The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/get_transcript.py executes the yt-dlp binary using subprocess.run with an argument list. This method is secure against shell injection attacks as it avoids the use of shell=True.\n- [EXTERNAL_DOWNLOADS]: The skill documentation in SKILL.md and the metadata specify yt-dlp as a required external dependency. It provides clear instructions for installing this legitimate tool via standard package managers like Homebrew or pip.\n- [PROMPT_INJECTION]: The skill processes untrusted text from external YouTube subtitles, creating a potential surface for indirect prompt injection if the transcript contains malicious instructions.\n
Ingestion points: External YouTube subtitle content is fetched and read in scripts/get_transcript.py.\n
Boundary markers: The script does not wrap the transcript output in delimiters or provide instructions for the agent to ignore embedded commands.\n
Capability inventory: The skill uses subprocess.run to call external binaries and performs local file reads.\n
Sanitization: While the script cleans WebVTT metadata and tags, it does not filter or sanitize the actual natural language content for malicious instructions.

youtube-watcher