The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/extract_transcript.py utilizes subprocess.run to call the yt-dlp binary. The call is implemented using an argument list rather than a shell string, which is a security best practice that prevents shell-level command injection.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install yt-dlp via pip. This is a well-known and standard tool for media metadata extraction.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (YouTube video titles, descriptions, and subtitles), which could be leveraged for indirect prompt injection if the resulting Markdown files are subsequently processed by an AI agent. \n
Ingestion points: Data is fetched from YouTube URLs using yt-dlp in scripts/extract_transcript.py. \n
Boundary markers: The output is structured with YAML frontmatter and Markdown headers, but lacks explicit delimiters or instructions to the agent to ignore instructions embedded in the video content. \n
Capability inventory: The skill performs filesystem writes to ~/Brains/brain/ and executes the yt-dlp tool. \n
Sanitization: While filenames are sanitized to remove illegal characters, the content of the transcript and description is passed through without filtering for potential prompt injection patterns.

youtube-transcript