The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided story content and photographs to generate character descriptors and visual assets. This creates a surface where malicious instructions embedded in a story could potentially attempt to influence the image compilation logic.
Ingestion points: SKILL.md (story, character cards, and references), references/own-photos.md (user-provided images via drive upload or links).
Boundary markers: While no explicit delimiters are used, the skill instructs the agent to generate "concrete" and "line-keyed" descriptors, which helps constrain the output format and mitigate some free-text injection risks.
Capability inventory: picsart_film_compile_prompt (image generation), picsart_film_save_asset (persistent asset storage), picsart_scene_export (media rendering).
Sanitization: No specific sanitization or filtering is documented for incoming story text before it is used to draft prompts.
[DYNAMIC_EXECUTION]: The skill dynamically generates JSON-based scene compositions in references/views.md for processing by the picsart_scene_export tool. This is used for technical media tasks such as precisely cropping and scaling individual character views from larger design sheets. The generated structure is a rigid configuration format for a media engine and does not represent arbitrary code execution.

film-cast-design