From 0ce1def8d4d346dcf7e9aa2bbc7c77dfc6a2d35c Mon Sep 17 00:00:00 2001
From: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
Date: Thu, 27 Aug 2026 07:47:47 -0700
Subject: [PATCH] fix(automations): wait for Codex prompt lifecycle

---
 .../hooks/useAutomationDispatchEvents.test.ts    |  7 +++++--
 .../src/hooks/useAutomationDispatchEvents.ts     | 16 +++++++++++++---
 .../lib/launch-agent-background-session.test.ts  | 13 +++++++------
 3 files changed, 25 insertions(+), 11 deletions(-)

diff --git a/src/renderer/src/hooks/useAutomationDispatchEvents.test.ts b/src/renderer/src/hooks/useAutomationDispatchEvents.test.ts
index 81fc23a5728..19f220a479e 100644
--- a/src/renderer/src/hooks/useAutomationDispatchEvents.test.ts
+++ b/src/renderer/src/hooks/useAutomationDispatchEvents.test.ts
@@ -537,7 +537,7 @@ describe('useAutomationDispatchEvents setup launch', () => {
     )
   })
 
-  it('finalizes a fresh non-reuse terminal only after completed result persistence', async () => {
+  it('finalizes fresh Codex only after working and completed result persistence', async () => {
     const order: string[] = []
     let launchArgs: { onAgentStatus?: (payload: { state: string }) => void } = {}
     mockMarkDispatchResult.mockImplementation(
@@ -569,7 +569,10 @@ describe('useAutomationDispatchEvents setup launch', () => {
       }
     })
 
-    await registerAndDispatch()
+    await registerAndDispatch(makeAutomation({ agentId: 'codex' }))
+    launchArgs.onAgentStatus?.({ state: 'done' })
+    expect(mockFinalizeTerminalOwnership).not.toHaveBeenCalled()
+    launchArgs.onAgentStatus?.({ state: 'working' })
     launchArgs.onAgentStatus?.({ state: 'done' })
     await vi.waitFor(() => expect(mockFinalizeTerminalOwnership).toHaveBeenCalledOnce())
 
diff --git a/src/renderer/src/hooks/useAutomationDispatchEvents.ts b/src/renderer/src/hooks/useAutomationDispatchEvents.ts
index 1e46b315f28..98d8ec952cb 100644
--- a/src/renderer/src/hooks/useAutomationDispatchEvents.ts
+++ b/src/renderer/src/hooks/useAutomationDispatchEvents.ts
@@ -595,6 +595,7 @@ export function useAutomationDispatchEvents(): void {
               }
             }
           }
+          let freshLaunchSawWorking = false
           const result = await launchAgentBackgroundSession({
             agent: automation.agentId,
             worktreeId: worktree.id,
@@ -607,8 +608,15 @@ export function useAutomationDispatchEvents(): void {
             onAgentStatus: (payload) => {
               latestAssistantMessage =
                 payload.lastAssistantMessage?.trim() || latestAssistantMessage
-              // Why: session-boundary done = launch connect, not run completion (see observeAgentStatus).
-              if (payload.state !== 'done' || payload.sessionBoundary === true) {
+              if (payload.state === 'working') {
+                freshLaunchSawWorking = true
+              }
+              // Why: Claude marks launch boundaries done; Codex reports initial idle before its prompt turn works.
+              if (
+                payload.state !== 'done' ||
+                payload.sessionBoundary === true ||
+                (automation.agentId === 'codex' && !freshLaunchSawWorking)
+              ) {
                 return
               }
               handleAgentDone()
@@ -634,7 +642,9 @@ export function useAutomationDispatchEvents(): void {
             releaseTerminalOwnership()
           }
           const launchedTabId = result.tabId
-          observeAgentStatus(result.paneKey, dispatchStartedAt)
+          observeAgentStatus(result.paneKey, dispatchStartedAt, {
+            requireWorkingAfterStart: automation.agentId === 'codex'
+          })
           try {
             await markDispatchResult({
               runId: run.id,
diff --git a/src/renderer/src/lib/launch-agent-background-session.test.ts b/src/renderer/src/lib/launch-agent-background-session.test.ts
index eedd1e9fb32..33b29523a10 100644
--- a/src/renderer/src/lib/launch-agent-background-session.test.ts
+++ b/src/renderer/src/lib/launch-agent-background-session.test.ts
@@ -90,11 +90,11 @@ describe('launchAgentBackgroundSession', () => {
     })
   })
 
-  it('spawns a PTY first and creates the inactive tab already bound to it', async () => {
+  it('spawns the Codex PTY first and publishes the inactive tab already bound to it', async () => {
     const { launchAgentBackgroundSession } = await import('./launch-agent-background-session')
 
     const result = await launchAgentBackgroundSession({
-      agent: 'claude',
+      agent: 'codex',
       worktreeId: 'wt-1',
       prompt: 'run the automation',
       title: 'Nightly audit'
@@ -123,7 +123,8 @@ describe('launchAgentBackgroundSession', () => {
     expect(mockSpawn).toHaveBeenCalledWith(
       expect.objectContaining({
         cwd: '/repo/worktree',
-        command: "claude '--dangerously-skip-permissions' 'run the automation'",
+        command: "codex '--dangerously-bypass-approvals-and-sandbox' 'run the automation'",
+        startupCommandDelivery: 'shell-ready',
         env: expect.objectContaining({
           ORCA_TAB_ID: tabId,
           ORCA_WORKTREE_ID: 'wt-1'
@@ -146,11 +147,11 @@ describe('launchAgentBackgroundSession', () => {
     expect(mockSetTabLayout.mock.calls.at(-1)?.[1]).not.toHaveProperty('titlesByLeafId')
     expect(mockSpawn.mock.calls[0]?.[0]).toMatchObject({
       launchConfig: {
-        agentCommand: "claude '--dangerously-skip-permissions'",
-        agentArgs: '--dangerously-skip-permissions',
+        agentCommand: "codex '--dangerously-bypass-approvals-and-sandbox'",
+        agentArgs: '--dangerously-bypass-approvals-and-sandbox',
         agentEnv: {}
       },
-      launchAgent: 'claude',
+      launchAgent: 'codex',
       launchToken: expect.stringMatching(UUID_RE)
     })
     expect(mockSpawn.mock.calls[0]?.[0].launchToken).toBe(