The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess style execution for its internal scripts (generate_title_variants.py and update_sheet.py) as part of its primary workflow. These scripts interact with local JSON files and the Google Sheets API.
[CREDENTIALS_SAFE]: The skill correctly manages sensitive credentials. ANTHROPIC_API_KEY is loaded from a .env file, and Google OAuth credentials use a token.json or credentials.json flow, which are standard and secure practices for these APIs.
[EXTERNAL_DOWNLOADS]: The skill downloads data from the Google Sheets API (sheets.googleapis.com) and sends prompts to the Anthropic API. These are well-known services required for the skill's stated functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests YouTube titles from external sources (JSON files or Google Sheets). While this is a potential ingestion point for untrusted data, the skill uses these titles as literal strings for analysis and does not execute them as commands. The generated output is restricted to JSON arrays of strings.

title-variants