The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts and system utilities to perform its tasks.
Evidence: scripts/lib/launch-check.mjs and scripts/lib/run-ego.mjs use child_process.spawn and spawnSync to invoke ego-browser.
Evidence: The skill executes Node.js modules for configuration and login verification.
[DYNAMIC_EXECUTION]: The skill dynamically generates a temporary entry script to facilitate execution within a specialized browser runtime.
Evidence: scripts/lib/launch-check.mjs constructs a JavaScript string containing configuration and imports, which is written to a temporary file (entry.mjs) and executed via the ego-browser Node.js environment.
Evidence: Safe practices are used, such as JSON.stringify for variable interpolation into generated code.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from the external YouTube Studio web interface, representing a potential ingestion point for untrusted content.
Ingestion points: scripts/lib/studio-dom.mjs scrapes body text, titles, and element attributes from studio.youtube.com.
Boundary markers: The skill returns results in a structured JSON format to the calling agent.
Capability inventory: The skill has the ability to write to /tmp and execute local shell commands via the platform's browser utility.
Sanitization: Content is filtered using strict regular expressions in scripts/lib/pure.mjs to extract only valid video IDs and processing status labels.

youtube-video