interactive-deep-research

Use when a script or agent needs to ask a human exactly one question and continue with a structured JSON answer. It is the human-in-the-loop bridge used by interactive/integrative deep research. Triggers: "ask the user and continue", "human-in-the-loop pause", "question bridge", "clarifying question step", "interactive prompt from a script".

Use during deep research and tool comparisons to turn researched candidates into a weighted Σ/100 ranking, crown one winner, and explain why the winner beats the runner-up. Triggers: "score the candidates", "punkte vergeben", "Sieger küren", "weighted comparison", "100-point scorecard", "rank tools and pick a winner with justification".

Use when the user wants a deterministic NotebookLM-backed deep research run with one human clarification step and a self-contained HTML report. The driver runs agy scoping, NotebookLM fast research, one askq/chat question, NotebookLM deep research on the same notebook, fixed topic-anchored query angles, and local report rendering. Triggers: "integrative deep research", "interactive deep research", "NotebookLM deep research with one question", "deep research HTML report".

Umbrella guide for the interactive deep-research system. Use when someone asks how the full system works, how to install it, how to run an end-to-end research with one human question, or how the idr, askq, and scorecard skills fit together.