baoyu-image-gen

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes external commands using spawn and execFileSync. Specifically, it invokes curl to facilitate API requests through proxies and the codex CLI to support image generation via the user's Codex subscription.
  • [EXTERNAL_DOWNLOADS]: The skill connects to numerous well-known technology service providers (including Google, OpenAI, Microsoft Azure, Alibaba Cloud, and Replicate) to perform image generation tasks and retrieve the resulting media files.
  • [DATA_EXPOSURE]: The skill reads and writes configuration files in the user's home directory (~/.baoyu-skills/) and the project directory (.baoyu-skills/) to manage user preferences and API credentials provided via environment variables.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing untrusted data from local prompt files (--promptfiles) and reference images.
  • Ingestion points: Prompt content is read from user-specified markdown/text files.
  • Boundary markers: None detected in the prompt construction logic.
  • Capability inventory: Network POST requests to various AI APIs and local filesystem writes for generated images.
  • Sanitization: Prompts are passed to the providers without explicit filtering or sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 08:18 AM
Security Audit — agent-trust-hub — baoyu-image-gen