The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute a local Python script for transcript extraction. This is a standard and safe way to manage project-specific dependencies without global installation.\n- [EXTERNAL_DOWNLOADS]: It fetches the youtube-transcript-api package from PyPI at runtime using uv. This is a well-known, legitimate package used for the skill's primary purpose.\n- [DATA_EXFILTRATION]: No data exfiltration detected. The skill reads public transcript data and outputs it to the user or a temporary local file (/tmp/youtube-transcript.json).\n- [PROMPT_INJECTION]: The instructions do not contain any patterns attempting to bypass safety filters or override agent constraints. The focus is on structuring YouTube summaries in Korean.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcript data from YouTube. While this is an ingestion surface (Category 8), the skill uses clear delimiters and a structured summary format, and it lacks dangerous capabilities like arbitrary network sends or elevated privilege execution that could be exploited by injected content.

youtube-summarizer