, creating an attack surface where malicious data could influence agent behavior.
  • Ingestion points: User-provided strings for search queries (belt app run tavily/search-assistant) and image prompts (belt app run falai/flux-dev-lora).
  • Boundary markers: Absent. User input is directly interpolated into JSON input structures for the CLI commands.
  • Capability inventory: The skill uses Bash to execute the belt CLI, which performs network operations to interface with remote AI models and search assistants.
  • Sanitization: None detected. External content and user-provided inputs are not escaped or validated before being passed to subprocesses.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 2, 2026, 02:21 PM