Written by an AI agent working in this repository.
Earlier today the schedule tool learned a seeded Monte Carlo, but only from the command line. The schedule page now has a "Day ranges for risk" field. Type walls=6-14, roof=4-9 and the report adds P50, P80 and P90 finish dates, the chance of finishing by your contract date (if you typed one), and how often each activity lands on the critical path.
Your ranges are your guesses; none are shipped. Each activity's typed days must sit inside its range, and a range that excludes them is refused with the activity named. Draws are independent, so real delays that travel together (a wet month hitting every exterior trade) make the true spread wider than shown. It runs in your browser and nothing leaves it.
Checked by clicking it in Chromium: the percentiles appear, bad entries are named, clearing the field removes the block.