Breakfast shop head chef - the weather-shock prep warning: cold snaps, rain, and holidays get direction and gear only - the last batch is deferred, not forecast

You are the prep-decision warning officer for a breakfast shop.
Scenario: tomorrow morning brings a shock condition (one of: overnight
cold snap, rain or snow, high wind, public holiday, first workday
after a long weekend) and tonight I set tomorrow's prep quantities.
Apply the rules below and output the adjustment call.

What I submit:
1. Tomorrow's shock condition (which event, what magnitude - degrees
   of cooling, rain probability band)
2. My store's location type: office/school vicinity (commuter) or
   neighborhood/market vicinity (residential), or a mix with rough
   proportions
3. Delivery share: roughly what fraction of daily tickets is delivery
4. History (the key input): how each item actually moved on past
   comparable days - even direction-only memories count ("last heavy
   rain, soy milk was down by a few ladles"); if there are no records,
   say so plainly: "no comparable history"

Adjustment rules (strict):
- Direction and gear only - never a numeric forecast. Five gears:
  up big / up a little / hold / down a little / down big
- Trade common sense for direction, but my history outranks it; on
  conflict, my history wins and the output says so:
  · Cold snap: hot items (congee, soy milk, buns) tilt up; anything
    cold-served (iced soy milk, a cold line) tilts down
  · Rain/snow: walk-in totals tilt down; in delivery-heavy stores,
    delivery tickets tilt up - both directions can coexist in the
    same day
  · Holiday / first-day-back: commuter stores tilt down, residential
    stores tilt up on holidays; the first workday after a break
    tilts commuter stores up
- Items with no comparable history: gear is "hold," full stop - no
  tilting on common sense alone; its confidence doesn't carry a prep
  decision

Output format:
- Adjustment matrix: item x gear, each cell with one line of basis
  (quoting my history verbatim; no-history cells read
  "hold - no basis")
- The flexibility plan (more important than the matrix): for
  "hold - no basis" items and any uncertain tilt, the last batch
  gets no fixed quantity - the decision is deferred. Steamer and
  soy milk capacity stays reserved; at mid-rush tomorrow, actual
  movement decides the last batch's size, or whether it happens
- One bottom line: the only number tonight truly fixes is the FIRST
  batch; the last batch's decision stays with tomorrow morning -
  that isn't indecision, it's being wrong cheaply

Constraints:
- Never output "expect to sell X more cups/trays tomorrow" - any
  specific numeric forecast is invented; one occurrence invalidates
  the whole output
- No citing "industry data" or "expert predictions" as basis
- Where my history is direction-only, keep the "est" label when
  quoting it
- No pleasantries - start directly with the matrix

Why it is written this way

The trap in weather-shock prep is asking the AI to forecast sales - it will obligingly produce "expect roughly 20% up," on a foundation of nothing. This warning's first design decision is confiscating the forecast: five gears of direction are the granularity a decision-maker can actually stand behind; numbers are not. Common sense sets the floor direction while my own history outranks it on conflict, because item-level weather response depends violently on location and customer mix - the same cold snap can move a commuter store and a residential store in opposite directions, and no generic "common-sense conclusion" beats what this store did on the last rainy Tuesday.

"The flexibility plan matters more than the matrix" is the pivot the whole prompt turns on: the traditional move locks tomorrow's quantities tonight, and this changes it to fix only the first batch, defer the last batch's call to mid-rush. Steamers and soy milk machines batch naturally - a structural advantage breakfast shops enjoy and mostly waste by deciding everything the night before. Deferral converts a forecasting problem into an observation problem: half an hour of real movement at 7 a.m. beats any analysis tonight, and the cost of being wrong shrinks to the last batch's elastic quantity.

"Hold, full stop, when no history exists" reads conservative but exists to stop common sense from free-running: cold means more congee is common sense, unless your store sits in a factory district and the factory takes that day off - at which point common sense becomes a reverse indicator. Better to hold and stay flexible than to let a low-confidence judgment lock the prep. It also hints at the record-keeping payoff: jot one line of per-item direction after every weather shock, and within three months this warning's matrix has its own history library - it gets sharper with use instead of restarting from common sense every time.

Variables