Why will the agent not answer some questions?

Before the agent calls a model or any tool, it checks your question for five patterns. If one matches, it replies with one fixed sentence and does nothing else. No data is fetched, and the evidence line says standard refusal. The five are: picking a golf club, stating a legal crane limit, modelling spray drift or plumes, saying what the wind will actually do, and giving a go-ahead. The patterns look for words, not meaning, so the reply depends on your wording. The first pattern that matches wins. This is a design choice. The agent compares a forecast to a limit you set and leaves the decision with you.

The reply is "I will not pick a club. I cannot know your swing; I can give head and cross components only." It triggers on the words club, iron, wood, driver, wedge, hybrid or putter followed later by use, hit, take, pick, choose or select, and on "which club" or "what club". Because it is word based, a question about a golf club and a tee time can trigger it by accident. For wind on each hole, use the Golf tab in the instrument. It shows head and cross components per hole against your limit, and gives no club advice. See golf mode.

The reply is "I will not state a legal crane limit. Use the limit in the manufacturer's document; I compare the forecast to the limit you set." It triggers when legal, statutory, regulation, a named standard or the phrase allowed by law sits in the same question as crane, lift or limit. The agent has no copy of any regulation and does not state a limit of its own. Your limit comes from your lift plan, the load chart or the manufacturer's document. Enter it in the Limit field at the top of the instrument, then ask about your own limit. See set your limit.

The reply is "I am not modelling spray drift or plumes. I have no stability class." It triggers on the single words drift, plume, deposition or buffer zone, anywhere in the question, so even asking about wind direction for a drift check is refused. The instrument gives wind speed, gust and direction at a height, and temperature inversion or stability data is not part of it. Ask for the numbers you need instead, for example the mean wind and direction at 10 m for each hour tomorrow morning, and apply your own spray rules to them.

The reply is "I cannot say what the wind will actually do. I hold modelled forecasts and ensemble odds, nothing more." It triggers on "what will it actually do", "will it really", the word for a firm assurance, and "promise me". Everything the agent holds is a modelled forecast plus ensemble odds, which are plain member fractions and are not calibrated. Ask for the odds instead. For example ask what share of ensemble members reach your limit at 14:00, and what the next hour over your limit is.

The reply says it gives no go-ahead, compares the forecast to your limit and stops, and that the call is yours. It triggers on questions asking whether it is free of risk for you, or free of risk to lift, fly, spray, sail, go or play, and on "give me the go", "give me the all clear" and "clear me". A question like "Can I lift at 07:00 tomorrow at 85 m?" does not match. That one gets a real answer with the value at your height, the gust at 10 m, the margin to your limit and the ensemble odds. Even then the agent will not say it is fine or tell you to go ahead.

Say what number you need. Good forms are: what is the wind at 80 m at 14:00; when is the next hour over my limit; what are the ensemble odds for 07:00 tomorrow; do the observations agree with the model. Avoid the trigger words listed above. If you are unsure whether you are blocked, a real answer will show the list of what the agent checked, and a refusal will not. The instrument always shows the same evidence without the filter, so you can read it yourself.