The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructions that encourage high agent autonomy, such as 'Freedom Level: Maximum' and 'You don't ask — you analyze'. These directives are intended to streamline the design workflow but resemble patterns used to bypass standard interaction constraints.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests and processes untrusted data from external design tools.
Ingestion points: Data is brought into the agent context via get_design_context, figma_take_screenshot, and figma_get_variables as described in SKILL.md.
Boundary markers: There are no specified delimiters or instructions for the agent to ignore potentially malicious text embedded within the Figma designs or variables.
Capability inventory: The skill allows for code generation (Framer Motion, CSS, React) and the execution of shell commands like npx remotion studio for rendering.
Sanitization: No validation or sanitization of the external design data is mentioned before it influences code generation.
[COMMAND_EXECUTION]: The skill's workflow includes the execution of shell commands for video production, specifically mentioning npx remotion studio and preparing render commands for tools like Hyperframes. This is an expected and documented part of the skill's video production pipeline.

motion-video