The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The troubleshooting section in SKILL.md provides the official installation command for the uv dependency manager: curl -LsSf https://astral.sh/uv/install.sh | sh. This pattern downloads and executes a shell script from a remote server.
[EXTERNAL_DOWNLOADS]: The skill recommends installing the uv tool from https://astral.sh/uv/install.sh if it is not already present on the system.
[COMMAND_EXECUTION]: The skill provides a Python script (scripts/youtube_api.py) meant to be executed via the command line (uv run youtube_api.py) to perform search, video, channel, and comment retrieval tasks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from YouTube, which could contain malicious instructions designed to influence agent behavior.
Ingestion points: Video transcripts, descriptions, and user comments fetched by the search, video, comments, and transcript subcommands in scripts/youtube_api.py.
Boundary markers: No specific delimiters or instructions to treat the external content as untrusted are included in the skill's instructions.
Capability inventory: The youtube_api.py script performs network operations (API calls) and accesses local environment configuration files.
Sanitization: The script returns raw structured data (JSON) containing external text without specific sanitization for prompt injection patterns.

youtube-data