The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands via the subprocess module in scripts/render.py and scripts/tts.py to call ffmpeg and ffprobe. These calls are used for media encoding and metadata extraction. They are implemented using argument lists with shell=False to prevent command injection.
[EXTERNAL_DOWNLOADS]: The skill instructions and scripts facilitate the installation and use of external libraries including edge-tts and playwright. It also triggers the download of the Chromium browser through Playwright's CLI. These are established tools from well-known providers.
[DYNAMIC_EXECUTION]: The render.py script uses Playwright to execute JavaScript within a headless browser context to perform frame-by-frame rendering of the video content. This is a core part of the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text from script.txt and incorporates it into a generated HTML file for video production. This represents an indirect prompt injection surface.
Ingestion points: User-provided script text in script.txt.
Boundary markers: None identified to separate instructions from data in the processed script.
Capability inventory: The skill performs shell execution (ffmpeg), network operations (via edge-tts to Microsoft services), and file system writes.
Sanitization: The assets/engine.js script uses textContent to inject words into the DOM, which mitigates script execution risks within the browser environment, though the agent is responsible for the initial HTML structure.

voice-to-video