The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process YouTube transcripts, which are external, untrusted data sources.
Ingestion points: Data is fetched from YouTube's infrastructure via the fetchTranscript function in youtube-transcript.js and the youtube-transcript-plus package in transcript.js.
Boundary markers: While output is structured (SRT, JSON, etc.), there are no explicit instructions or wrappers to prevent the agent from following potential commands embedded in the video subtitles.
Capability inventory: The skill possesses network access to YouTube and file-writing capabilities for caching and output generation.
Sanitization: The script includes robust XML unescaping and text normalization logic, though this handles formatting rather than malicious instructions.
[COMMAND_EXECUTION]: The skill provides a CLI wrapper for Node.js scripts (youtube-transcript.js and transcript.js) intended to be executed by the agent to perform its primary function.
[EXTERNAL_DOWNLOADS]: The skill relies on the youtube-transcript-plus dependency listed in package.json, which is a standard library for this functionality.

youtube-transcript