manual_tests.infinite_block_tests

Automating serial execution of infinite-block tests involves delegating work to sub-agents running in strict order with per-test resets. This prevents cross-test interference and ensures deterministic blocking behavior is observable by recording queue status after each scenario.

Verifying non-promise and promise blocking behavior requires running test cases that observe both bypass and blocking outcomes. By checking queue status entries and hook activity after each test, you can confirm correct firing and timing to validate the behavior.

Preventing cross-test interference when running infinite-block scenarios requires executing tests in strict serial order. Performing a mandatory reset after each test and checking the rules queue ensures that previous test states do not affect subsequent blocking behavior.

The Haiku sub-agent model supports infinite-block testing by enforcing a controlled environment with a strict limit of 5 turns. This configuration ensures deterministic sub-agent execution while observing blocking behavior and recording the results in a queue.

Consolidating infinite-block test results into a single artifact requires running all test scenarios and capturing hook activity alongside queue status. After executing the tests in serial order with resets, you return the recorded outcomes as an infinite_block_results artifact.

Limitations of using sub-agents for infinite-block testing include a strict max_turns limit of 5 and forced serial execution. These constraints ensure controlled testing but may restrict complex, multi-turn interactions or parallel test case processing.