The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which creates a vulnerability to indirect prompt injection attacks.
Ingestion points: The scripts/get_transcript.py script fetches transcripts directly from YouTube via the youtube-transcript-api library.
Boundary markers: Absent. The instructions in SKILL.md do not provide delimiters or warnings to the agent to ignore any instructions found within the fetched transcript text.
Capability inventory: The skill allows the agent to execute shell commands (uv run) and write data to the filesystem as described in the Output section of SKILL.md ("save the transcript to a specific file").
Sanitization: Absent. The transcript content is printed to stdout and ingested into the agent's context without filtering or escaping.

youtube-transcript