The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using uv run to trigger the Python transcript fetcher and markdownlint to validate the generated markdown notes.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill uses the youtube-transcript-api library to fetch data. This is a well-known library used for its intended purpose within the script.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (YouTube transcripts) and includes it in the agent's context, creating a surface where malicious instructions hidden in video subtitles could influence the agent.
Ingestion points: YouTube transcripts are fetched via scripts/fetch_transcript.py.
Boundary markers: Absent; transcript content is placed directly into markdown notes without delimiters or 'ignore' instructions.
Capability inventory: File system writing (creating notes) and command execution (uv run, markdownlint).
Sanitization: Absent; the transcript text is printed directly to stdout and then saved to files without validation or escaping.

youtube-ingestion