ai-content-pipeline

Pass

Audited by Gen Agent Trust Hub on May 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches configuration and related skills from the author's public repository (github.com/inference-sh/skills).
  • [COMMAND_EXECUTION]: Utilizes the belt CLI tool for media generation tasks. The skill uses the allowed-tools configuration to restrict Bash execution strictly to the belt command, which is a security best practice for minimizing the attack surface.
  • [PROMPT_INJECTION]: Documents workflow patterns that process external data, creating an attack surface for indirect prompt injection.
  • Ingestion points: External text from search tools or user-provided content (e.g., ) is interpolated into model prompts as shown in SKILL.md.
  • Boundary markers: The workflows use JSON key-value pairs for tool inputs, providing structural separation, but they lack explicit prompt delimiters or instructions to the agent to ignore instructions embedded in the data variables.
  • Capability inventory: The skill orchestrates various text-to-speech, image, and video generation models through the belt CLI, as defined in SKILL.md.
  • Sanitization: No specific sanitization, filtering, or escaping of the interpolated strings is documented before the content is passed to the models in the pipeline.
Audit Metadata
Risk Level
SAFE
Analyzed
May 8, 2026, 10:25 PM