Decide which of 2 tasks an agent runs unattended

You have 2 tasks this week. Task A: a small new CLI tool with argument parsing, 3 subcommands, tests and a README. Task B: adding a paused state to the billing state machine, threaded through 9 call sites with subtle ordering rules. A colleague wants to start an autonomous agent on both and leave it alone.

Which task can run unattended, and how should your supervision differ between them?