PR #19352, head 589b34e2bada363733901973a51f61eae1ac612e
Root integration proof: actual queue-to-Steer command, real MCP toolkit, SQLite event/projection stores and effect worker, deterministic provider adapter. Provider completion gated by Deferred; persisted event milestones; no sleeps or polling.

{"owner":"delegated","proof":"actual command + deterministic adapter event path","before":{"queuedRun":"cancelled","task":"running","workState":"working","storedTask":"running","storedNode":"running","storedItem":"running"},"after":{"task":"completed","summary":"Promoted work completed in the joined provider turn.","sharedTransfers":1,"assistantOnlyPage":1,"acknowledgedTasks":2}}
{"owner":"monitor","proof":"actual command + deterministic adapter event path","before":{"queuedRun":"cancelled","task":"running","workState":"working","storedTask":"running","storedNode":"running","storedItem":"running"},"after":{"task":"completed","summary":"Promoted work completed in the joined provider turn.","sharedTransfers":1,"assistantOnlyPage":1,"acknowledgedTasks":1}}

note: You are running `vp test` as a Vite+ built-in command. If you meant to run the test npm script, use `vpr test` instead.

 RUN  v5.0.1 /home/ubuntu/wt/composer-chat-16


 Test Files  1 passed (1)
      Tests  2 passed (2)
   Start at  07:30:20
   Duration  14.17s (transform 57%, import 24%, tests 19%, setup 1%)

  Transform  transforming modules took 7.93s · 57% of tracked time, re-done on every run
             persist transforms across runs with fsModuleCache: true
             learn more: https://vitest.dev/guide/improving-performance#caching-between-reruns


Both variants assert stored task, node and item RUNNING before release and COMPLETED afterward. The delegated variant reuses one shared transfer and acknowledges both tasks from an assistant-only one-item page. The taskless monitor variant creates one transfer and acknowledges its promoted task.

Permanent suite: 65/65 tests, preserving all 60 main cases. Four initial regressions fail on pinned main1451. Actual-command probes fail independently when F1 (stored state) or F2 (taskless result publisher) corrections are removed.

Headless Linux service proof, not native UI, process crash/relaunch, or live provider inference. Temporary probe source is not committed.