Prompt · Analysis
Meeting notes prompt
Turning an hour of transcript into the four things that actually have to happen, without inventing an owner for the task nobody claimed.
Extract the outcomes from this meeting.
<transcript>
{{transcript}}
</transcript>
Produce three sections. Omit any section that would be empty.
**Decisions** — things settled in the meeting. For each: what was decided, and what it replaces if it reverses an earlier position.
**Actions** — things someone will do. For each: the action, the owner, and the date if one was said. If no owner was named, write "unassigned" — do not infer one from who talked most or who raised the topic.
**Open** — questions raised and not resolved, and disagreements that were not settled. This section matters more than the other two; it is the part that is lost by the next meeting.
Rules:
- Only what was actually said. Discussion of a possibility is not a decision.
- Distinguish "we should" from "I will". The first is an opinion, the second is an action.
- Quote the phrase for anything ambiguous, so the reader can check.
- If the transcript contains no decisions or actions, say so. Many meetings do not produce any.What to fill in
{{transcript}}- Raw transcript with speaker labels if you have them — ownership attribution depends almost entirely on knowing who said what.
Why it is written this way
Every rule in the prompt is there because of a specific failure it prevents. Knowing which is which is what lets you adapt it instead of only pasting it.
Unassigned is a permitted answer
The most common failure of meeting-notes prompts is inventing owners. Someone raised a topic, so they get the action, and now a person who never agreed to anything has a task with their name on it. "Unassigned" is accurate and actionable — it tells you what to settle.
"We should" and "I will" are separated
This distinction is where most phantom action items come from. Meetings are full of agreement in principle, and only some of it becomes commitment.
Open questions rank highest
Decisions and actions get remembered because someone owns them. Unresolved disagreements evaporate between meetings and resurface a month later as a surprise.
Empty sections are dropped
A heading with nothing under it reads as an omission. Removing it is honest about a meeting that decided nothing.
The version most people write, and what it costs
Summarise this meeting and list action items:
{{transcript}}It produces six action items, three of which nobody agreed to, each assigned to whoever mentioned the topic. The two genuine disagreements are smoothed into a summary sentence that implies consensus. The notes circulate, nobody objects because objecting is awkward, and the wrong record becomes the record.
What it still gets wrong
- Transcript quality dominates. Without speaker labels, ownership is guesswork the prompt is instructed not to do — so most actions come back unassigned.
- It cannot hear tone. Sarcastic agreement and reluctant agreement look identical in text.
- Long meetings need chunking, and decisions that were revisited across chunks can appear twice with different outcomes.
Check it before you ship it
- LLM Token CounterReal BPE tokenization, not characters ÷ 4. Shows which counts are exact and which are estimates.
- Text Splitter for RAG and EmbeddingsToken-accurate chunking with configurable overlap. Respects sentence and paragraph boundaries.
- LLM API Cost CalculatorRequests per month in, dollars out. Input, cached input and output priced separately.
To build one of these from scratch for a task not covered here, the prompt generator assembles the same structure — delimiters, output contract, edge cases — from an expert-authored blueprint, and the prompt review checklist is what to run over the result.