Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Everyday tasks

Pack for the forecast, checked again at pack time

Written October 28, 2026

For a traveler × packing for a trip × a chatbot, ending with a pack list keyed to the forecast for the actual days away — built once when the trip is planned, and re-checked the morning the bag gets packed, because forecasts move.

Before you start

You need the destination, the dates away, and what the days hold: city walking, a hike, a wedding, or all three. The forecast the agent works from comes from Open-Meteo's forecast API, which needs no key and returns up to 16 days of daily figures for a place: high and low temperatures, rain chance, and a weather code for each day. The weather codes are the WMO set: 0 is a clear sky, 1 to 3 run from mainly clear to overcast, 45 and 48 are fog, 51 to 67 are drizzle and rain in increasing strength, 71 to 77 are snowfall, 80 to 82 are rain showers, 85 and 86 are snow showers, and 95 to 99 are thunderstorms, with and without hail. The daily forecast needs a time zone to draw its day boundaries, and the API updates as the underlying weather models refresh — global models every 6 hours, some regional models hourly — so a forecast read a week out is a draft, not a promise.

Pack for the worst day in the window, not the average. An average of 14 degrees across five days hides the one day at 4 degrees with rain, and that day is the one that decides whether the trip is comfortable. The agent's list should name, for each day, the high, the low, the rain chance, and the weather code in plain words — and the pack list should follow from those four numbers, not from the season or the destination's reputation.

What "done" looks like

At the end of one sitting you have a pack list with a line for each day's weather and the item that answers it, a short "if the forecast moves" rule for swaps, and a second check scheduled for pack day. The bag gets packed from the second check, not the first: the first list is the draft, the re-check is the list.

Steps

  1. Give the agent the destination, the dates, and the days' shape. Name the town or city, the first and last dates away, and what each day holds. A day of city walking in rain needs waterproof shoes; a wedding day needs the outfit protected from the same rain; a hike needs layers and a shell. Without the shape of the days, the agent packs for weather in general, which fits no day in particular.
  2. Ask for the daily forecast in plain words, with the numbers. Ask the agent to fetch the forecast for the destination and dates, and to report each day as four figures: high, low, rain chance, and the weather code translated into plain words — "moderate rain," not "code 63." If the trip is more than 16 days out, the forecast does not reach it yet: the agent should say so and build the list from the trip's first 16 days plus a note to re-check, not invent figures for the days beyond the forecast.
  3. Turn the forecast into items, one decision per day. Ask for the pack list grouped by decision, not by suitcase: rain days get the waterproof layer and the shoes; cold mornings get the layer that fits under the jacket; a thunderstorm day gets the indoor backup plan for the outdoor activity. Each item should name the day and the figure that earned it — "rain shell: Tuesday, 80% rain chance, moderate rain." An item with no day and no figure behind it is packing by habit, and habit is how the heavy jacket ends up in a heat wave.
  4. Add the "if the forecast moves" rule. Ask the agent to write the swap rules next to the list: if a dry day turns wet, which item comes out and which goes in; if a cold snap appears, which layer is added; if a heat day appears, which layer stays home. The rules are short — one line each — and they are what makes the pack-day re-check fast instead of a second full repack.
  5. Re-check the morning you pack, and pack from the second list. On pack day, ask the agent for the forecast again for the same dates, and compare it with the first list day by day. Apply the swap rules where the forecast moved, and pack from the new list. A forecast read at booking time and trusted at pack time is a week-old guess; the models refresh several times a day, and the re-check costs one question.
  6. After the trip, keep the list that worked. Note which items were used, which stayed packed, and which day surprised you. The next trip to a similar place starts from that note, not from scratch. One line per trip, kept with the trip notes, is enough.

Check the agent's work

What can go wrong

Limits

Related playbooks

Sources

Sources: Open-Meteo forecast API docs, Open-Meteo.

Back to: Everyday playbooks hub · Also: Plan a weekend trip itinerary on a budget