The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several Python scripts (scripts/composite_layers.py, scripts/build_prompt.py, scripts/validate_poster_text.py, scripts/validate_project.py, scripts/validate_skill.py) that are designed to be executed locally by the agent or user for image compositing, prompt construction, and package validation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided stories, genre keywords, and visual references to generate image prompts, which represents a potential surface for indirect prompt injection.
Ingestion points: Untrusted narrative data enters the context through user inputs describing movie plots, characters, and references, which are then processed by the SKILL.md instructions and scripts/build_prompt.py script.
Boundary markers: The skill utilizes structured JSON templates (templates/project-brief.json) and specific negative restriction lists (assets/negative-prompt.txt) to guide the image generation process and prevent unwanted artifacts.
Capability inventory: The skill possesses capabilities for local file writing through image compositing scripts and command execution via the provided Python utility scripts.
Sanitization: The skill includes basic validation logic in scripts/validate_poster_text.py and scripts/validate_project.py to enforce credit naming conventions and ensure the structure of the project data.

fantasy-movie-poster