The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): The skill accesses local configuration files (.env, EXTEND.md) and image files provided as command-line arguments. This data is transmitted exclusively to legitimate AI provider endpoints (Google, OpenAI, Zhipu) as required for image generation.
[Indirect Prompt Injection] (LOW): The skill accepts external text through prompts and file inputs, which could be used to influence the generated output.
Ingestion points: Command-line arguments --prompt and files referenced by --promptfiles (scripts/types.ts).
Boundary markers: Absent; the skill passes concatenated prompt data directly to the providers.
Capability inventory: Network operations via fetch for API communication and file system read access for configuration and reference images (scripts/providers/).
Sanitization: None; the skill relies on the safety filters of the underlying AI models.
[External Downloads] (LOW): The skill utilizes npx to execute scripts, which may involve downloading the bun runtime from the npm registry if not present. Additionally, openai.ts and zhipu.ts download generated images from URLs returned by the respective AI services.

baoyu-image-gen