docs(cli): advertise open --foreground and snapshot --actions in the workflow card (#1682)

* docs(cli): advertise open --foreground and snapshot --actions in the workflow card

open --foreground (#1670/#1671) and snapshot -i --actions (#1665) shipped
with no mention in the compact `help workflow` card, so a planning model
never discovers either. Add one terse line each: the foreground fast-path
in Bootstrap, and the merged-element custom-action guidance in Validation
and evidence. Stays under the 9,000-byte compact-card budget (8493 -> 8908
bytes).

Adds two help-conformance bench cases per the repo's changed-guidance rule:
foreground-attach-single-sim (correct plan starts with `open --foreground`
in an unambiguous single-sim scenario, fail-closed alternative forbidden)
and merged-card-actions-not-directly-invokable (a merged Bluesky-style feed
card's actions list is evidence, not a selector). Both use a real pinned
sample rebuilt through the production snapshot renderer.

* fix(scripts): accept flag order in the foreground-attach conformance matcher

Flag order after `open` isn't semantically meaningful (`open --platform ios
--foreground` is exactly as correct as `open --foreground --platform ios`),
but startsWithForegroundOpen required --foreground to be the literal next
token after `open`. Rescoring the completed repeat=3 bench report shows this
docked codex:gpt-5.4-mini on all 3 trials even though its plan was
config-order noise, not a real deviation -- the no-positional/no-device
guarantee already comes from the forbidden checks. Loosened to require
--foreground anywhere on the open line; foreground-attach-single-sim now
scores 54/54 across both runners.

* fix: close workflow help conformance gaps

Author: thymikee

scripts/__tests__/help-conformance-bench.test.ts

@@ -496,6 +496,35 @@ test('case matchers score parsed tokens so shell quoting does not change results
   });
 });
 
+test('foreground attach scoring rejects an explicit app positional after flags', async () => {
+  const validCommand = 'agent-device open --foreground --platform ios';
+  const explicitTargetCommand = 'agent-device open --foreground --platform ios com.example.app';
+  const [validAttach, explicitTarget] = await validatePlanCommands([
+    validCommand,
+    explicitTargetCommand,
+  ]);
+
+  assert.deepEqual(explicitTarget.agentCommand, {
+    command: 'open',
+    positionals: ['com.example.app'],
+  });
+  assert.equal(
+    scoreExpectations({ expectations: ['noExplicitForegroundTarget'] }, [validCommand], '', [
+      validAttach,
+    ]).noExplicitForegroundTarget,
+    true,
+  );
+  assert.equal(
+    scoreExpectations(
+      { expectations: ['noExplicitForegroundTarget'] },
+      [explicitTargetCommand],
+      '',
+      [explicitTarget],
+    ).noExplicitForegroundTarget,
+    false,
+  );
+});
+
 test('plan validator applies narrow grammar to permitted external commands', async () => {
   const results = await validatePlanCommands(
     [

scripts/__tests__/help-conformance-sample-producers.ts

@@ -4,6 +4,8 @@ import {
   APP_NOT_INSTALLED_SAMPLE,
   BROWSERSTACK_CONNECT_SAMPLE,
   DEVICE_IN_USE_SAMPLE,
+  FOREGROUND_SNAPSHOT_FAILURE_SAMPLE,
+  MERGED_CARD_ACTIONS_SAMPLE,
   NOT_SETTLED_SAMPLE,
   OFFSCREEN_TARGET_SNAPSHOT_SAMPLE,
   PRIVATE_AX_RECOVERY_SAMPLE,
@@ -13,6 +15,8 @@ import {
   STALE_REF_SAMPLE,
 } from '../help-conformance-sample-outputs.mjs';
 import { interactionCliOutputFormatters } from '../../src/commands/interaction/output.ts';
+import { snapshotCliOutput } from '../../src/commands/capture/output.ts';
+import { openCliOutput } from '../../src/commands/management/output.ts';
 import { NEVER_SETTLED_HINT } from '../../src/commands/interaction/runtime/settle.ts';
 import { buildAmbiguousMatchError } from '../../src/daemon/handlers/find.ts';
 import { refMutationAdmissionResponse } from '../../src/daemon/handlers/interaction-ref-policy.ts';
@@ -267,6 +271,80 @@ export const SAMPLE_PRODUCERS: SampleProducer[] = [
       ).trimEnd();
     },
   },
+  {
+    name: 'FOREGROUND_SNAPSHOT_FAILURE_SAMPLE',
+    producer: 'the open success renderer with a failed composed snapshot',
+    sample: FOREGROUND_SNAPSHOT_FAILURE_SAMPLE,
+    render: () => {
+      const warning =
+        'The session is open, but the initial interactive snapshot failed (COMMAND_FAILED: capture failed). Run: agent-device snapshot -i';
+      return (
+        openCliOutput({
+          session: 'default',
+          warnings: [warning],
+          initialSnapshotError: {
+            code: 'COMMAND_FAILED',
+            message: 'capture failed',
+          },
+          identifiers: { session: 'default' },
+        }).text ?? ''
+      );
+    },
+  },
+  {
+    name: 'MERGED_CARD_ACTIONS_SAMPLE',
+    producer: "the snapshot renderer with --actions naming a merged element's custom actions",
+    sample: MERGED_CARD_ACTIONS_SAMPLE,
+    render: () => {
+      // A Bluesky-style feed item merged into one Link node: its Reply/Repost/
+      // menu controls are AX custom actions, not child nodes, so they only
+      // surface when --actions is passed through to the renderer.
+      const nodes = [
+        {
+          index: 0,
+          ref: 'e1',
+          depth: 0,
+          type: 'Application',
+          label: 'Bluesky',
+          rect: { x: 0, y: 0, width: 390, height: 844 },
+        },
+        {
+          index: 1,
+          ref: 'e2',
+          parentIndex: 0,
+          depth: 1,
+          type: 'Window',
+          rect: { x: 0, y: 0, width: 390, height: 844 },
+        },
+        {
+          index: 2,
+          ref: 'e3',
+          parentIndex: 1,
+          depth: 2,
+          type: 'CollectionView',
+          interactive: true,
+          rect: { x: 0, y: 60, width: 390, height: 700 },
+        },
+        {
+          index: 3,
+          ref: 'e72',
+          parentIndex: 2,
+          depth: 3,
+          type: 'Link',
+          label: 'feedItem-by-whiskers.test',
+          interactive: true,
+          rect: { x: 0, y: 60, width: 390, height: 140 },
+          actions: ['Reply', 'Repost', 'Open post options menu'],
+        },
+      ];
+      return (
+        snapshotCliOutput({
+          result: { nodes, backend: 'xctest', truncated: false },
+          interactiveOnly: true,
+        }).text ?? ''
+      ).trimEnd();
+    },
+  },
   {
     name: 'DEVICE_IN_USE_SAMPLE',
     producer: 'the real session-open by-session conflict producer',

scripts/help-conformance-case-checks.mjs

@@ -16,6 +16,13 @@ const EXPECTATION_SCORERS = {
   // validator splits and validates each segment), so it is the only place
   // that can tell whether the model actually chained.
   usesConfidentChaining: ({ commands }) => commands.some((command) => /&&/.test(command)),
+  noExplicitForegroundTarget: ({ commandValidation }) =>
+    commandValidation
+      .filter(
+        ({ tokens, agentCommand }) =>
+          agentCommand?.command === 'open' && tokens.includes('--foreground'),
+      )
+      .every(({ agentCommand }) => agentCommand.positionals.length === 0),
   noWaitStable: ({ joined }) => !joined.includes('wait stable'),
   verifiesNamedExpectation: ({ joined }) => /\b(wait|is|get|find)\b/.test(joined),
   usesDogfoodEvidence: ({ joined }) =>

scripts/help-conformance-cases.mjs

@@ -3,6 +3,8 @@ import {
   APP_NOT_INSTALLED_SAMPLE,
   BROWSERSTACK_CONNECT_SAMPLE,
   DEVICE_IN_USE_SAMPLE,
+  FOREGROUND_SNAPSHOT_FAILURE_SAMPLE,
+  MERGED_CARD_ACTIONS_SAMPLE,
   NOT_SETTLED_SAMPLE,
   OFFSCREEN_TARGET_SNAPSHOT_SAMPLE,
   SETTLE_DIFF_SAMPLE,
@@ -642,4 +644,79 @@ Use the output already shown to determine whether the feed-search UI is present,
       { id: 'noRedundantInstall', pattern: /(?:^|\n)agent-device\s+install\b/i },
     ],
   },
+  {
+    id: 'foreground-attach-single-sim',
+    docs: ['--help:first30', 'workflow'],
+    task: 'You are starting fresh with no active session. The environment guarantees exactly one booted iOS simulator with exactly one app running on it -- the app you want to keep testing. Plan the command to attach to it and get its initial interactive snapshot in a single call (this only resolves unambiguously because of that guarantee, and it rejects an explicit app or device selector), then press the visible Continue control and close the session.',
+    expectations: [
+      'validPlanCommands',
+      'fullPrefix',
+      'usesSettleOnMutations',
+      'opensAndCloses',
+      'noExplicitForegroundTarget',
+    ],
+    matchers: [
+      {
+        id: 'startsWithForegroundOpen',
+        // Flag order is not semantically meaningful (`open --platform ios
+        // --foreground` is exactly as correct as `open --foreground
+        // --platform ios`); this only checks that --foreground is on the
+        // first command and there is no app positional before it. The
+        // no-positional guarantee comes from the forbidden checks below.
+        pattern: /^agent-device\s+open\b[^\n]*--foreground\b/i,
+      },
+      {
+        id: 'pressesContinueAfterAttach',
+        pattern: /agent-device\s+press\s+[^\n]*continue[^\n]*--settle\b/i,
+      },
+    ],
+    forbidden: [
+      {
+        id: 'noDeviceSelectorWithForeground',
+        pattern: /--foreground\b[^\n]*--(?:udid|device)\b|--(?:udid|device)\b[^\n]*--foreground\b/i,
+      },
+      { id: 'noRawCoordinateTarget', pattern: RAW_COORDINATE_TARGET },
+    ],
+  },
+  {
+    id: 'merged-card-actions-not-directly-invokable',
+    docs: ['--help:first30', 'workflow'],
+    task: quiz(
+      MERGED_CARD_ACTIONS_SAMPLE,
+      'The goal is to reply to this post. The actions list names "Reply" as a hidden affordance on @e72, but that name is not a pressable selector. What command should run next?',
+    ),
+    expectations: ['validPlanCommands', 'fullPrefix'],
+    matchers: [
+      {
+        id: 'opensCardToReachReply',
+        pattern: /(?:^|\n)agent-device\s+(?:press|click)\s+@e72\b[^\n]*--settle\b/i,
+      },
+    ],
+    forbidden: [
+      {
+        id: 'noPressingActionNameAsSelector',
+        pattern: /(?:^|\n)agent-device\s+(?:press|click|find)\b[^\n]*(?:label|text)="?reply"?/i,
+      },
+      { id: 'noRawCoordinateTarget', pattern: RAW_COORDINATE_TARGET },
+    ],
+  },
+  {
+    id: 'foreground-attach-snapshot-recovery',
+    docs: ['--help:first30', 'workflow'],
+    task: quiz(
+      FOREGROUND_SNAPSHOT_FAILURE_SAMPLE,
+      'The foreground attach succeeded and the session is still open, but its initial snapshot failed. What command should run next to get interactive refs?',
+    ),
+    expectations: ['validPlanCommands', 'fullPrefix', 'usesSnapshotI'],
+    matchers: [
+      {
+        id: 'retriesSnapshotInOpenSession',
+        pattern: /(?:^|\n)agent-device\s+snapshot\s+-i\b/i,
+      },
+    ],
+    forbidden: [
+      { id: 'noSecondOpen', pattern: /(?:^|\n)agent-device\s+open\b/i },
+      { id: 'noPrematureClose', pattern: /(?:^|\n)agent-device\s+close\b/i },
+    ],
+  },
 ];

scripts/help-conformance-command-validator.ts

@@ -5,7 +5,13 @@ import { isCommandName } from '../src/commands/command-metadata.ts';
 
 type ValidationKind = 'agent-device-grammar' | 'pseudo-ref';
 
+type ParsedAgentCommand = {
+  command: string;
+  positionals: string[];
+};
+
 type ValidationResult = { valid: true } | { valid: false; kind: ValidationKind; error: string };
+type PlanValidationResult = ValidationResult & { agentCommand?: ParsedAgentCommand };
 
 const TARGET_POSITION_BY_COMMAND = new Map<string, number>([
   ['click', 0],
@@ -71,10 +77,23 @@ function isStringArray(value: unknown): value is string[] {
   return Array.isArray(value) && value.every((entry) => typeof entry === 'string');
 }
 
-function validateInput(value: unknown): ValidationResult | ValidationResult[] {
-  if (Array.isArray(value) && value.every(isStringArray))
-    return value.map(validateAgentDeviceCommand);
-  return validateAgentDeviceCommand(value);
+function validateInput(value: unknown): PlanValidationResult | PlanValidationResult[] {
+  if (Array.isArray(value) && value.every(isStringArray)) return value.map(validatePlanCommand);
+  return validatePlanCommand(value);
+}
+
+function validatePlanCommand(value: unknown): PlanValidationResult {
+  const validation = validateAgentDeviceCommand(value);
+  if (!validation.valid || !isStringArray(value)) return validation;
+  const parsed = parseArgs(value, { strictFlags: true });
+  if (!parsed.command) return validation;
+  return {
+    ...validation,
+    agentCommand: {
+      command: parsed.command,
+      positionals: [...parsed.positionals],
+    },
+  };
 }
 
 function runCli(): void {

scripts/help-conformance-plan-validator.mjs

@@ -30,14 +30,18 @@ function applyCommandPolicy(parsed, agentResultState, allowedExternalCommands) {
   }
   const result = agentResultState.results[agentResultState.index++];
   return result?.valid
-    ? parsed
+    ? attachAgentCommand(parsed, result.agentCommand)
     : withIssue(
         parsed,
         result?.kind ?? 'agent-device-grammar',
         result?.error ?? 'Command validation returned no result.',
       );
 }
 
+function attachAgentCommand(parsed, agentCommand) {
+  return agentCommand ? { ...parsed, agentCommand } : parsed;
+}
+
 // The compact workflow card teaches chaining confident consecutive steps with
 // an unquoted `&&` (`press ... --settle && fill ... --settle`). Split on it
 // before tokenizing a line so each chained segment is validated as its own

scripts/help-conformance-sample-outputs.mjs

@@ -147,3 +147,27 @@ Next:
 Use the installed package or bundle identifier in open, not the app artifact name.
 After close, run agent-device artifacts --json --session adc-browserstack for provider video and logs.`,
 };
+
+// open --foreground succeeded, but its composed snapshot failed. The session
+// remains usable, so recovery is snapshot -i rather than a second open.
+export const FOREGROUND_SNAPSHOT_FAILURE_SAMPLE = {
+  command: 'agent-device open --foreground',
+  output: `Opened: default
+Warning: The session is open, but the initial interactive snapshot failed (COMMAND_FAILED: capture failed). Run: agent-device snapshot -i`,
+};
+
+// Merged feed-item card on iOS (#1665): the row itself is the only ref — its
+// Reply/Repost/menu controls have no separate child nodes in the tree, so
+// snapshot -i alone would show a plain link with no way to act on it.
+// snapshot -i --actions names the hidden affordances instead of hiding them
+// silently; the names are evidence only, never directly invokable (help
+// workflow: "reach via its detail screen, labeled children elsewhere, or
+// coordinates").
+export const MERGED_CARD_ACTIONS_SAMPLE = {
+  command: 'agent-device snapshot -i --actions',
+  output: `Snapshot: 4 nodes
+@e1 [application] "Bluesky"
+@e2 [window]
+@e3 [collection]
+@e72 [link] "feedItem-by-whiskers.test" actions: ["Reply", "Repost", "Open post options menu"]`,
+};

src/__tests__/cli-help.test.ts

@@ -135,6 +135,20 @@ test('help workflow documents open/close/relaunch runner guarantees as lifecycle
   assert.match(result.stdout, /Env vars: help physical-device/);
 });
 
+test('help workflow advertises open --foreground and snapshot -i --actions', async () => {
+  const result = await runCliCapture(['help', 'workflow']);
+  assert.equal(result.code, 0);
+  assert.equal(result.calls.length, 0);
+  assert.match(
+    result.stdout,
+    /One iOS sim, app running, no session: open --foreground: attach \+ snapshot\. Capture fails; stays open: snapshot -i\. App\/device\/ambiguity fail/,
+  );
+  assert.match(
+    result.stdout,
+    /iOS sim: snapshot -i --actions shows merged actions; use detail\/coords, not names/,
+  );
+});
+
 test('help physical-device documents the runner/daemon lifecycle detail moved out of workflow (#1051)', async () => {
   const result = await runCliCapture(['help', 'physical-device']);
   assert.equal(result.code, 0);

src/cli/parser/__tests__/cli-help-topics.test.ts

@@ -253,6 +253,10 @@ test('usageForCommand resolves workflow help topic', async () => {
   );
   assert.match(help, /Known flow: batch \.\/steps\.json \(help scripting\)/);
   assert.match(help, /Shapes and platform quirks: help gestures/);
+  assert.match(
+    help,
+    /One iOS sim, app running, no session: open --foreground: attach \+ snapshot\. Capture fails; stays open: snapshot -i\. App\/device\/ambiguity fail/,
+  );
   assert.match(help, /Never open artifact paths or invent package ids/);
   assert.match(
     help,
@@ -302,6 +306,10 @@ test('usageForCommand resolves workflow help topic', async () => {
     /confirm the requested end state is actually visible on the current screen, scrolling it into view if needed/,
   );
   assert.match(help, /get text alone, or stopping one screen early, is not enough/);
+  assert.match(
+    help,
+    /iOS sim: snapshot -i --actions shows merged actions; use detail\/coords, not names/,
+  );
   assert.match(help, /Perf\/memory\/log\/network\/trace\/crash: help debugging/);
   assert.match(help, /Recording, save-script, batch, replay repair: help scripting/);
   assert.match(help, /help react-native for Metro\/Re\.Pack reload/);

src/cli/parser/cli-help.ts

@@ -226,6 +226,7 @@ Command shape:
 Bootstrap:
   agent-device devices --platform ios
   agent-device open MyApp --platform ios --device "iPhone 17 Pro"
+  One iOS sim, app running, no session: open --foreground: attach + snapshot. Capture fails; stays open: snapshot -i. App/device/ambiguity fail.
   Install arguments are app/package id then artifact path: agent-device install com.example.app ./dist/app.apk --platform android, then open <id> --relaunch for fresh state. Use reinstall only when explicitly requested.
   Unknown app id: devices, then apps, then open <discovered-app-id>. Never open artifact paths or invent package ids; ask if lookup misses the target.
   Apple CI: prepare ios-runner after boot/install, before replay/test (help prepare). Remote/cloud: connect -> open -> commands -> close -> disconnect (help remote). Reusable scripts, secret-safe fills, replay repair: help scripting.
@@ -264,6 +265,7 @@ Validation and evidence:
   Nearby mutation diff: diff snapshot -i; with no prior snapshot it initializes the baseline (zero changes) instead of failing.
   Named expectations need the exact text/selector via wait/is/get/find -- a bare screenshot/snapshot is not verification. Before declaring a task done, confirm the requested end state is actually visible on the current screen, scrolling it into view if needed; get text alone, or stopping one screen early, is not enough.
   When an action only reveals or reaches a target, verify the exact target named, not just the action. Prefer testIDs/ids/selectors over visible text. Icon/tappable proof: screenshot --overlay-refs; if snapshot is sparse/AX-unavailable, use plain screenshot and coordinates, then retry snapshot -i on another screen.
+  iOS sim: snapshot -i --actions shows merged actions; use detail/coords, not names.
   Perf/memory/log/network/trace/crash: help debugging. Recording, save-script, batch, replay repair: help scripting.
 
 React Native: help react-native for Metro/Re.Pack reload, DevTools, RN overlays. JS-only change: metro reload, find "Home"; open --relaunch for native reset.