✔ JSON migration survives process exit before commit and retains exact legacy backups (3977.4956ms) ✔ process exit after_state rolls back state, event and outbox together (3391.3139ms) ✔ process exit after_event rolls back state, event and outbox together (3285.313ms) ✔ crash recovery at accepted retains execution identity and never replays side effects (2372.3403ms) ✔ crash recovery at queued retains execution identity and never replays side effects (3930.6046ms) ✔ crash recovery at approval_wait retains execution identity and never replays side effects (3370.4176ms) ✔ crash recovery at approved retains execution identity and never replays side effects (3850.0741ms) ✔ crash recovery at effect_unrecorded retains execution identity and never replays side effects (3829.418ms) ✔ crash recovery at result_committed retains execution identity and never replays side effects (2395.8607ms) ✔ crash recovery at before_next_model retains execution identity and never replays side effects (2528.3302ms) ✔ state and event failures publish no success and preserve the last committed memory snapshot (591.7228ms) ✔ tool result artifacts and event cursors retain project ownership across restart (1435.9465ms) ✔ token budgets pause before tools execute and continuation keeps original task and accumulated usage (1465.7178ms) ✔ a time budget interrupts an unresponsive model and persists a resumable run (873.9656ms) ✔ a corrupted large result artifact cannot become a recovered success (940.3009ms) ✔ a stop from the event loop interrupts immediate model replies before later tools (2629.0773ms) ✔ a stop after the first tool preserves a complete interrupted multi-call protocol group (3175.8609ms) ✔ the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands (7337.2425ms) ✔ denial prevents command execution and produces a paired denied result (3011.3365ms) ✔ stopping during approval interrupts pending calls without executing or replaying them (1856.5033ms) ✔ stopping a running fixture command terminates its own process tree and retains uncertain state (2445.1602ms) ✔ command output decodes split UTF-8 and excludes the workbench key from child environment (1430.8124ms) ✔ model-list testing uses a tentative key without changing the saved connection and redacts service errors (6.8615ms) ✔ truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities (3211.0294ms) ✔ request projection keeps exact tool pairs and complete recent user instructions while preserving disk history (1875.46ms) ✔ HTML without tests ends at reviewed file delivery and never offers automatic rechecking (4264.3105ms) ✔ diagnostic answers retain the unfinished task, limits and check result, and enforce read-only tools (5555.3294ms) ✔ review continuation is a system action and adds no testing demand or fake user instruction (3232.9329ms) ✔ a provider cannot execute tests against an explicit no-test task even if it ignores advertised tools (3099.2957ms) ℹ tests 29 ℹ suites 0 ℹ pass 29 ℹ fail 0 ℹ cancelled 0 ℹ skipped 0 ℹ todo 0 ℹ duration_ms 38384.793