The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection by processing untrusted user-supplied data and incorporating it into instructions for generative tools.
Ingestion points: User inputs for 'Subject and action', 'Medium', and 'Genre' are collected in SKILL.md and used to construct prompts using the SCLCAM structure.
Boundary markers: The skill does not define delimiters or specific 'ignore embedded instructions' warnings when merging user input into the prompt template.
Capability inventory: The skill utilizes the vg generate tool (referenced in the Genmedia workflow) to perform image and video generation and writes files to the local ./outputs/cinema/ directory.
Sanitization: There are no instructions for sanitizing or validating user-provided strings before interpolation.

cinematography