The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes transcripts from arbitrary YouTube videos. An attacker could include malicious instructions in the video subtitles designed to hijack the agent's context during the transformation phase.
Ingestion points: Fetches transcript data from external YouTube URLs via the fetch_transcript.py script.
Boundary markers: The skill workflow does not specify the use of delimiters or 'ignore instructions' warnings when processing the fetched transcript.
Capability inventory: The skill uses python3 for execution and has the ability to install packages via pip.
Sanitization: No sanitization or filtering is performed on the transcript text before it is used to generate summaries or blog posts.
[COMMAND_EXECUTION]: The skill requires the execution of a local Python script (scripts/fetch_transcript.py) to function and instructs the agent to run pip install to manage its dependencies.

youtube-content