The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run within scripts like finalize_pet_run.py and render_animation_videos.py to execute internal Python scripts and the ffmpeg utility for media processing tasks. These calls are used for orchestration and are not exposed to direct arbitrary user input.
[EXTERNAL_DOWNLOADS]: The generate_pet_images.py script makes network requests to OpenAI's official Image API to generate or edit pet artwork. This is a secondary fallback mechanism that requires a user-provided API key.
[DYNAMIC_EXECUTION]: The workflow involves several scripts that dynamically invoke other bundled Python utilities using sys.executable. This modular approach is used for deterministic spritesheet assembly and validation.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied concepts and reference images to construct prompts for image generation. It mitigates potential injection risks by using structured manifests, applying 'slugify' to identifiers, and incorporating strict 'Identity lock' and 'Transparency' rules for delegated generation tasks. Additionally, the fallback generation script includes explicit path traversal validation to ensure file operations remain within the project directory.

hatch-pet