feat: add diff snapshot command

README.md

@@ -14,8 +14,8 @@ The project is in early development and considered experimental. Pull requests a
 
 ## Features
 - Platforms: iOS (simulator + physical device core automation) and Android (emulator + device).
-- Core commands: `open`, `back`, `home`, `app-switcher`, `press`, `longpress`, `focus`, `type`, `fill`, `scroll`, `scrollintoview`, `wait`, `alert`, `screenshot`, `close`, `reinstall`.
-- Inspection commands: `snapshot` (accessibility tree), `appstate`, `apps`, `devices`.
+- Core commands: `open`, `back`, `home`, `app-switcher`, `press`, `long-press`, `focus`, `type`, `fill`, `scroll`, `scrollintoview`, `wait`, `alert`, `screenshot`, `close`, `reinstall`.
+- Inspection commands: `snapshot` (accessibility tree), `diff snapshot` (structural baseline diff), `appstate`, `apps`, `devices`.
 - Device tooling: `adb` (Android), `simctl`/`devicectl` (iOS via Xcode).
 - Minimal dependencies; TypeScript executed directly on Node 22+ (no build step).
 
@@ -39,9 +39,11 @@ Use `press` as the canonical tap command; `click` is an equivalent alias.
 ```bash
 agent-device open Contacts --platform ios # creates session on iOS Simulator
 agent-device snapshot
+agent-device diff snapshot # first run initializes baseline
 agent-device press @e5
 agent-device fill @e6 "John"
 agent-device fill @e7 "Doe"
+agent-device diff snapshot # subsequent runs compare against previous baseline
 agent-device press @e3
 agent-device close
 ```
@@ -135,8 +137,8 @@ agent-device swipe 540 1500 540 500 120 --count 8 --pause-ms 30 --pattern ping-p
 ## Command Index
 - `boot`, `open`, `close`, `reinstall`, `home`, `back`, `app-switcher`
 - `batch`
-- `snapshot`, `find`, `get`
-- `press` (alias: `click`), `focus`, `type`, `fill`, `longpress`, `swipe`, `scroll`, `scrollintoview`, `pinch`, `is`
+- `snapshot`, `diff snapshot`, `find`, `get`
+- `press` (alias: `click`), `focus`, `type`, `fill`, `long-press`, `swipe`, `scroll`, `scrollintoview`, `pinch`, `is`
 - `alert`, `wait`, `screenshot`
 - `trace start`, `trace stop`
 - `settings wifi|airplane|location on|off`
@@ -149,6 +151,20 @@ Notes:
 - iOS snapshots use XCTest on simulators and physical devices.
 - Scope snapshots with `-s "<label>"` or `-s @ref`.
 - If XCTest returns 0 nodes (e.g., foreground app changed), agent-device fails explicitly.
+- `diff snapshot` uses the same snapshot flags and compares the current capture with the previous session baseline, then updates baseline.
+
+Diff snapshots:
+- Run `diff snapshot` once to initialize baseline for the current session.
+- Run `diff snapshot` again after UI changes to get unified-style output (`-` removed, `+` added, unchanged context).
+- Use `--json` to get `{ mode, baselineInitialized, summary, lines }`.
+
+Efficient snapshot usage:
+- Default to `snapshot -i` for iterative agent loops.
+- Add `-s "<label>"` (or `-s @ref`) for screen-local work to reduce payload size.
+- Add `-d <depth>` when lower tree levels are not needed.
+- Re-snapshot after UI mutations before reusing refs.
+- Use `diff snapshot` for low-noise structural change verification between adjacent states.
+- Reserve `--raw` for troubleshooting and parser/debug investigations.
 
 Flags:
 - `--version, -V` print version and exit

skills/agent-device/SKILL.md

@@ -64,10 +64,16 @@ agent-device snapshot -c               # Compact output
 agent-device snapshot -d 3             # Limit depth
 agent-device snapshot -s "Camera"      # Scope to label/identifier
 agent-device snapshot --raw            # Raw node output
+agent-device diff snapshot             # Structural diff against previous session baseline
 ```
 
 XCTest is the iOS snapshot engine: fast, complete, and no Accessibility permission required.
 
+Snapshot diff notes:
+- First `diff snapshot` call initializes baseline for the current session.
+- Subsequent `diff snapshot` calls compare current UI to prior baseline and then update baseline.
+- Use this for compact change tracking between adjacent UI states.
+
 ### Find (semantic)
 
 ```bash
@@ -241,6 +247,11 @@ agent-device apps --platform android --user-installed
 - Snapshot refs are the core mechanism for interactive agent flows.
 - Use selectors for deterministic replay artifacts and assertions (e.g. in e2e test workflows).
 - Prefer `snapshot -i` to reduce output size.
+- Prefer scoped snapshots (`-s "<label>"` or `-s @ref`) for screen-local tasks.
+- Add `-d <depth>` when only upper tree levels matter; avoid full-tree snapshots by default.
+- Use `diff snapshot` after mutations to detect structural changes with less output than full re-read.
+- Refresh refs immediately after navigation/modal/list mutations before issuing next ref-targeted action.
+- Use `--raw` only for debugging parser/tree edge-cases; avoid it for normal agent loops due to size.
 - On iOS, snapshots use XCTest and do not require Accessibility permission.
 - If XCTest returns 0 nodes (foreground app changed), treat it as an explicit failure and retry the flow/app state.
 - `open <app|url> [url]` can be used within an existing session to switch apps or open deep links.

skills/agent-device/references/snapshot-refs.md

@@ -51,6 +51,22 @@ agent-device snapshot -i -s "Camera"
 agent-device snapshot -i -s @e3
 ```
 
+## Diff snapshots (structural)
+
+Use `diff snapshot` when you need compact state-change visibility between nearby UI states:
+
+```bash
+agent-device diff snapshot           # First run initializes baseline
+agent-device press @e5
+agent-device diff snapshot           # Shows +/− structural lines vs prior baseline
+```
+
+Efficient pattern:
+- Initialize once at a stable point.
+- Mutate UI (`press`, `fill`, `swipe`).
+- Run `diff snapshot` to confirm expected change shape.
+- Re-run full/scoped `snapshot` only when you need fresh refs for next step selection.
+
 ## Troubleshooting
 
 - Ref not found: re-snapshot.

src/__tests__/cli-diff.test.ts

@@ -0,0 +1,96 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import { runCli } from '../cli.ts';
+import type { DaemonRequest, DaemonResponse } from '../daemon-client.ts';
+
+class ExitSignal extends Error {
+  public readonly code: number;
+
+  constructor(code: number) {
+    super(`EXIT_${code}`);
+    this.code = code;
+  }
+}
+
+type RunResult = {
+  code: number | null;
+  stdout: string;
+  stderr: string;
+  calls: Omit<DaemonRequest, 'token'>[];
+};
+
+async function runCliCapture(argv: string[]): Promise<RunResult> {
+  let stdout = '';
+  let stderr = '';
+  let code: number | null = null;
+  const calls: Array<Omit<DaemonRequest, 'token'>> = [];
+
+  const originalExit = process.exit;
+  const originalStdoutWrite = process.stdout.write.bind(process.stdout);
+  const originalStderrWrite = process.stderr.write.bind(process.stderr);
+
+  (process as any).exit = ((nextCode?: number) => {
+    throw new ExitSignal(nextCode ?? 0);
+  }) as typeof process.exit;
+  (process.stdout as any).write = ((chunk: unknown) => {
+    stdout += String(chunk);
+    return true;
+  }) as typeof process.stdout.write;
+  (process.stderr as any).write = ((chunk: unknown) => {
+    stderr += String(chunk);
+    return true;
+  }) as typeof process.stderr.write;
+
+  const sendToDaemon = async (req: Omit<DaemonRequest, 'token'>): Promise<DaemonResponse> => {
+    calls.push(req);
+    return {
+      ok: true,
+      data: {
+        mode: 'snapshot',
+        baselineInitialized: false,
+        summary: { additions: 1, removals: 1, unchanged: 1 },
+        lines: [
+          { kind: 'unchanged', text: '@e2 [window]' },
+          { kind: 'removed', text: '  @e3 [text] "67"' },
+          { kind: 'added', text: '  @e3 [text] "134"' },
+        ],
+      },
+    };
+  };
+
+  try {
+    await runCli(argv, { sendToDaemon });
+  } catch (error) {
+    if (error instanceof ExitSignal) code = error.code;
+    else throw error;
+  } finally {
+    process.exit = originalExit;
+    process.stdout.write = originalStdoutWrite;
+    process.stderr.write = originalStderrWrite;
+  }
+
+  return { code, stdout, stderr, calls };
+}
+
+test('diff snapshot renders human-readable unified diff text', async () => {
+  const result = await runCliCapture(['diff', 'snapshot']);
+  assert.equal(result.code, null);
+  assert.equal(result.calls.length, 1);
+  assert.match(result.stdout, /^@e2 \[window\]/m);
+  assert.match(result.stdout, /^-  @e3 \[text\] "67"$/m);
+  assert.match(result.stdout, /^\+  @e3 \[text\] "134"$/m);
+  assert.match(result.stdout, /1 additions, 1 removals, 1 unchanged/);
+  assert.equal(result.stderr, '');
+});
+
+test('diff snapshot --json passes daemon payload through unchanged', async () => {
+  const result = await runCliCapture(['diff', 'snapshot', '--json']);
+  assert.equal(result.code, null);
+  assert.equal(result.calls.length, 1);
+  const payload = JSON.parse(result.stdout);
+  assert.equal(payload.success, true);
+  assert.equal(payload.data.mode, 'snapshot');
+  assert.equal(payload.data.baselineInitialized, false);
+  assert.equal(Array.isArray(payload.data.lines), true);
+  assert.equal(result.stderr, '');
+});

src/cli.ts

@@ -1,6 +1,6 @@
 import { parseArgs, toDaemonFlags, usage, usageForCommand } from './utils/args.ts';
 import { asAppError, AppError, normalizeError } from './utils/errors.ts';
-import { formatSnapshotText, printHumanError, printJson } from './utils/output.ts';
+import { formatSnapshotDiffText, formatSnapshotText, printHumanError, printJson } from './utils/output.ts';
 import { readVersion } from './utils/version.ts';
 import { pathToFileURL } from 'node:url';
 import { sendToDaemon } from './daemon-client.ts';
@@ -189,6 +189,11 @@ export async function runCli(argv: string[], deps: CliDeps = DEFAULT_CLI_DEPS):
         if (logTailStopper) logTailStopper();
         return;
       }
+      if (command === 'diff' && positionals[0] === 'snapshot') {
+        process.stdout.write(formatSnapshotDiffText((response.data ?? {}) as Record<string, unknown>));
+        if (logTailStopper) logTailStopper();
+        return;
+      }
       if (command === 'get') {
         const sub = positionals[0];
         if (sub === 'text') {

src/core/__tests__/capabilities.test.ts

@@ -54,6 +54,7 @@ test('core commands support iOS simulator, iOS device, and Android', () => {
     'boot',
     'click',
     'close',
+    'diff',
     'fill',
     'find',
     'focus',

src/core/capabilities.ts

@@ -23,6 +23,7 @@ const COMMAND_CAPABILITY_MATRIX: Record<string, CommandCapability> = {
   click: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },
   close: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },
   fill: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },
+  diff: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },
   find: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },
   focus: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },
   get: { ios: { simulator: true, device: true }, android: { emulator: true, device: true, unknown: true } },

src/daemon/__tests__/snapshot-diff.test.ts

@@ -0,0 +1,99 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import { attachRefs, type RawSnapshotNode } from '../../utils/snapshot.ts';
+import { buildSnapshotDiff } from '../snapshot-diff.ts';
+
+function nodes(raw: RawSnapshotNode[]) {
+  return attachRefs(raw);
+}
+
+test('buildSnapshotDiff reports unchanged lines when snapshots are equal', () => {
+  const previous = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeButton', label: 'Increment' },
+  ]);
+  const current = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeButton', label: 'Increment' },
+  ]);
+
+  const diff = buildSnapshotDiff(previous, current);
+  assert.equal(diff.summary.additions, 0);
+  assert.equal(diff.summary.removals, 0);
+  assert.equal(diff.summary.unchanged, 2);
+  assert.deepEqual(diff.lines.map((line) => line.kind), ['unchanged', 'unchanged']);
+});
+
+test('buildSnapshotDiff reports added and removed lines', () => {
+  const previous = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeStaticText', label: '67' },
+    { index: 2, depth: 1, type: 'XCUIElementTypeButton', label: 'Increment' },
+  ]);
+  const current = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeStaticText', label: '134' },
+    { index: 2, depth: 1, type: 'XCUIElementTypeButton', label: 'Increment' },
+  ]);
+
+  const diff = buildSnapshotDiff(previous, current);
+  assert.equal(diff.summary.additions, 1);
+  assert.equal(diff.summary.removals, 1);
+  assert.equal(diff.summary.unchanged, 2);
+  assert.deepEqual(diff.lines.map((line) => line.kind), ['unchanged', 'removed', 'added', 'unchanged']);
+});
+
+test('buildSnapshotDiff treats value changes as remove plus add', () => {
+  const previous = nodes([{ index: 0, depth: 0, type: 'XCUIElementTypeTextField', label: 'Amount', value: '67' }]);
+  const current = nodes([{ index: 0, depth: 0, type: 'XCUIElementTypeTextField', label: 'Amount', value: '134' }]);
+
+  const diff = buildSnapshotDiff(previous, current);
+  assert.equal(diff.summary.additions, 1);
+  assert.equal(diff.summary.removals, 1);
+  assert.equal(diff.summary.unchanged, 0);
+  assert.deepEqual(diff.lines.map((line) => line.kind), ['removed', 'added']);
+});
+
+test('buildSnapshotDiff preserves surrounding context ordering', () => {
+  const previous = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeStaticText', label: 'Count' },
+    { index: 2, depth: 1, type: 'XCUIElementTypeStaticText', label: '67' },
+    { index: 3, depth: 1, type: 'XCUIElementTypeButton', label: 'Increment' },
+  ]);
+  const current = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeStaticText', label: 'Count' },
+    { index: 2, depth: 1, type: 'XCUIElementTypeStaticText', label: '134' },
+    { index: 3, depth: 1, type: 'XCUIElementTypeButton', label: 'Increment' },
+  ]);
+
+  const diff = buildSnapshotDiff(previous, current);
+  assert.equal(diff.lines[0]?.kind, 'unchanged');
+  assert.equal(diff.lines[1]?.kind, 'unchanged');
+  assert.equal(diff.lines[2]?.kind, 'removed');
+  assert.equal(diff.lines[3]?.kind, 'added');
+  assert.equal(diff.lines[4]?.kind, 'unchanged');
+});
+
+test('buildSnapshotDiff flatten option uses flat snapshot line shape', () => {
+  const previous = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeOther', label: '335' },
+    { index: 2, depth: 2, type: 'XCUIElementTypeStaticText', label: '335' },
+  ]);
+  const current = nodes([
+    { index: 0, depth: 0, type: 'XCUIElementTypeWindow' },
+    { index: 1, depth: 1, type: 'XCUIElementTypeOther', label: '402' },
+    { index: 2, depth: 2, type: 'XCUIElementTypeStaticText', label: '402' },
+  ]);
+
+  const diff = buildSnapshotDiff(previous, current, { flatten: true });
+  assert.equal(diff.summary.additions, 2);
+  assert.equal(diff.summary.removals, 2);
+  const changed = diff.lines.filter((line) => line.kind !== 'unchanged');
+  assert.equal(changed.length, 4);
+  for (const line of changed) {
+    assert.equal(line.text.startsWith('  '), false);
+  }
+});