ai-content-pipeline

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the belt-sh/cli tool via npx and references installation guidelines hosted on GitHub by the inference-sh organization. These are legitimate dependencies for the specified platform.
  • [COMMAND_EXECUTION]: The skill uses the belt CLI tool to execute remote AI models. It utilizes Bash for automation, including loops to process multiple items sequentially, which is standard for media processing workflows.
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to chain outputs from one AI model (e.g., text from a script generator) into the input of another (e.g., a text-to-speech generator). This creates an inherent attack surface where untrusted output could influence downstream actions.
  • Ingestion points: Inputs are read from intermediate JSON files like script.json, voice.json, and video.json generated in previous steps of the pipeline (SKILL.md).
  • Boundary markers: The instructions use placeholder syntax like <script-text> and <url-from-previous> but do not define explicit delimiters or escaping mechanisms to prevent instruction injection within the data.
  • Capability inventory: The skill has the capability to execute shell commands and interact with various remote AI inference models via the belt tool (SKILL.md).
  • Sanitization: There is no evidence of input sanitization or validation of the data being passed between pipeline stages.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 11:00 PM
Security Audit — agent-trust-hub — ai-content-pipeline