The Agent Skills Directory

[SAFE]: The skill's primary function is to act as a cinematographic expert, expanding simple scene descriptions into technically detailed prompts for AI video models.
[SAFE]: Script execution is handled safely using double-quoted variables, preventing shell command injection from user-supplied subject or intent strings.
[SAFE]: The skill relies on local core scripts for the actual video generation, avoiding external downloads or remote code execution from untrusted sources.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided scene descriptions and interpolates them into a technical prompt template. While this is a standard design for prompt-expansion skills, it represents a surface where a user could attempt to inject instructions for the downstream video model.
Ingestion points: scripts/generate-film.sh via the --subject argument.
Boundary markers: Absent; the subject is directly embedded into the SCENE field of the constructed prompt.
Capability inventory: Execution of media generation scripts (core/media/generate-video.sh).
Sanitization: No escaping or validation of the subject string is performed before interpolation, relying on the downstream model's own safety guardrails.

muapi-cinema-director