The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses directive language to override default agent behavior and maximize autonomy.
Evidence: "Freedom Level: Maximum" and "You don't ask — you analyze... specify animation details, create motion specs, generate code, and produce delivery-ready output" (SKILL.md). These instructions encourage the agent to bypass user confirmation and take autonomous actions.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection due to the ingestion of untrusted external data combined with code generation capabilities.
Ingestion points: The agent reads design data via get_design_context, figma_take_screenshot, and figma_get_variables (SKILL.md).
Boundary markers: No delimiters or "ignore embedded instructions" warnings are specified for the ingested design context.
Capability inventory: The skill generates and prepares for the execution of React/Remotion code and shell commands like npx remotion studio (SKILL.md).
Sanitization: No sanitization or validation of the ingested design metadata is described before it is interpolated into code generation tasks.
[EXTERNAL_DOWNLOADS]: The skill references the use of external tools and packages from public registries.
Evidence: Mentions npx remotion studio, Framer Motion, GSAP, and Lottie (SKILL.md). These involve downloading and executing code from the NPM registry.
Metadata: The note.json file references a GitHub repository at https://github.com/sarveshsea/memi.
[COMMAND_EXECUTION]: The skill instructs the agent to prepare and run shell commands for rendering and studio environments.
Evidence: References to npx remotion studio and render command preparation (SKILL.md).

motion-video