The point of monetary deployment should be balance sheet repair. Only when people's balance sheet's are repaired can normal spending resume and the layoffs stop. The point of an intervention now should not be to get the unemployment rate back down to 4%, but to prevent it from going to 15%. Therefore, the thing to do is to repair balance sheets by printing money. If you only give money to people repairing roads then consumer spending still drops off a cliff, and millions of people still lose their jobs.
1) Must be able to start very soon
2) Must have a clear horizon (i.e. a project with a final deadline, so that it doesn't keep costing money forever)
3) Must be useful
One example it gave of the best way to spend that money would be actually to repair existing roads. That's usually very easy to do right away, provides immediate short term benefits, and reduces your long term costs (unlike building roads, which increases them, as you then have to maintain all these new roads).