notebooklm-teacher-workflow

name	notebooklm-teacher-workflow
description	Use when Codex should consult a fixed NotebookLM notebook before answering paper-backed questions, preserve NotebookLM citations, report no-solution when sources do not support an answer, and then combine the teacher's answer with local coding or analysis work.

NotebookLM Teacher Workflow

Use this skill when a task depends on paper-backed knowledge stored in a fixed NotebookLM notebook.

This skill is notebook-agnostic. Bind it to your own teacher notebook by:

• passing --notebook <alias-or-id> to the bundled script, or
• setting NLM_TEACHER_NOTEBOOK=<alias-or-id> in the environment

What This Skill Does

1. Identifies which parts of the task need source-backed knowledge.
2. Queries the fixed NotebookLM teacher before answering from memory.
3. Preserves NotebookLM citation markers such as [1][2].
4. Surfaces no_solution when the notebook cannot support an answer.
5. Uses the teacher result to drive local coding, experiments, file search, or summarization.

Query Workflow

Use the bundled wrapper:

uv run ~/.codex/skills/notebooklm-teacher-workflow/scripts/ask_teacher.py \
  --notebook "<NOTEBOOK_ALIAS_OR_ID>" \
  --question "What does the notebook say about this method?"

Or set a default notebook once:

export NLM_TEACHER_NOTEBOOK="<NOTEBOOK_ALIAS_OR_ID>"
uv run ~/.codex/skills/notebooklm-teacher-workflow/scripts/ask_teacher.py \
  --question "What does the notebook say about this method?"

Reuse --conversation-id only for follow-up questions on the same thread.

Required Rules

• Ask the teacher before answering source-backed questions from memory.
• Keep NotebookLM citation markers unchanged in your own response.
• If the wrapper returns status=no_solution, say the teacher does not support the answer and do not fabricate details.
• Do not write notes, code, or experiment results back into NotebookLM.

Output Contract

For research-engineering tasks, prefer this delivery skeleton:

## Teacher Says
- source-backed findings with original citations

## What I Did
- code, commands, experiments, or local findings

## Conclusion
- final answer grounded in teacher citations when relevant

## Not Covered by the Teacher
- unresolved or weakly supported points

Resources

• scripts/ask_teacher.py: normalized wrapper around nlm notebook query
• references/workflow.md: behavior and failure conditions
• references/prompt-template.md: reusable NotebookLM persona template