The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow where the agent ingests untrusted data from user-provided animation specifications to generate and execute Python code. This represents a potential surface for indirect prompt injection.
Ingestion points: The skill refers to processing external specification documents such as scenes.md and scene_with_position.md in rules/scene-planning.md and templates/scene-spec-template.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are prescribed for the agent to use when reading these external specifications.
Capability inventory: The skill is designed to generate Python scripts and execute them via the manim and manimgl command-line interfaces, providing a path from processed data to code execution.
Sanitization: There are no instructions for validating or sanitizing the content of the scene specifications before they are used to generate executable code.

animation-composer