The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Python subprocess module to execute external binaries such as yt-dlp, ffmpeg, ffprobe, and whisper-cli. All command invocations in scripts/fetch_transcript.py are performed using list-based arguments rather than shell strings, which effectively mitigates the risk of shell injection attacks from malicious URLs or file paths.
[EXTERNAL_DOWNLOADS]: The skill utilizes yt-dlp to fetch metadata and caption files from YouTube. These network requests target a well-known service and are strictly necessary for the skill's primary function of transcribing online content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external video captions and audio files, which represents a potential attack surface for indirect prompt injection.
Ingestion points: Data is ingested from YouTube's caption endpoints and local media files via the fetch_transcript.py script.
Boundary markers: The skill produces transcript.json and metadata.json files; while these files provide structured data, they do not utilize specific delimiters to isolate the transcribed text for the consuming agent.
Capability inventory: The skill possesses capabilities for network access (via yt-dlp), file system writes (to the designated --out directory), and shell command execution.
Sanitization: The skill implements a comprehensive deterministic quality gate that assesses transcripts for repetition, timing coverage, and character density. This automated vetting process helps detect and flag potentially malicious or malformed content before it is processed by the agent.

video-transcribe