數據工程
Research Pipeline — Token-Efficient Deep Research
Token-efficient research pipeline combining NotebookLM (free analysis) with sub-agent
validation and Obsidian vault storage. Reduces token usage by ~80% compared to /deep-research
by offloading heavy analysis to Google's servers. Use when user says "research pipeline",
"research this efficiently", "deep research but cheaper", "analyze and validate",
or wants thorough research without burning excessive tokens. Also trigger when user has
multiple YouTube videos, PDFs, or web sources to analyze comprehensively.
Prefer this over /deep-research for most research tasks.