The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several command-line tools to perform its tasks. The script scripts/render.mjs uses child_process.execFileSync to call ffmpeg for encoding image frames into an MP4 video. Additionally, the skill instructions in SKILL.md and references/replica-and-qc.md specify the use of node to run the rendering engine and yt-dlp or ffprobe to process external video references.
[EXTERNAL_DOWNLOADS]: The skill uses yt-dlp to download external video content for style replication and analysis. It also utilizes puppeteer-core to launch a headless browser instance for rendering HTML-based animations. These actions are standard for the skill's documented functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could potentially harbor indirect injection attacks.
Ingestion points: The agent ingests user-provided text scripts, video transcriptions, and external video metadata (SKILL.md, references/replica-and-qc.md).
Boundary markers: The instructions do not define specific delimiters or "ignore previous instructions" guards when processing this content.
Capability inventory: The agent has access to shell execution (ffmpeg, yt-dlp, node), file system writes (mkdirSync), and web rendering (puppeteer).
Sanitization: There is no explicit sanitization or validation mentioned for the content interpolated into the HTML templates before rendering.

space-video-broll