Verification Contracts

Verification contracts ensure task completion by generating verifiable contracts with specific criteria for automated testing, linting, and type-checking before delegating work to agents. This prevents ambiguity by defining exact success criteria.

A verification contract in agent task management is a defined agreement specifying required checks, such as unit tests or linting, that an agent must pass to prove task completion. It supports verifiable, advisory, and skip contract types.

You can set up automated testing criteria by generating a verification contract from the task description, specifying required tests like e2e tests for login flows or unit tests for components, which the system then executes and records.

Yes, you can skip verification for simple agent tasks by using the 'skip' or 'advisory' contract types. These match lower task complexity levels where strict automated testing, linting, and type-checking are not required.

The best way to manage quality assurance for multiple delegated tasks is using verifiable contracts that manage contract storage and result reporting within the chunk state, ensuring robust task management and automated verification.

Agent delegation lacks clear success criteria when it does not use verification contracts. Generating a contract based on expected outcomes defines exact requirements like automated testing and linting to ensure correct task completion.