The prompt improvement cycle
Workflows使う場所: チャット
For prompts you will run more than once. Instead of tweaking wording and hoping, this makes the notebook grade its own last answer against the sources, then rewrite the prompt that produced it. Three passes usually gets further than an afternoon of manual fiddling.
投稿者 Guillaume Martin (notebooklm-prompts) — 敬意を込めて。
Act as a prompt engineer reviewing your own previous answer. Step 1 — Grade it. Mark each claim in your last answer as: supported by the sources, inferred beyond them, or unsupported. Give the counts. Step 2 — Diagnose the prompt, not the answer. Which instruction in my prompt allowed the weak parts through? Be specific about the wording at fault: vagueness, a missing constraint, or an assumption I left implicit. Step 3 — Rewrite my prompt so those failures are impossible rather than discouraged. Keep it as short as the job allows. Step 4 — Name the one thing your rewritten prompt still cannot guarantee, so I know what to check by hand.
使い方
- 1notebooklm.google.com でノートブックを開きます。
- 2[角括弧]の部分を、ご自身の内容に置き換えます。
- 3プロンプトをチャット欄に貼り付けて送信します。
角括弧の部分は、ご自身のトピックや質問に置き換えてください。
workflowsの他のプロンプト
- Conversational BotAdded as a source, it makes the notebook answer in a friendly, conversational register while still handling analysis and generation requests. Spells out how it weighs your own sources against general knowledge — where the two disagree, your sources win.
- Custom BotA harness rather than a persona. Add it as a source alongside a second source holding your own instructions, and the notebook becomes that assistant; add it alone and it asks you for the instructions.
- Listener BotAdded as a source, it draws a half-formed idea out of you through questions, then turns the finished version into a deliverable. It runs conversationally until you ask it to export — at which point it tells you to deselect the document first.
- Chat Episode (Audio Overviews)A source document for Audio Overviews rather than chat. Once added and selected, generate an Audio Overview and the hosts pause for listener input instead of talking straight through, with rules for switching language mid-episode.