The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates image generation tasks using local shell commands and scripts.
It utilizes a user-defined script directory (set via the $SK variable) to invoke batch.sh and poll.sh for processing character candidates.
It employs the cohub CLI utility (a tool associated with the vendor's ecosystem) to perform complex multi-input image generations for wardrobe and age variants.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a prompt-generation engine, presenting a vulnerability surface where external data influences model behavior.
Ingestion points: The skill reads character descriptions from project-specific text files (e.g., films/<slug>/prompts/char-santiago.txt).
Boundary markers: The prompt skeleton utilizes descriptive tags (e.g., <NAME>, <ethnicity>) to structure input, but does not implement explicit "ignore embedded instructions" delimiters for the ingested text.
Capability inventory: The system possesses command execution capabilities through shell scripts (batch.sh) and image generation tools (cohub), though these are limited to the creative pipeline scope.
Sanitization: The instructions do not describe specific sanitization or filtering logic for the content of the prompt files before they are interpolated into the generation commands.

film-character-sheet