The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied narrative details and interpolates them into prompts for media generation models, creating a surface for indirect prompt injection.
Ingestion points: In SKILL.md, the agent is instructed to collect user inputs including 'Main subject', 'Character', 'Product', 'Location', 'Continuity anchors', and 'Visual prompt'.
Boundary markers: The instructions lack the use of delimiters or 'ignore' directives when placing user input into the prompt structures defined in the 'Prompt build order' section.
Capability inventory: The skill utilizes the vg generate tool to communicate with external AI endpoints (e.g., OpenAI, ByteDance, Alibaba) and performs file system writes to save generated assets to ./outputs/story/.
Sanitization: There are no sanitization or validation steps specified for the user-provided strings before they are incorporated into the executable prompts.

storytelling