The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's setup process downloads the 'Chrome for Testing' binary and installs Node.js dependencies.
Evidence: engine/src/setup/chrome.js uses the @puppeteer/browsers library to fetch the browser and includes hardcoded SHA-256 hashes for integrity verification across different platforms.
Evidence: engine/src/setup/deps.js utilizes npm ci to install dependencies from a pinned package-lock.json file into a local directory (~/.animator/deps).
[COMMAND_EXECUTION]: The skill executes various Node.js scripts and system utilities required for rendering and video encoding.
Evidence: engine/bin/batch.js and engine/bin/make.js use spawn and execFileSync to run internal animation scripts and invoke ffmpeg for video processing.
Evidence: engine/src/render/frames.js uses puppeteer-core to manage a headless browser instance for rendering SVG frames.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied SVG data which is evaluated by subagents during the analysis and verification steps.
Ingestion points: User-provided SVGs are parsed by engine/src/svg/parse.js and rendered in the browser.
Boundary markers: The skill workflow uses separate subagent contexts to isolate the processing and visual evaluation of untrusted illustration data.
Capability inventory: The engine possesses capabilities for file system operations (scoped to the style folders and temporary directories) and execution of local scripts and system tools (node, ffmpeg, browser).
Sanitization: SVGs are processed through a custom parser before being rendered, reducing the surface area for certain types of injection.

illustration-animator