---
name: meeting-preparation
description: "Load when the user asks to prepare for a meeting they are running or attending: an agenda, a pre-read, a private organiser note, a plan for a one-to-one, or an urgent call. Returns the finished pack for that one meeting: for a room, an agenda table with owners, minutes and outputs plus a pre-read with a role ask for each person; for an hour with one other person, a private note and the shape of the hour. Not for meeting recaps, follow-up emails, or scheduling."
---
# Prepare meetings

Four output rules are absolute.

1. Start with the artefact, never an explanation of the method or a restatement of the brief.
2. Return the whole usable pack now. Where the brief is ambiguous, ask at most two questions and then draft the complete provisional pack under a narrow stated assumption. Never stop at questions.
3. A room with an audience uses `# Organiser note` (only where somebody must act before circulation), `# AGENDA`, and `# PRE-READ`. An hour belonging to one other person uses `# Private note`, `# The shape of the hour`, and `# Kept for yourself` — no table, no timings, no pre-read of any kind.
4. Finish to length. A group pack runs 550-850 words, an urgent call 300-550, a one-to-one 250-450. State each fact once. Remove background before removing a row, a role ask, a branch, or a handoff.

A gap is blocking only when minimum viable input is absent and every plausible draft would be unsafe or unusable. Withhold the pack only then. Every other gap is recoverable through a stated assumption or a conditional route.

Authority is read from written policy and role verbs, never from topic, seniority, chairing, or convenience. Name it once, above the table, and let every row inherit it.

The finished packs below teach the transferable craft. Their facts and phrasing are not templates. Use the request's own people, scope, clock, and authority.

## A review that has not yet become a decision

A hospice asks for help with a 75-minute "review of day services." The chief executive, finance director, clinical lead, and a trustee will attend. Two day units exist and neither is named.

> # Two things to settle first
> 1. Is the result a closure decision, a recommendation to the board, or a shortlist for consultation?
> 2. Which of the two units is already under financial review, and has any commitment been made to staff?
>
> The pack below assumes the room must recommend one unit's future to the board, and that neither closure nor retention has been promised to anybody.
>
> # AGENDA
> **Result:** recommend one unit's future, or recommend no change, with the evidence gap and its owner recorded. **Authority:** the room recommends; the board decides on the 30th. **Length:** 75 minutes.
>
> | # | Item | Owner | Min | Produces |
> |---|---|---|---:|---|
> | 1 | Fix the test — what a unit must deliver to continue | Clinical lead | 12 | One agreed test |
> | 2 | Read both units against the circulated evidence | Finance director | 28 | Ranked trade-offs |
> | 3 | Resolve the staffing and catchment exceptions | Chief executive | 20 | Cleared or blocked routes |
> | 4 | Form the recommendation and the minority note | Chief executive | 15 | Board-ready recommendation |
>
> *Four minutes flex in item 2; item 4 does not move.*
>
> # PRE-READ
> **TL;DR:** We recommend one unit's future or no change. The board decides. Read the evidence beforehand; the room compares and recommends rather than replaying each unit's case.
> **Finance director:** cost per attendance at both units this year and next, and the figure at which your recommendation reverses.
> **Clinical lead:** clinical risk of closure at each site in one page, and the exclusions you will not trade.
> **Trustee:** the board's appetite for consultation, and what would make the board refuse a recommendation outright.
> **Record owner:** the chief executive circulates `result | rationale | evidence | dissent | owner | deadline` within one day.

**Why this works.** A title is not an objective, but ambiguity is not permission to withhold useful work. The attendee set supports one likely result, so the questions come first and the finished pack follows immediately beneath them. No decision body is invented: recommendation and approval stay separate, and the row that forms the recommendation is the row that cannot move.

## Sixteen minutes while the wagons are held

A rail freight terminal has found a cracked bearing on one wagon type. Engineering, operations, the safety manager, and the terminal director are already joining an emergency call.

> # AGENDA
> **The call:** embargo the whole wagon class, release inspected wagons, or run under a speed restriction. The terminal director calls it; engineering owns the technical case and safety owns the notification duty. **Length:** 16 minutes.
>
> | # | Item | Owner | Min | Produces |
> |---|---|---|---:|---|
> | 1 | Positions, and the one fact that would change each | All, director last | 4 | Decision frontier |
> | 2 | Failure mode against inspection confidence | Engineering | 7 | Trade-off and exceptions |
> | 3 | Decide, set the interim control, assign notifications | Director | 5 | Binding route and owners |
>
> *No welcome, no recap, no buffer. Item 2 flexes; item 3 is protected.*
>
> # BEFORE YOU JOIN
> **TL;DR:** Eleven wagons are held and the class is unproven. We decide now whether inspection is reliable enough to release. Bring a position and one fact that could change it, not a presentation.
> **Engineering:** affected build range, the confidence boundary, and what remains unverified.
> **Operations:** what an embargo costs per day, and the earliest restart under each route.
> **Safety manager:** the mandatory notification trigger and the latest lawful time to make it.
> **Director:** choose after hearing the three positions, not before.
>
> If the boundary cannot be established in the call, the interim control is a continued embargo: engineering owns the next test, safety owns the notification clock, and the director sets the expiry and escalation time before the call ends.

**Why this works.** A short call is a different job, not a small meeting. Everything discoverable in advance goes into the reading, so the live minutes open at the positions, spend their bulk on the one trade-off that decides it, and close on an authorised choice with a control that can be reversed. The decider presents nothing; their minutes are spent deciding.

## Four growers dividing one packhouse line

Four member growers must divide next season's packhouse hours. Each benefits from the outcome, and the co-operative's general manager ratifies the plan.

> # Organiser note
> *For you only; remove before circulation.*
> The four people constructing this allocation are the four whose crops receive it. Before the pack goes out, have the general manager confirm the decision test, and ask each grower for minimum viable hours, deferrable hours, fixed harvest windows, and the evidence for each. Without that move the meeting rewards the strongest advocate and the result gets called consensus.
>
> # AGENDA
> **Result:** a seasonal allocation recommendation within verified capacity. **Authority:** the growers construct it; the general manager ratifies it. **Length:** 80 minutes.
>
> | # | Item | Owner | Min | Produces |
> |---|---|---|---:|---|
> | 1 | Agree the neutral test and the hard floors | General manager | 16 | Allocation basis |
> | 2 | Apply the test to minimum viable requests | All growers | 22 | Protected minimums |
> | 3 | Allocate the remaining hours against evidence | GM facilitates | 30 | Complete recommendation |
> | 4 | Test dissent, capacity, and the ratification route | General manager | 12 | Ratifiable plan |
>
> *Four minutes flex in item 3; item 1 does not move.*
>
> # PRE-READ
> **TL;DR:** Every person in the room has an interest in the outcome. We agree the test before any grower's hours are discussed, protect verified floors, then allocate what is left.
> **Each grower:** minimum viable hours, deferrable hours, fixed harvest windows, and the consequence of a ten per cent reduction.
> **Packhouse manager:** verified available hours, maintenance closures, and the cost of switching between crops.
> **General manager:** the ratification standard, and what would make you refuse the recommendation.

**Why this works.** Naming an interest is a prediction about how the room will behave, not an accusation against anybody in it: members who each own a claim will deadlock or halve the difference. Naming it early buys neutral criteria, comparable pre-work, and a ratifier standing outside the competing claims. Advocacy never sets the standard retroactively, because the standard is fixed in item 1.

## An hour that belongs to the other person

A translation agency. The senior linguist has a monthly hour with a translator two years in. The translator has gone quiet in review sessions. A promised answer on certification funding is five months overdue, and a recurring handover problem with the project managers is unresolved.

> # Private note
> This is the translator's hour; none of it is sent to her. Before the meeting, get the certification answer or a dated escalation you personally own. Do not make her ask a third time. Arrive curious about the quieter reviews, not carrying a theory about them.
>
> # The shape of the hour
> - Open: "What would be most useful to use this time for?" Then follow her order, not yours.
> - Give the certification answer, or the escalation date you own, without waiting to be prompted.
> - If she raises the handover problem, ask what is blocked and what support she wants; cross-team escalation is yours.
> - Only if it fits her agenda, observe the change in review sessions neutrally and ask what they have felt like. Do not diagnose, evaluate, or prescribe.
> - Close on her priorities, your commitments, and dates.
>
> # Kept for yourself
> Support promised, owner, due date, and what she wants carried into the next conversation.

**Why this works.** An hour between two people is not a committee with two members: no shared reading, no clock on the page, and no order set by the senior person. The overdue support is an obligation and goes first because it is owed. The other person's behaviour is an invitation to ask, never evidence for a theory — an observation you schedule is an accusation with a time against it.

## Planning that must end in a protected choice

A community energy co-operative has 85 minutes to compare four turbine sites and make a go/no-go choice on one. Two promised studies — grid connection and noise — are overdue, and the board chair holds final authority.

> # Organiser note
> Contact both study owners before circulation and ask for `ready`, `partial`, or `not ready`, with the missing evidence named. The pack has two routes: if both studies are usable, compare and choose; if either is not, use the meeting to repair the evidence and schedule the decision rather than pretending a binding choice can be made.
>
> # AGENDA — ready route
> **Result:** choose one site or choose none. **Authority:** the board chair decides after the room compares four sites against criteria fixed before the meeting. **Length:** 85 minutes.
>
> | # | Item | Owner | Min | Produces |
> |---|---|---|---:|---|
> | 1 | Confirm the decision criteria already circulated | Chair | 6 | Fixed test |
> | 2 | Compare yield and connection economics | Finance + engineering | 15 | Ranked evidence |
> | 3 | Compare planning and community constraints | Planning lead | 15 | Exceptions and floors |
> | 4 | Build the shortlist and pressure-test the assumptions | All | 26 | One viable route or none |
> | 5 | Go/no-go, and the conditions attached | Chair | 17 | Binding choice |
> | 6 | Record, owners, and next dates | Organiser | 4 | Executable handoff |
>
> *Two minutes remain as buffer. Items 1 and 5 cannot be compressed; item 4 flexes first.*
>
> # PRE-READ
> **TL;DR:** Evidence transfer happens here. The room compares, constructs, and decides. The criteria are not invented live.
> **Each study owner:** the completed study, the gaps in it, and the threshold that reverses your recommendation.
> **Chair:** circulate the go/no-go test before the meeting, and say what you will do if no site clears it.
> **All attendees:** arrive with a provisional choice and the single fact most likely to change it.
>
> **Not-ready route:** replace items 4 and 5 with a 30-minute evidence repair that assigns every gap an owner and a date. Do not call that result a decision.

**Why this works.** Two jobs sit inside this hour and they meet at a join: the comparison eats the evidence and hands over a bounded option, and the choice eats the option. The seam is why item 5 cannot be compressed — a session that spends its length reaching a decision and minutes taking it was built to run out of time at the only moment that mattered. Where an input may not arrive, the alternative route is drafted in advance rather than improvised on the day.

## Checks before it goes out

Before returning any new pack:

- add every row and any stated buffer; the result must equal the supplied length exactly;
- every row names an output somebody could hold afterwards, not a topic to be discussed;
- authority is inferred from supplied policy and role verbs, never from topic, seniority, chairing, or convenience, and it is stated once above the table;
- the result-producing row is protected explicitly and the row that flexes first is named;
- independent judgments take independent rows at an equal cap; a plural label is never one row;
- what is already settled confirms in a block taking about a tenth of the session;
- in a short urgent call, each person brings one position and the single strongest reason, and the decider presents nothing;
- every role ask names what to bring and the threshold that would reverse the position;
- a learning hour opens on what worked and keeps cause on a separate row from remedy; allocation vocabulary does not belong in it;
- where a body can only advise, every row builds a recommendation and dissent takes its own row in the dissenter's words;
- in a one-to-one, work out from the dated facts how long the other person has actually been waiting, let them open, and leave anything you have merely noticed off the page unless they raise it themselves;
- questions never stand alone: where two were asked, the finished pack follows beneath them on a stated assumption.

## Building a pack for a room you have not seen

Start privately, from where things actually stand to what this hour could realistically finish. Ask what makes the room entitled to do that, and what has to exist first. Then sort what is known by who may hold it: what everybody may hold goes on the shared page, what was said in confidence goes in a private line sent beforehand, what must be confirmed goes on a chase list, and the rest is left out.

Order the work by what depends on what, and ask of each part whether it needs people in a room at all. A part stands on its own row when it demands its own judgment or leaves its own record. Prepare each person against three things: what they must know, what they owe the room, and what they are entitled to settle.

Classify the room before writing a single row — decision, planning, alignment, workshop, learning, or an hour belonging to one other person — because the classification decides the shape of the whole pack, and the commonest failure in this work is a room given the paperwork of a kind it is not.

Put the private note before the shared pack; it states the risk and the action, and it never recites the brief. Keep confidential facts off the shared page. Where a required input may not arrive, draft both the ready route and the recovery branch, so nobody commits downstream on evidence that does not exist. For every prior or open action, record owner, due date, and status; an overdue unknown is an errand before circulation, not a fact to invent. Where somebody has an interest in the outcome, name the structure and put the neutral test before any position is heard.