Meeting scheduling
Finds an hour when three people are free and books it. Inside working hours it books on its own; outside them it asks.
- Written by our coding agent
- 4 automatic tests
The job
Read a meeting request and each attendee’s calendar, find a one-hour slot when everyone is free between 9:00 and 17:00, and book it.
The agent gets this description and the permissions on the right, nothing else. It never sees how “done” will be checked.
How it was built
Written from a task description
One attempt, kept unchanged. It worked out the right cut-off on its own: 16:00, since a meeting starting then ends at 17:00.
Review
Review found two weak spots, both from limits capsule couldn’t yet express. Fixing them added number ranges and a done-check for “inside a range” to capsule itself.
Automatic tests
Each test plays out a run with the agent's moves written in advance, instead of asking an AI model, and checks that capsule allows, stops or refuses each one correctly. They run every time capsule's code changes.
- Normal run
- 14:00, the one slot all three are free, is booked without asking. Done.
- After hours
- 18:00 waits for you, is allowed, and is booked.
- Too early
- 7:00 waits for you and is denied.
- Impossible time
- 25:00 is refused outright. Nobody is asked.
What’s proved
These hold for every run, not just the ones above. Each is proved, and the proof is checked by computer.
- A number outside its allowed range is never accepted.
- Combining permissions never turns “ask me first” into “go ahead.”
What this doesn’t show
- This job hasn’t had a real run with an AI model yet. Its tests run on every code change.