The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several scripts (scripts/render.mjs, scripts/preview.mjs, scripts/words.py, and scripts/check-film.mjs) to execute external system binaries including ffmpeg, ffprobe, and nice. These tools are used for silence detection, media mixing, and video encoding. The commands are executed using shell-safe list-based arguments.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection via scripts/read-look.mjs.
Ingestion points: External URLs provided by the user to scrape branding information (CSS, fonts, and metadata).
Boundary markers: The instructions in SKILL.md direct the agent to cross-reference claims against the product's site and report contradictions, providing a layer of manual verification.
Capability inventory: The skill has the capability to render high-fidelity video (render.mjs) and interact with an external editor via the Fil MCP (references/fil-editor.md).
Sanitization: read-look.mjs extracts values like fontFamily and textTransform from the browser context, though it normalizes colors and dimensions.
[EXTERNAL_DOWNLOADS]: The skill requires external dependencies including playwright and ffmpeg. It also references an external Model Context Protocol (MCP) hosted at fil.design/for-agents for final video assembly. These are documented as standard requirements for the skill's media generation functionality.

html-explainer-video