Carry an agent's first prompt on its command line instead of pasting it (step 3 of 7)

.github/workflows/pr.yml

@@ -1206,6 +1206,7 @@ jobs:
           src/main/providers/windows-conpty-wide-char-duplication.node-pty.test.ts
           src/main/providers/pty-repaint-wide-char-buffer.node-pty.test.ts
           src/shared/child-process/windows-command-line.win32.test.ts
+          src/shared/cmd-startup-arg-quoting.win32.test.ts
           src/shared/child-process/windows-cmd-shim-resolution.test.ts
           src/shared/child-process/windows-cmd-shim-resolution.win32.test.ts
           src/main/agent-hooks/windows-hook-payload-delivery.test.ts

config/scripts/pr-code-change-scope.mjs

@@ -345,6 +345,7 @@ const WINDOWS_PACKAGE_TESTS = [
   'src/main/providers/windows-conpty-wide-char-duplication.node-pty.test.ts',
   'src/main/providers/pty-repaint-wide-char-buffer.node-pty.test.ts',
   'src/shared/child-process/windows-command-line.win32.test.ts',
+  'src/shared/cmd-startup-arg-quoting.win32.test.ts',
   'src/shared/child-process/windows-cmd-shim-resolution.test.ts',
   'src/shared/child-process/windows-cmd-shim-resolution.win32.test.ts',
   'src/main/agent-hooks/windows-hook-payload-delivery.test.ts',

src/main/agent-hooks/first-work-branch-rename.test.ts

@@ -1,3 +1,4 @@
+import { buildLaunchFilePointer } from '../../shared/launch-prompt-file'
 import { beforeEach, describe, expect, it, vi } from 'vitest'
 import type { GlobalSettings } from '../../shared/global-settings-types'
 import type { Repo } from '../../shared/repo-types'
@@ -258,6 +259,33 @@ describe('maybeAutoRenameBranchOnFirstWork', () => {
     expect(onRenamed).toHaveBeenCalledWith(REPO_ID)
   })
 
+  it('never names the branch after a launch-file pointer, and still renames on a later prompt', async () => {
+    const { deps, onRenamed } = makeDeps()
+    await maybeAutoRenameBranchOnFirstWork(
+      workingEvent({ prompt: buildLaunchFilePointer('/tmp/orca-launch-file-a1/task-context.md') }),
+      deps
+    )
+    expect(generateBranchNameMock).not.toHaveBeenCalled()
+    expect(onRenamed).not.toHaveBeenCalled()
+
+    await maybeAutoRenameBranchOnFirstWork(workingEvent({ prompt: 'Fix auth from note #1' }), deps)
+    expect(generateBranchNameMock).toHaveBeenCalledTimes(1)
+  })
+
+  it('names the branch from the prompt a launch file carried, when main kept it', async () => {
+    const { deps, onRenamed } = makeDeps({ showLaunchPrompt: () => 'Fix auth from note #1' })
+    await maybeAutoRenameBranchOnFirstWork(
+      workingEvent({ prompt: buildLaunchFilePointer('/tmp/orca-launch-file-a1/task-context.md') }),
+      deps
+    )
+    expect(generateBranchNameMock).toHaveBeenCalledWith(
+      expect.objectContaining({ firstPrompt: 'Fix auth from note #1' }),
+      expect.anything(),
+      expect.anything()
+    )
+    expect(onRenamed).toHaveBeenCalled()
+  })
+
   it('asks to align the on-disk folder with the generated slug after renaming', async () => {
     const { deps, renameWorktreeFolder } = makeDeps()
     await maybeAutoRenameBranchOnFirstWork(workingEvent(), deps)

src/main/agent-hooks/first-work-branch-rename.ts

@@ -1,4 +1,5 @@
 // On first agent work in a fresh workspace, replace the auto-generated creature branch (e.g. `you/Nautilus`) with a short work-derived name.
+import { isLaunchFilePointer } from '../../shared/launch-prompt-file'
 import type { GlobalSettings } from '../../shared/global-settings-types'
 import type { Repo } from '../../shared/repo-types'
 import { isFolderRepo } from '../../shared/repo-kind'
@@ -67,6 +68,8 @@ export type FirstWorkBranchRenameDeps = {
   resolveWorktreeIdForTab: (tabId: string) => string | undefined
   /** Invalidate caches + notify the renderer so the new branch name surfaces. */
   onRenamed: (repoId: string) => void
+  /** The prompt Orca shows for a pane's reported one (`launchPromptShownForPane`). */
+  showLaunchPrompt?: (paneKey: string, prompt: string) => string
 }
 
 // inFlight blocks concurrent generation; settled caches definitive verdicts (transient bails stay unsettled to retry later).
@@ -114,7 +117,14 @@ export async function maybeAutoRenameBranchOnFirstWork(
   if (settledWorktreeIds.has(worktreeId) || inFlightWorktreeIds.has(worktreeId)) {
     return
   }
-  const prompt = event.prompt?.trim()
+  const hookPrompt = event.prompt?.trim()
+  // Why the pointer: a launch that carried its prompt in a file names only that file here, which
+  // would name the branch after "read the task file". Its real prompt, when main kept it, names it
+  // instead; otherwise (no longer kept, as after a restart) a later prompt of the user's own can.
+  const shown = hookPrompt
+    ? (deps.showLaunchPrompt?.(event.paneKey, hookPrompt) ?? hookPrompt).trim()
+    : undefined
+  const prompt = shown && !isLaunchFilePointer(shown) ? shown : undefined
   if (!prompt) {
     return
   }

src/main/agent-hooks/first-work-rename-runtime.ts

@@ -1,3 +1,4 @@
+import { launchPromptShownForPane } from './launch-file-prompt-by-pane'
 import { existsSync } from 'node:fs'
 import { parseWorkspaceKey } from '../../shared/workspace-scope'
 import { getRepoIdFromWorktreeId } from '../../shared/worktree/id'
@@ -21,6 +22,7 @@ export function firstWorkRenameDeps(
   >
 ): FirstWorkBranchRenameDeps {
   return {
+    showLaunchPrompt: launchPromptShownForPane,
     getSettings: () => store.getSettings(),
     getRepo: (repoId) => store.getRepo(repoId),
     getAgentEnvResolvers: () => runtime.getCommitMessageAgentEnvironmentResolvers(),

src/main/agent-hooks/launch-file-prompt-by-pane.ts

@@ -0,0 +1,34 @@
+import { opensLikeLaunchFilePointer, type LaunchFile } from '../../shared/launch-prompt-file'
+
+// Why bounded: an entry outlives its launch until newer ones push it out; the rename reads only a
+// prompt's opening, and a pane's rename runs within minutes of its launch.
+const KEPT_PANES = 50
+const KEPT_CHARS = 4_000
+const promptByPane = new Map<string, string>()
+
+/**
+ * Keeps a launch file's prompt for the pane it launched: the agent's hook reports only the pointer
+ * to the file, and Orca shows the prompt instead (`launchPromptShownForPane`).
+ */
+export function rememberLaunchFilePrompt(
+  paneKey: string,
+  launchFile: LaunchFile | undefined
+): void {
+  if (!launchFile) {
+    return
+  }
+  promptByPane.delete(paneKey)
+  promptByPane.set(paneKey, launchFile.content.slice(0, KEPT_CHARS))
+  for (const oldest of promptByPane.keys()) {
+    if (promptByPane.size <= KEPT_PANES) {
+      break
+    }
+    promptByPane.delete(oldest)
+  }
+}
+
+/** The one place Orca reads a pane's reported prompt or session title: a launch file's pointer,
+ *  whole or cut short, reads as the prompt it points at, when this host kept it. */
+export function launchPromptShownForPane(paneKey: string, prompt: string): string {
+  return opensLikeLaunchFilePointer(prompt) ? (promptByPane.get(paneKey) ?? prompt) : prompt
+}

src/main/agent-hooks/server-launch-file-prompt.test.ts

@@ -0,0 +1,26 @@
+import { describe, expect, it } from 'vitest'
+import { makePaneKey } from '../../shared/stable-pane-id'
+import { buildLaunchFilePointer, carryInLaunchFile } from '../../shared/launch-prompt-file'
+import { AgentHookServer } from './server'
+import { rememberLaunchFilePrompt } from './launch-file-prompt-by-pane'
+
+const LAUNCHED = makePaneKey('tab-launched', '55555555-5555-4555-8555-555555555555')
+const POINTER = buildLaunchFilePointer('/tmp/orca-launch-file-a1/task-context.md')
+
+describe('a status row whose agent reports its launch file’s pointer', () => {
+  // Why: the agent's own hook sees only the pointer sentence; every Orca reader of the row (status
+  // row, dashboard, phone) must see the prompt the user wrote.
+  it('shows the prompt the file carries', () => {
+    rememberLaunchFilePrompt(LAUNCHED, carryInLaunchFile('fix the flaky test').launchFile)
+    const server = new AgentHookServer()
+    server.ingestTerminalStatus({
+      paneKey: LAUNCHED,
+      tabId: 'tab-launched',
+      worktreeId: 'wt-1',
+      payload: { state: 'working', prompt: POINTER }
+    })
+    expect(server.getStatusSnapshot()).toEqual([
+      expect.objectContaining({ paneKey: LAUNCHED, prompt: 'fix the flaky test' })
+    ])
+  })
+})

src/main/agent-hooks/server/server-status-update.ts

@@ -19,6 +19,7 @@ import {
 import { isStaleGrokTurnEnd } from './server-grok-status-rules'
 import { resolveCancelVerdictLatch } from './server-cancel-verdict-latch'
 import { AgentHookServerStatusApplication } from './server-status-application'
+import { launchPromptShownForPane } from '../launch-file-prompt-by-pane'
 
 export abstract class AgentHookServerStatusUpdate extends AgentHookServerStatusApplication {
   protected applyNormalizedStatus(
@@ -28,6 +29,12 @@ export abstract class AgentHookServerStatusUpdate extends AgentHookServerStatusA
     observedAt?: number,
     mutationBefore?: EnrichedAgentHookEventPayload
   ): EnrichedAgentHookEventPayload | undefined {
+    // Why here: every status row (local, relayed, replayed) passes through, so no reader of the
+    // prompt (status row, dashboard, mobile) shows a launch file's pointer instead of the prompt.
+    const prompt = launchPromptShownForPane(incoming.paneKey, incoming.payload.prompt)
+    if (prompt !== incoming.payload.prompt) {
+      incoming = { ...incoming, payload: { ...incoming.payload, prompt } }
+    }
     const transitioned = transitionHookPresence(
       incoming,
       this.state.lastStatusByPaneKey.get(incoming.paneKey)

src/main/agent-launch/agent-launch-executor.test.ts

@@ -27,7 +27,7 @@ function harness(options: {
   structuredCreateError?: Error
   deliveredMessageId?: string | null
   terminalPromptDelivered?: boolean
-  /** Whether the surface reports that its typed line took the offered prompt. */
+  /** Whether the surface reports the offered prompt rode the launch command. */
   lineCarriesPrompt?: boolean
 }) {
   const calls: string[] = []
@@ -355,7 +355,17 @@ describe('delivering a launch prompt to a terminal agent', () => {
     expect(h.deliverTerminalPrompt).not.toHaveBeenCalled()
   })
 
-  it('pastes an argv agent’s prompt after start when the surface reports its typed line could not carry it', async () => {
+  it('never pastes an argv agent’s prompt, however long: its launch command carries it', async () => {
+    const h = harness({ createSupport: { supported: false, reason: 'wsl' } })
+    const long = { delivery: 'submit' as const, text: `line one\n${'x'.repeat(20_000)}` }
+    const result = await h.run({ ...CREATE_INTENT, prompt: long })
+
+    expect(result.prompt).toEqual({ delivery: 'submit', outcome: 'handed-to-terminal' })
+    expect(h.createTerminalAgent.mock.calls[0]?.[0]).toMatchObject({ startupPrompt: long.text })
+    expect(h.deliverTerminalPrompt).not.toHaveBeenCalled()
+  })
+
+  it('pastes after start when the surface reports the prompt needed a file the agent cannot read', async () => {
     const h = harness({
       createSupport: { supported: false, reason: 'wsl' },
       lineCarriesPrompt: false
@@ -375,7 +385,7 @@ describe('delivering a launch prompt to a terminal agent', () => {
     })
   })
 
-  it('pastes into an agent-first create’s startup terminal when its typed line could not carry the prompt', async () => {
+  it('pastes into an agent-first create’s startup terminal that did not take the prompt', async () => {
     const h = harness({ settings: null, lineCarriesPrompt: false })
     const result = await h.run({ ...CREATE_INTENT, prompt: SUBMIT })
 
@@ -388,6 +398,14 @@ describe('delivering a launch prompt to a terminal agent', () => {
     })
   })
 
+  it('never pastes into an agent-first create’s startup terminal that took the prompt', async () => {
+    const h = harness({ settings: null })
+    const result = await h.run({ ...CREATE_INTENT, prompt: SUBMIT })
+
+    expect(result.prompt).toEqual({ delivery: 'submit', outcome: 'handed-to-terminal' })
+    expect(h.deliverTerminalPrompt).not.toHaveBeenCalled()
+  })
+
   it('writes into a reused terminal, whose process started before the launch existed', async () => {
     const h = harness({})
     const result = await h.run({

src/main/agent-launch/agent-launch-executor.ts

@@ -220,7 +220,7 @@ async function resolveWorkspace(
   startupTerminalHandle: string | undefined
   startupTerminalPaneKey?: string
   warning?: string
-  /** True when this create folded the prompt into the agent's startup command. */
+  /** True when this create put the prompt on the agent's startup command. */
   promptRodeLaunchCommand?: boolean
 }> {
   const { intent } = execution
@@ -257,7 +257,7 @@ export type CreatedSurface = {
   outcome: AgentLaunchResult['outcome']
   warning?: string
   structured?: AgentLaunchStructuredSurface
-  /** True when this create folded the prompt into the agent's launch command. */
+  /** True when this create put the prompt on the agent's launch command. */
   promptRodeLaunchCommand?: boolean
 }
 

src/main/agent-launch/agent-launch-prompt-delivery.ts

@@ -5,15 +5,17 @@
  * surface exists and in what order; everything here decides how the text reaches whichever surface
  * that turned out to be, and each surface takes it differently:
  *
- *   structured session   ->  committed to the transcript, named by a message id  ->  journaled
- *   terminal, line fits  ->  folded into the command that execs the agent        ->  handed-to-terminal
- *   terminal, otherwise  ->  bracketed paste into the live PTY once it is ready  ->  handed-to-terminal
- *   anything unproven    ->                                                      ->  not-delivered
+ *   structured session  ->  committed to the transcript, named by a message id  ->  journaled
+ *   terminal, argv CLI  ->  on the command that execs the agent                 ->  handed-to-terminal
+ *   terminal, otherwise ->  bracketed paste into the live PTY once it is ready  ->  handed-to-terminal
+ *   anything unproven   ->                                                      ->  not-delivered
  *
- * argv has no readiness race, so it is offered wherever the agent's CLI takes a prompt argument
- * (`agentPromptRidesLaunchCommand`). But that command is TYPED into the user's shell, and a long or
- * multi-line typed line fails in ways argv itself does not, so whether the offer was taken is decided
- * where the line is built (`startup-line-prompt-carry`) and reported back, never predicted here.
+ * argv has no readiness race, so it carries the prompt wherever the agent's CLI takes a prompt
+ * argument (`agentPromptRidesLaunchCommand`). Where it rides is decided once, where the line is built
+ * (`carryLaunchPrompt`): on the line, which the host stages when it is too long or multi-line to
+ * type, or in a launch file the line points at. A prompt the rule leaves for the paste (an agent
+ * not known to read a launch file, or a host that cannot write its staging folder) is pasted, and
+ * the surface reports that back.
  */
 
 import type {
@@ -38,7 +40,7 @@ export async function settleLaunchPromptDisposal(
     const messageId = await deliverStructuredLaunchPrompt(execution, created.structured)
     return messageId ? { outcome: 'journaled', messageId } : NOT_DELIVERED
   }
-  // The surface reported that the startup command carried the prompt; there is nothing left to write.
+  // The startup command carried the prompt; there is nothing left to write.
   if (created.promptRodeLaunchCommand) {
     return HANDED_TO_TERMINAL
   }
@@ -102,10 +104,8 @@ function launchSubmitText(intent: AgentLaunchIntent): string | undefined {
   return intent.prompt?.delivery === 'submit' && intent.prompt.text ? intent.prompt.text : undefined
 }
 
-/**
- * The prompt offered to a terminal's launch command, which is an argv-mode agent's and only an
- * argv-mode agent's. An offer, not a decision: the surface reports whether the typed line took it.
- */
+/** The prompt a terminal's launch command carries, which is an argv-mode agent's and only an
+ *  argv-mode agent's. */
 export function argvLaunchPrompt(intent: AgentLaunchIntent): string | undefined {
   const text = launchSubmitText(intent)
   return text && agentPromptRidesLaunchCommand(intent.agent) ? text : undefined

src/main/agent-launch/agent-launch-surface-factories.ts

@@ -26,7 +26,8 @@ export type AgentLaunchSurfaceFactory = {
     agent: TuiAgent
     options?: Readonly<Record<string, unknown>>
     /** Offered only for an agent whose CLI takes the prompt on argv. It rides the launch command
-     *  only when the typed line can carry it; `promptRodeLaunchCommand` reports which happened. */
+     *  unless `carryLaunchPrompt` leaves it for the paste; `promptRodeLaunchCommand` reports which
+     *  happened. */
     startupPrompt?: string
     /** Replaces the settings default for this launch only; `null` means no arguments at all. */
     agentArgs?: string | null
@@ -40,7 +41,7 @@ export type AgentLaunchSurfaceFactory = {
     /** The pane this create minted; a factory whose runtime reports none omits it, never invents. */
     paneKey?: string
     warning?: string
-    /** Reported by the surface that built the typed line, never predicted by the executor. */
+    /** Reported by the surface that built the launch command, never predicted by the executor. */
     promptRodeLaunchCommand?: boolean
   }>
   /**
@@ -59,8 +60,8 @@ export type AgentLaunchSurfaceFactory = {
    * Writes the launch text into a terminal agent's live PTY, answering whether it landed.
    *
    * The other half of `startupPrompt`, for the cases the launch command cannot serve: a
-   * `stdin-after-start` agent, whose CLI takes no prompt argument; a prompt the typed line cannot
-   * carry; and a reused terminal, whose process was already running before this launch existed. `false` for every failure, on the same rule the structured
+   * `stdin-after-start` agent, whose CLI takes no prompt argument; a prompt `carryLaunchPrompt`
+   * leaves for the paste; and a reused terminal, whose process was already running before this launch existed. `false` for every failure, on the same rule the structured
    * twin follows — a launch whose agent is running must not fail because its text did not land.
    */
   deliverTerminalPrompt?(args: {
@@ -105,7 +106,7 @@ export type AgentLaunchWorkspaceFactory = {
      *  must await that gate explicitly instead. */
     startupAgent: TuiAgent | undefined
     /** Offered only alongside a `startupAgent` whose CLI takes the prompt on argv: agent-first
-     *  creation builds the startup command, so that is where the typed line is measured. */
+     *  creation builds the startup command, so that is where the carry is decided. */
     startupPrompt?: string
     /** Inputs needed when this terminal is created as the worktree's startup surface. */
     agentArgs?: string | null
@@ -121,7 +122,7 @@ export type AgentLaunchWorkspaceFactory = {
     startupTerminalPaneKey?: string
     /** Created, but incomplete — surfaced on the launch result rather than dropped. */
     warning?: string
-    /** Reported by the create that built the startup command's typed line. */
+    /** Reported by the create that built the startup command. */
     promptRodeLaunchCommand?: boolean
   }>
 }

src/main/ai-vault/session-title-launch-prompt.ts

@@ -0,0 +1,39 @@
+import type {
+  AiVaultSessionTitleRequest,
+  AiVaultSessionTitlesResult
+} from '../../shared/ai-vault-session-title'
+import { launchPromptShownForPane } from '../agent-hooks/launch-file-prompt-by-pane'
+import { normalizeTitleText } from './session-scanner-text-normalization'
+
+/**
+ * Titles a launch file's session with the prompt it carried. Its transcript opens with Orca's
+ * pointer, so the tab label and agent row would otherwise show "The full task is in the file …".
+ */
+export function showLaunchPromptsInSessionTitles(
+  requests: readonly AiVaultSessionTitleRequest[],
+  result: AiVaultSessionTitlesResult
+): AiVaultSessionTitlesResult {
+  const paneBySession = new Map(
+    requests.flatMap((request) =>
+      request.paneKey ? [[`${request.agent}\0${request.sessionId}`, request.paneKey] as const] : []
+    )
+  )
+  if (paneBySession.size === 0) {
+    return result
+  }
+  return {
+    titles: result.titles.map((title) => {
+      const paneKey = paneBySession.get(`${title.agent}\0${title.sessionId}`)
+      const shown = paneKey ? launchPromptShownForPane(paneKey, title.title) : title.title
+      return shown === title.title
+        ? title
+        : { ...title, title: normalizeTitleText(shown) ?? title.title }
+    })
+  }
+}
+
+/** A request as another host receives it: the pane is this host's to read. */
+export function withoutPaneKey(request: AiVaultSessionTitleRequest): AiVaultSessionTitleRequest {
+  const { paneKey: _paneKey, ...forwarded } = request
+  return forwarded
+}