The Agent Skills Directory

[SAFE]: The skill implements a robust security and quality posture by requiring a 'JSON Prompt Pipeline'. This process forces the conversion of untrusted user input into a structured format that must pass a manual or automated review checklist before being sent to the Image API.
[EXTERNAL_DOWNLOADS]: The skill references several external documentation and research URLs (OpenAI, arXiv, Baymard Institute, Shutterstock). These references are used for grounding the agent's research and prompt engineering logic and target well-known, trusted professional and academic services.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes user requirements and external source images. However, it mitigates this risk by:
Ingestion points: Explicitly defining input sources in the source_inputs and user_requirements_summary JSON fields.
Boundary markers: Using a JSON schema to separate creative direction from API execution settings.
Capability inventory: Limiting actions to image generation and file saving within the workspace.
Sanitization: Including a mandatory review_checklist that checks for safety, rights, and brand risks before any generation occurs.
[COMMAND_EXECUTION]: The skill references the use of an image_api or imagegen system skill. There is no evidence of arbitrary shell command execution or unauthorized file system access. File operations are limited to saving project-bound images to the workspace.

image-generation