The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill constructs JavaScript code blocks at runtime by interpolating user-provided After Effects timeline JSON data into templates (e.g., Step 11 and Step 13). This generated code is then executed in Figma using the figma_execute or use_figma tools. If the external JSON contains specially crafted strings, it could potentially lead to execution of unintended JavaScript within the Figma plugin context.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection through its ingestion of external data.
Ingestion points: The skill accepts untrusted JSON data via clipboard paste or file references (Step 4).
Boundary markers: While Step 5 mentions validation against a schema, there are no explicit instructions or delimiters used to sanitize the data before it is interpolated into executable JavaScript code.
Capability inventory: The skill utilizes powerful tools including figma_execute, use_figma, figma_navigate, and figma_take_screenshot.
Sanitization: The instructions lack specific requirements for escaping or sanitizing string values (like layer names or property values) before they are placed into the characters property of Figma text nodes or the __PROPERTIES_JSON__ variable.

create-motion