PR #21493 real-provider verification, 2026-10-28
github/main: 8031991591
PR head: 44bb6f810e7c8e055a7907b54e6aa15e80e7776c
Codex CLI 0.164.0; model gpt-6.1-sol; local CLIProxyAPI; no mocks or fixtures in live runs.
Created /home/user/native-child-check as a real git repository, then added it via the web UI.

MAIN
2026-10-28T05:46:27.949Z OUT thread/read {"threadId":"01a1810b-a9fb-7cb1-a2e0-11051488870e","includeTurns":false}
2026-10-28T05:46:27.967Z IN subAgentActivity {"kind":"started","model":"gpt-6.1-sol","reasoningEffort":"high","agentThreadId":"01a1810b-a9fb-7cb1-a2e0-11051488870e"}
2026-10-28T05:46:28.149Z IN thread/read result {"id":"01a1810b-a9fb-7cb1-a2e0-11051488870e","model":"gpt-6.1-sol","reasoningEffort":"high","parentThreadId":"01a1810b-3b2d-7f51-a857-d05c493615d2","status":{"type":"active","activeFlags":[]}}
2026-10-28T05:47:16.296Z IN agentMessage {"threadId":"01a1810b-a9fb-7cb1-a2e0-11051488870e","text":"CHILD_OK: 323"}
2026-10-28T05:47:17.992Z IN turn/completed {"threadId":"01a1810b-a9fb-7cb1-a2e0-11051488870e","status":"completed"}
2026-10-28T05:47:17.995Z IN subAgentActivity {"kind":"completed","model":null,"reasoningEffort":null,"agentThreadId":"01a1810b-a9fb-7cb1-a2e0-11051488870e"}
2026-10-28T05:47:41.736Z IN agentMessage {"threadId":"01a1810b-3b2d-7f51-a857-d05c493615d2","text":"PARENT_OK\n\n> CHILD_OK: 323"}
2026-10-28T05:47:44.727Z IN turn/completed {"threadId":"01a1810b-3b2d-7f51-a857-d05c493615d2","status":"completed"}

PR
2026-10-28T05:52:46.087Z OUT thread/read {"threadId":"01a18111-6ffc-7ee3-b7a3-7227adf43464","includeTurns":false}
2026-10-28T05:52:46.087Z IN subAgentActivity {"kind":"started","model":"gpt-6.1-sol","reasoningEffort":"high","agentThreadId":"01a18111-6ffc-7ee3-b7a3-7227adf43464"}
2026-10-28T05:52:46.098Z IN thread/read result {"id":"01a18111-6ffc-7ee3-b7a3-7227adf43464","model":"gpt-6.1-sol","reasoningEffort":"high","parentThreadId":"01a18111-5cc1-7dd1-b9e6-c5a5da43a5e4","status":{"type":"idle"}}
2026-10-28T05:52:54.352Z IN agentMessage {"threadId":"01a18111-6ffc-7ee3-b7a3-7227adf43464","text":"CHILD_OK: 323"}
2026-10-28T05:52:54.603Z IN turn/completed {"threadId":"01a18111-6ffc-7ee3-b7a3-7227adf43464","status":"completed"}
2026-10-28T05:52:54.606Z IN subAgentActivity {"kind":"completed","model":null,"reasoningEffort":null,"agentThreadId":"01a18111-6ffc-7ee3-b7a3-7227adf43464"}
2026-10-28T05:52:56.899Z IN agentMessage {"threadId":"01a18111-5cc1-7dd1-b9e6-c5a5da43a5e4","text":"PARENT_OK\n\n> CHILD_OK: 323"}
2026-10-28T05:52:57.311Z IN turn/completed {"threadId":"01a18111-5cc1-7dd1-b9e6-c5a5da43a5e4","status":"completed"}
2026-10-28T05:55:21.526Z IN agentMessage {"threadId":"01a18111-5cc1-7dd1-b9e6-c5a5da43a5e4","text":"FOLLOWUP_OK: 42"}

Browser assertions: PR child displayed GPT-6.1-Sol / High / Completed in 8.0s and CHILD_OK: 323. Reload retained the result and selection. Parent completed PARENT_OK quoting the child. Parent follow-up at Medium returned FOLLOWUP_OK: 42; reopening child still showed High.

Blacksmith: CodexAdapterV2.test.ts: 86 passed. Three known-model/missing-effort regression cases failed against the PR base adapter (expected metadataRequests=2, actual=0).

LIMITATION: main also displayed High correctly. Live gpt-6.1-sol emitted subAgentActivity, whose registration begins with model:null on both revisions. The exact known-model/missing-effort collabAgentToolCall branch was not exercised by a real provider. Test-only replay results are not presented as live reproduction. Native desktop/mobile, other OSes, null/blank real metadata, and a real stale-response/rerouting race were not verified.