The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill is designed to author and modify JSX code at runtime through tools such as create_motion_graphic_from_code and edit_asset.
The execution risk is mitigated by a mandatory 'Motion Graphic Code Contract' that restricts the JSX environment to pure JavaScript (no TypeScript), prohibits all import and export statements, and limits access to a specific set of pre-injected globals (e.g., React, spring, useCurrentFrame).
The code generation is based on established library references and project-specific metadata retrieved through vendor-provided tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, such as user-provided brand references, video transcripts, and library search results, which are used to derive graphic properties and layouts.
Ingestion points: Untrusted data enters the agent's context through project transcripts, video frame evidence, and library metadata fetched via browse_library.
Boundary markers: The instructions do not define specific text delimiters, but they mandate the use of the host's design spec and visual verification of composed frames to ensure alignment with project styles.
Capability inventory: The skill possesses the ability to create new assets (create_motion_graphic_from_code), modify existing asset code (edit_asset), and place items on the project timeline (edit_item).
Sanitization: The skill includes a validation mechanism (metadata.renderContract) that checks for syntax errors, API violations, and renderer incompatibilities before code execution.

create-motion-graphics