The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses spawnSync in scripts/youtube.ts to execute yt-dlp (or uvx/python3 alternatives) for retrieving video metadata and captions. It also passes environment-supplied browser cookie parameters to the process.
[EXTERNAL_DOWNLOADS]: The skill fetches content from YouTube's official domains (www.youtube.com, i.ytimg.com). It also utilizes yt-dlp's capability to load remote components from GitHub via the --remote-components ejs:github flag.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted content from external video sources for subsequent AI processing.
Ingestion points: Untrusted video descriptions and transcript snippets are fetched and stored in scripts/main.ts and scripts/youtube.ts.
Boundary markers: The skill does not employ explicit delimiters or system instructions to distinguish or protect the agent from malicious instructions embedded within the transcript text before it is processed by the speaker identification sub-agent.
Capability inventory: The environment has the capability to execute shell commands (via yt-dlp) and perform file system writes.
Sanitization: While the skill performs basic cleaning such as HTML unescaping and tag stripping, it does not filter for adversarial prompt patterns in the ingested text.

baoyu-youtube-transcript