project-schedulelisted
Install: claude install-skill teodstr/workbench
# Project schedule
People underestimate how long their own projects will take, even when they know that their past projects ran late: Buehler, Griffin and Ross called it the planning fallacy. This skill builds a schedule that corrects for it.
## When to use
- After `work-breakdown`, once tasks, owners and estimates exist.
- To check whether a deadline is realistic, or to work backward from a fixed date.
- Not for the cost side: use `project-budget`.
## Inputs
1. The task list with estimates and dependencies (`work-breakdown`).
2. Fixed dates: deadline, holidays, availability of each person, delivery dates.
3. Past projects of the same kind, if any, and how long they really took.
## Method
1. **Order the tasks** by their dependencies and the availability of each person; nobody works on two full-time tasks at once.
2. **Find the critical path**: the chain of dependent tasks that sets the end date. Any delay on it delays the whole project.
3. **Check against reality** (Flyvbjerg's reference class forecasting): how long did similar projects take, and how far did they overrun? Adjust the plan toward that outside view, not toward the best case.
4. **Add buffers** where the risk is: one shared buffer at the end of the critical path and before each fixed external date, rather than padding every task.
5. **Set milestones**: a few dated checkpoints that anyone can verify (permit received, walls painted, beta tested).
6. **Work backward** when the date is fixed: start from the de