The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input that is subsequently interpolated into a prompt for a video generation model, creating a potential injection surface.
Ingestion points: The --subject parameter in the scripts/generate-film.sh shell script accepts arbitrary user-supplied text.
Boundary markers: The script uses a SCENE: header for the input but lacks explicit delimiters or instructions telling the downstream AI model to ignore instructions embedded within the user text.
Capability inventory: The skill is capable of executing shell scripts to call media generation primitives (core/media/generate-video.sh).
Sanitization: No filtering, escaping, or validation is performed on the $SUBJECT variable before it is included in the DIRECTOR_PROMPT string.
[COMMAND_EXECUTION]: The script scripts/generate-film.sh invokes a platform primitive located at a relative path (../../../../core/media/generate-video.sh). While the user-supplied SUBJECT variable is double-quoted to prevent direct shell injection within this script, the pattern relies on the secure implementation of the external core script to handle the passed data safely.

muapi-cinema-director