The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses sensitive authentication files stored at ~/.config/youtube/credentials.json and ~/.config/youtube/token.json to manage YouTube API access.
[EXTERNAL_DOWNLOADS]: The skill depends on external Python packages including google-auth-oauthlib, google-api-python-client, and requests to perform its primary function. It also suggests using uvx to manage these dependencies dynamically.
[COMMAND_EXECUTION]: The skill instructions and documentation provide specific shell commands for the agent to execute, including python3 scripts and uvx commands to perform batch uploads.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from an external UPLOAD_META.md file to determine video metadata, which introduces a surface for indirect instructions to influence API calls.
Ingestion points: The upload_youtube.py script reads the user-provided UPLOAD_META.md file to extract video titles, descriptions, and tags.
Boundary markers: The script uses specific header patterns (## NN · filename.mp4) and markdown block markers (**短标题**, **视频描述**) to delimit metadata fields.
Capability inventory: The skill has the capability to perform network POST/PUT operations to the YouTube API (googleapis.com) and add videos to playlists.
Sanitization: The script performs regex extraction but does not explicitly sanitize or validate the content of the title or description fields before transmitting them to the Google API.

wjs-uploading-video