dsh-builtin-browser 0.4.0 → 0.4.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (44) hide show
  1. package/README.en.md +68 -30
  2. package/README.md +66 -30
  3. package/client.js +60 -6
  4. package/cordis.patch.yml +21 -0
  5. package/desktop-bridge/plugin-browser-bridge.js +12 -4
  6. package/docs/architecture.md +12 -2
  7. package/docs/tool-reference.md +17 -17
  8. package/docs/user-guide.md +13 -4
  9. package/docs/why-browser.md +2 -2
  10. package/lib/browser/types.d.ts +25 -1
  11. package/lib/browser-electron/bridge-connection.d.ts +31 -5
  12. package/lib/browser-electron/bridge-connection.js +143 -21
  13. package/lib/browser-electron/entry.d.ts +1 -1
  14. package/lib/browser-electron/entry.js +61 -35
  15. package/lib/browser-electron/ephemeral-profile.d.ts +37 -0
  16. package/lib/browser-electron/ephemeral-profile.js +146 -0
  17. package/lib/browser-electron/history-store.d.ts +25 -0
  18. package/lib/browser-electron/history-store.js +51 -4
  19. package/lib/browser-electron/host-main.js +75 -27
  20. package/lib/browser-electron/provider.d.ts +44 -1
  21. package/lib/browser-electron/provider.js +379 -48
  22. package/lib/browser-electron/remote-host.d.ts +52 -3
  23. package/lib/browser-electron/remote-host.js +244 -23
  24. package/lib/browser-electron/settings-route.d.ts +28 -0
  25. package/lib/browser-electron/settings-route.js +28 -0
  26. package/lib/browser-electron/settings-store.d.ts +40 -0
  27. package/lib/browser-electron/settings-store.js +107 -11
  28. package/lib/browser-electron/system-browser.d.ts +71 -3
  29. package/lib/browser-electron/system-browser.js +355 -38
  30. package/lib/tool-browser/index.js +143 -34
  31. package/package.json +5 -3
  32. package/scripts/verify-host-compat.mjs +195 -0
  33. package/src/browser/types.ts +25 -1
  34. package/src/browser-electron/bridge-connection.ts +142 -21
  35. package/src/browser-electron/entry.ts +58 -33
  36. package/src/browser-electron/ephemeral-profile.ts +152 -0
  37. package/src/browser-electron/history-store.ts +51 -4
  38. package/src/browser-electron/host-main.ts +105 -35
  39. package/src/browser-electron/provider.ts +401 -51
  40. package/src/browser-electron/remote-host.ts +261 -36
  41. package/src/browser-electron/settings-route.ts +46 -0
  42. package/src/browser-electron/settings-store.ts +121 -11
  43. package/src/browser-electron/system-browser.ts +342 -21
  44. package/src/tool-browser/index.ts +144 -36
@@ -8,7 +8,7 @@
8
8
  */
9
9
 
10
10
  import { randomUUID } from 'node:crypto'
11
- import { existsSync, mkdirSync, writeFileSync } from 'node:fs'
11
+ import { existsSync, lstatSync, mkdirSync, realpathSync, writeFileSync } from 'node:fs'
12
12
  import { homedir } from 'node:os'
13
13
  import { dirname, isAbsolute, join, resolve, sep } from 'node:path'
14
14
  import { HistoryStore, type VisitedPage } from './history-store.js'
@@ -306,7 +306,16 @@ export interface ElectronBrowserProviderConfig {
306
306
  readonly settings?: () => BrowserSettings
307
307
  /** Maximum snapshot elements before truncation. Default 60. */
308
308
  readonly snapshotMaxElements?: number
309
- /** Maximum content characters before truncation when no maxChars is given. Default 100_000. */
309
+ /**
310
+ * Explicit content cap, overriding the per-format default when set.
311
+ *
312
+ * Honoured again, and it is worth saying why that is a fix rather than a feature: this was
313
+ * `config.contentMaxChars ?? 100_000`, one number for every format, and when the caps became
314
+ * per-format (html and json 50 000, otherwise 20 000) the field was left assigned and never
315
+ * read. A configuration knob that silently does nothing is worse than no knob — it is an
316
+ * operator's decision that quietly went nowhere. Leaving it unset means the per-format
317
+ * default; setting it means the operator meant it.
318
+ */
310
319
  readonly contentMaxChars?: number
311
320
  /**
312
321
  * Directory `browser_download` save paths must resolve inside (prevents a
@@ -524,6 +533,57 @@ function defaultDownloadDir(): string {
524
533
  return join(homedir(), DOWNLOAD_DIR_NAMES[0] ?? 'Downloads')
525
534
  }
526
535
 
536
+ /**
537
+ * Whether a directory entry exists at `path`, judging the ENTRY rather than what it
538
+ * points at.
539
+ *
540
+ * `existsSync` follows links, so a DANGLING symlink — a link whose target does not
541
+ * exist — reads as "nothing here", and a write to that path creates the target
542
+ * instead: outside `downloadDir` whenever the link points out of it. `lstatSync`
543
+ * sees the link itself, which is the only correct answer to "is this name free".
544
+ * @param path - the path to judge.
545
+ * @returns true when any entry (file, directory, link) occupies the name.
546
+ */
547
+ function entryExists(path: string): boolean {
548
+ try {
549
+ lstatSync(path)
550
+ return true
551
+ } catch {
552
+ return false
553
+ }
554
+ }
555
+
556
+ /**
557
+ * The real path of the deepest existing ancestor of `path` — `path` itself when it
558
+ * exists, otherwise the nearest directory above it that does.
559
+ *
560
+ * `resolve()` normalizes `..` but never follows a link, so nothing textual can tell
561
+ * that `<downloadDir>/link/x.png` leaves the directory through `link`. The first
562
+ * ancestor that exists (judged by lstat, so a dangling link counts — the write would
563
+ * go through it) is the last component whose real location still determines where
564
+ * the bytes land, so that is the one to resolve and check.
565
+ * @param path - an absolute path that may not exist yet.
566
+ * @returns the resolved real path of its deepest existing ancestor.
567
+ */
568
+ function realAncestorOf(path: string): string {
569
+ let current = path
570
+ for (let depth = 0; depth < 64; depth += 1) {
571
+ if (entryExists(current)) {
572
+ try {
573
+ return realpathSync(current)
574
+ } catch {
575
+ // The entry vanished between the two calls, or cannot be resolved: the
576
+ // textual path is the best available answer, and the write reports the rest.
577
+ return current
578
+ }
579
+ }
580
+ const parent = dirname(current)
581
+ if (parent === current) return current
582
+ current = parent
583
+ }
584
+ return current
585
+ }
586
+
527
587
  /**
528
588
  * Browser provider over Electron views. Sessions hold an ordered list of
529
589
  * tabs; each tab is one view created by the host. The active tab receives
@@ -537,7 +597,7 @@ export class ElectronBrowserProvider implements BrowserProvider {
537
597
  private readonly sessions = new Map<BrowserSessionId, Session>()
538
598
  private readonly httpOnly: boolean
539
599
  private readonly snapshotMaxElements: number
540
- private readonly contentMaxChars: number
600
+ private readonly contentMaxChars: number | undefined
541
601
  private readonly downloadDir: string | undefined
542
602
  /**
543
603
  * Persistent browsing history, or `undefined` when the user turned it off.
@@ -553,7 +613,7 @@ export class ElectronBrowserProvider implements BrowserProvider {
553
613
  ) {
554
614
  this.httpOnly = config.httpOnly ?? true
555
615
  this.snapshotMaxElements = config.snapshotMaxElements ?? 60
556
- this.contentMaxChars = config.contentMaxChars ?? 100_000
616
+ this.contentMaxChars = config.contentMaxChars
557
617
  // Confine downloads AND screenshot saves to a dedicated directory by
558
618
  // default: the OS Downloads folder is the human-visible, browser-natural
559
619
  // place for written files, and it is the one directory the DSH file
@@ -850,13 +910,16 @@ export class ElectronBrowserProvider implements BrowserProvider {
850
910
  * Chromium drops `Input.*` events for a view with no display surface, so this
851
911
  * runs before click/type/key and fails loudly (BROWSER_VIEW_NOT_PRESENTED)
852
912
  * rather than reporting a success the page never saw.
913
+ * @param s - the session whose active view must be presented.
914
+ * @param signal - optional cancellation.
915
+ * @returns the view the session's active tab has AFTER the presentation barrier.
853
916
  */
854
- private async present(s: Session, signal?: AbortSignal): Promise<void> {
917
+ private async present(s: Session, signal?: AbortSignal): Promise<ElectronViewHandle> {
855
918
  const { handle } = this.activeTab(s)
856
919
  const present = this.host.presentView?.bind(this.host)
857
920
  if (present === undefined) {
858
921
  this.showActive(s)
859
- return
922
+ return this.activeTab(s).handle
860
923
  }
861
924
  signal?.throwIfAborted()
862
925
  const timeoutMs = 10_000
@@ -875,6 +938,48 @@ export class ElectronBrowserProvider implements BrowserProvider {
875
938
  { cause: error },
876
939
  )
877
940
  }
941
+ // Read the active tab AGAIN. `present` is a round-trip that can take up to the budget
942
+ // above, and the human is invited to drive the same window (a toolbar tab switch, a
943
+ // closed window) while the agent waits. Chromium answers `Input.*` with success even
944
+ // for a view that has no display surface, so dispatching to a handle that is no longer
945
+ // the visible one does not fail — it silently lands nowhere, or on the wrong page.
946
+ //
947
+ // Returning the re-read handle was wrong, though: `handle` above is the one that was
948
+ // actually presented, and if the active tab changed during the round trip then the
949
+ // re-read one was never presented at all. Handing it back meant the caller dispatched
950
+ // to a view with no display surface — the exact failure this method exists to prevent,
951
+ // arrived at from the other side. Either the presented view is still the active one, or
952
+ // the caller must be told rather than given something else to aim at.
953
+ const live = this.activeTab(s).handle
954
+ if (live !== handle) {
955
+ throw new BrowserError(
956
+ 'browser: the active tab changed while the page view was being presented, so synthesized input would go to an unshown view — retry, and the new page will be presented instead',
957
+ 'BROWSER_VIEW_CHANGED',
958
+ )
959
+ }
960
+ return handle
961
+ }
962
+
963
+ /**
964
+ * Refuse to send synthesized input to a view that is no longer the one it was located
965
+ * against.
966
+ *
967
+ * `locateHandle` was captured before a page-side locate that can consume its whole
968
+ * 10s budget; the session's active tab is re-read afterwards. If they differ, the human
969
+ * switched tabs mid-flight (the product's whole point is that they can take over), and
970
+ * the coordinates belong to one page while the dispatch would go to another. Comparing
971
+ * the handles turns that into a loud, retryable error instead of input the caller
972
+ * believes landed.
973
+ * @param session - the session id, for the message.
974
+ * @param locateHandle - the view the locate ran in.
975
+ * @param liveHandle - the view that is active now.
976
+ */
977
+ private assertSameView(session: BrowserSessionId, locateHandle: ElectronViewHandle, liveHandle: ElectronViewHandle): void {
978
+ if (locateHandle === liveHandle) return
979
+ throw new BrowserError(
980
+ `browser: the session's active view changed while the element was being located (session ${session}); the operation was not dispatched — retry it`,
981
+ 'BROWSER_VIEW_CHANGED',
982
+ )
878
983
  }
879
984
 
880
985
  /**
@@ -903,6 +1008,7 @@ export class ElectronBrowserProvider implements BrowserProvider {
903
1008
 
904
1009
  /** Execute JS in the active tab's page context. */
905
1010
  async execute(session: BrowserSessionId, request: BrowserExecuteRequest, signal?: AbortSignal): Promise<BrowserExecuteResult> {
1011
+ this.assertActionAllowed('execute')
906
1012
  const s = this.session(session)
907
1013
  const { handle } = this.activeTab(s)
908
1014
  signal?.throwIfAborted()
@@ -1140,7 +1246,9 @@ export class ElectronBrowserProvider implements BrowserProvider {
1140
1246
  const tab = this.activeTab(s)
1141
1247
  signal?.throwIfAborted()
1142
1248
  const includeHidden = request.includeHidden === true
1143
- const maxNodes = Math.max(10, Math.min(5000, Math.floor(request.maxNodes ?? 500)))
1249
+ // 150 rather than 500. The cap fills on any real page, and 500 nodes measured 38,953
1250
+ // characters — 10k-13k tokens for a single call. A caller that needs more asks for it.
1251
+ const maxNodes = Math.max(10, Math.min(5000, Math.floor(request.maxNodes ?? 150)))
1144
1252
  const script = `(() => {
1145
1253
  const includeHidden = ${String(includeHidden)}
1146
1254
  const maxNodes = ${String(maxNodes)}
@@ -1282,6 +1390,11 @@ export class ElectronBrowserProvider implements BrowserProvider {
1282
1390
  value: valueOf(el),
1283
1391
  states: statesOf(el),
1284
1392
  depth: baseDepth + depthOf(el),
1393
+ // Same derivation as the snapshot collector: an id or a name is what makes a node
1394
+ // something the caller can point at without guessing at its text. The tool-layer
1395
+ // schema declared this field before this line existed, so the schema was promising
1396
+ // a value the tool never sent.
1397
+ selector: el.id ? '#' + CSS.escape(el.id) : (el.name ? '[name=' + JSON.stringify(el.name) + ']' : ''),
1285
1398
  tag: el.tagName.toLowerCase(),
1286
1399
  x: Math.round(a.x + a.w / 2),
1287
1400
  y: Math.round(a.y + a.h / 2),
@@ -1330,9 +1443,18 @@ export class ElectronBrowserProvider implements BrowserProvider {
1330
1443
  async content(session: BrowserSessionId, request: BrowserContentRequest, signal?: AbortSignal): Promise<BrowserContentResult> {
1331
1444
  const tab = this.activeTab(this.session(session))
1332
1445
  signal?.throwIfAborted()
1333
- const maxChars = request.maxChars ?? this.contentMaxChars
1334
1446
  const selector = request.selector ?? ''
1335
1447
  const format = request.format
1448
+ // Per-format caps. A full HTML document is enormous, but the same limit applied to
1449
+ // plain text is no limit at all (a whole page of text measured 25,882 characters).
1450
+ // This bounds the common case instead of letting one call spend 25k-33k tokens.
1451
+ //
1452
+ // The plugin config overrides the lot when it is set — see `contentMaxChars`. The caller's
1453
+ // own `maxChars` still outranks both: an explicit per-call bound is the most specific
1454
+ // statement of intent there is.
1455
+ const maxChars = request.maxChars
1456
+ ?? this.contentMaxChars
1457
+ ?? (format === 'html' || format === 'json' ? 50_000 : 20_000)
1336
1458
  const script = `(() => {
1337
1459
  const root = ${selector === '' ? 'document.body' : `document.querySelector(${JSON.stringify(selector)})`}
1338
1460
  if (!root) return { ok: false, reason: 'selector not found' }
@@ -1376,7 +1498,12 @@ export class ElectronBrowserProvider implements BrowserProvider {
1376
1498
  }
1377
1499
  }
1378
1500
  else if (fmt === 'html') content = root.outerHTML || ''
1379
- else if (fmt === 'json') content = JSON.stringify(root)
1501
+ // A DOM element has no own enumerable properties — everything lives on the prototype
1502
+ // chain — so JSON.stringify(root) is always "{}". That is not a structured view of
1503
+ // the page, it is an empty object the caller will read as "no data here" and then
1504
+ // retry with other selectors. Serialise the markup instead, which is what a caller
1505
+ // asking for JSON of an element can actually use.
1506
+ else if (fmt === 'json') content = JSON.stringify({ html: root.outerHTML, tag: root.tagName.toLowerCase() })
1380
1507
  else {
1381
1508
  // markdown: headings, paragraphs, links, lists (best-effort)
1382
1509
  const parts = []
@@ -1431,7 +1558,9 @@ export class ElectronBrowserProvider implements BrowserProvider {
1431
1558
  signal?: AbortSignal,
1432
1559
  ): Promise<void> {
1433
1560
  const s = this.session(session)
1434
- const { handle } = this.activeTab(s)
1561
+ // The view the locate runs against. It can stop being the session's active view while
1562
+ // the locate waits (see prepareTarget), so every dispatch below re-reads the live one.
1563
+ const locateHandle = this.activeTab(s).handle
1435
1564
  signal?.throwIfAborted()
1436
1565
  let x = 0
1437
1566
  let y = 0
@@ -1465,8 +1594,13 @@ export class ElectronBrowserProvider implements BrowserProvider {
1465
1594
  y = request.y
1466
1595
  }
1467
1596
  // Input.* is only delivered to a view with a current display surface; the
1468
- // barrier also re-presents after a navigation replaced the renderer.
1469
- await this.present(s, signal)
1597
+ // barrier also re-presents after a navigation replaced the renderer. It returns the
1598
+ // handle that may still be receiving input when it settles.
1599
+ const handle = await this.present(s, signal)
1600
+ // The locate (up to 10s) and the present barrier both ran against the view captured
1601
+ // earlier; if the human switched tabs meanwhile, these coordinates belong to one page
1602
+ // and the dispatch would go to another — and Chromium answers success either way.
1603
+ if ('target' in request) this.assertSameView(session, locateHandle, handle)
1470
1604
  const timeoutMs = 30_000
1471
1605
  const send = (params: Record<string, unknown>, label: string): Promise<unknown> =>
1472
1606
  withTimeout(
@@ -1493,6 +1627,10 @@ export class ElectronBrowserProvider implements BrowserProvider {
1493
1627
  // tab" signal, so the bubble is what lets a human follow along without the log.
1494
1628
  this.showCursor(handle, x, y, 'click', 'click', true)
1495
1629
  const release = (): void => {
1630
+ // Back to the handle the PRESS may have landed on, deliberately not to the currently
1631
+ // active tab: this only runs after a failed press/release, and re-targeting a
1632
+ // compensating half of a click to a view the human just switched to would press a
1633
+ // button there. Best-effort either way.
1496
1634
  void handle
1497
1635
  .sendCommand('Input.dispatchMouseEvent', {
1498
1636
  type: 'mouseReleased',
@@ -1532,7 +1670,10 @@ export class ElectronBrowserProvider implements BrowserProvider {
1532
1670
  signal?: AbortSignal,
1533
1671
  ): Promise<void> {
1534
1672
  const s = this.session(session)
1535
- const { handle } = this.activeTab(s)
1673
+ // The view the locate runs against. It can stop being the session's active view while
1674
+ // the locate waits (the human is invited to drive the same window), so the input below
1675
+ // goes to the handle `present` reports instead of this one.
1676
+ const locateHandle = this.activeTab(s).handle
1536
1677
  signal?.throwIfAborted()
1537
1678
  const hasTarget = 'target' in request
1538
1679
  if (hasTarget) {
@@ -1549,8 +1690,13 @@ export class ElectronBrowserProvider implements BrowserProvider {
1549
1690
  const text = 'text' in request ? request.text : ''
1550
1691
  const timeoutMs = 30_000
1551
1692
  // Input.insertText goes through the Input domain: same display-surface
1552
- // requirement as click/key.
1553
- await this.present(s, signal)
1693
+ // requirement as click/key. It also reports which view is active once the barrier has
1694
+ // settled, which is what makes the tab-switch check below meaningful.
1695
+ const handle = await this.present(s, signal)
1696
+ // Focusing one page and inserting into another is exactly the silent mis-target this
1697
+ // refuses: the element was focused in the located view, and the text would have gone to
1698
+ // whichever page is now visible — possibly a password field there.
1699
+ if (hasTarget) this.assertSameView(session, locateHandle, handle)
1554
1700
  try {
1555
1701
  await withTimeout(
1556
1702
  handle.sendCommand('Input.insertText', { text } satisfies CdpInsertTextParams),
@@ -1674,7 +1820,6 @@ export class ElectronBrowserProvider implements BrowserProvider {
1674
1820
  /** Press one named key (Enter/Tab/arrows/…) via CDP key events. */
1675
1821
  async key(session: BrowserSessionId, request: BrowserKeyRequest, signal?: AbortSignal): Promise<void> {
1676
1822
  const s = this.session(session)
1677
- const { handle } = this.activeTab(s)
1678
1823
  signal?.throwIfAborted()
1679
1824
  const spec = KEY_SPECS[request.key]
1680
1825
  if (spec === undefined) {
@@ -1682,8 +1827,13 @@ export class ElectronBrowserProvider implements BrowserProvider {
1682
1827
  }
1683
1828
  const params = { key: spec.key, code: spec.code, windowsVirtualKeyCode: spec.vk, nativeVirtualKeyCode: spec.vk, ...spec.text !== undefined ? { text: spec.text } : {} }
1684
1829
  const timeoutMs = 15_000
1685
- // Input.dispatchKeyEvent is subject to the same display-surface rule.
1686
- await this.present(s, signal)
1830
+ // Input.dispatchKeyEvent is subject to the same display-surface rule. The handle comes
1831
+ // back from the presentation barrier rather than from a capture taken before it: this
1832
+ // session is shared with the human, who may switch tabs while the barrier round-trips,
1833
+ // and a key sent to the view they left is silently dropped (or lands on the new page).
1834
+ const handle = await this.present(s, signal)
1835
+ // Focus the SAME view the key is dispatched to; focusing the pre-switch handle would
1836
+ // grant web focus to a page that then receives nothing.
1687
1837
  await this.focusView(handle)
1688
1838
  const release = (): Promise<Record<string, unknown>> =>
1689
1839
  handle.sendCommand('Input.dispatchKeyEvent', { type: 'keyUp', ...params })
@@ -1721,7 +1871,13 @@ export class ElectronBrowserProvider implements BrowserProvider {
1721
1871
  name: f.name ?? null,
1722
1872
  label: f.label ?? null,
1723
1873
  placeholder: f.placeholder ?? null,
1724
- kind: f.kind ?? 'text',
1874
+ // A field with NO kind is a field with no kind FILTER: absent, not 'text'. Defaulting
1875
+ // it to 'text' here made `if (spec.kind)` above always true, so every unspecified
1876
+ // field was narrowed to an <input> — browser_fill on a <textarea> reported
1877
+ // `no "text" control matched`, and a contenteditable could not match any kind at all.
1878
+ // The value stays a string or null (never undefined): the in-page check tests truthiness,
1879
+ // and JSON drops an undefined member entirely, which is the same thing here.
1880
+ kind: f.kind ?? null,
1725
1881
  value: f.value,
1726
1882
  })))
1727
1883
  const submitFlag = request.submit === true
@@ -1754,10 +1910,34 @@ export class ElectronBrowserProvider implements BrowserProvider {
1754
1910
  }
1755
1911
  return false
1756
1912
  }
1913
+ // Every root this page exposes: the top document plus its shadow trees (recursively)
1914
+ // and same-origin iframes. The snapshot and waitFor already walk all of them, so an
1915
+ // element inside one is shown to the model and can be waited for — resolving a click or
1916
+ // a fill used to look at the top document alone and answer "not found" for something
1917
+ // the model had just been handed a reference to. The two must agree about what exists.
1918
+ // (Kept in step with the collector in the snapshot script above.)
1919
+ const allRoots = () => {
1920
+ const roots = []
1921
+ const seen = new Set()
1922
+ const visit = doc => {
1923
+ if (seen.has(doc)) return
1924
+ seen.add(doc)
1925
+ roots.push(doc)
1926
+ for (const el of doc.querySelectorAll('*')) {
1927
+ if (el.shadowRoot) visit(el.shadowRoot)
1928
+ if (el.tagName === 'IFRAME') {
1929
+ try { const d = el.contentDocument; if (d) visit(d) } catch { /* cross-origin */ }
1930
+ }
1931
+ }
1932
+ }
1933
+ visit(document)
1934
+ return roots
1935
+ }
1936
+ const roots = allRoots()
1757
1937
  const candidates = (spec) => {
1758
1938
  const all = spec.selector
1759
- ? [...document.querySelectorAll(spec.selector)]
1760
- : [...document.querySelectorAll('input, textarea, select, [contenteditable="true"]')].filter(el => matches(el, spec))
1939
+ ? roots.flatMap(root => [...root.querySelectorAll(spec.selector)])
1940
+ : roots.flatMap(root => [...root.querySelectorAll('input, textarea, select, [contenteditable="true"]')]).filter(el => matches(el, spec))
1761
1941
  const vis = all.filter(visible)
1762
1942
  return vis.length > 0 ? vis : all
1763
1943
  }
@@ -1775,7 +1955,9 @@ export class ElectronBrowserProvider implements BrowserProvider {
1775
1955
  let el = els[0]
1776
1956
  // The caller may name a kind. Honor it: the branches below read the element's own
1777
1957
  // tag/type, so without this a selector matching two different controls acted on
1778
- // whichever came first — usually right, which is what hid the mistake.
1958
+ // whichever came first — usually right, which is what hid the mistake. A field that
1959
+ // names NO kind gets no filter at all: the element's own tag then decides the
1960
+ // branch, which is what "kind defaults to text" was always trying to say.
1779
1961
  if (spec.kind) {
1780
1962
  const matchesKind = (node) => {
1781
1963
  const nodeTag = node.tagName
@@ -1783,6 +1965,12 @@ export class ElectronBrowserProvider implements BrowserProvider {
1783
1965
  if (spec.kind === 'select') return nodeTag === 'SELECT'
1784
1966
  if (spec.kind === 'textarea') return nodeTag === 'TEXTAREA'
1785
1967
  if (spec.kind === 'checkbox' || spec.kind === 'radio') return nodeType === spec.kind
1968
+ // 'text': the plain-value branch covers <input> AND <textarea>, and the batch
1969
+ // already collects [contenteditable="true"] as a settable control — so all three
1970
+ // must match. Requiring nodeTag === 'INPUT' here reported "no text control
1971
+ // matched" for a textarea and left the contenteditable branch below unreachable.
1972
+ if (nodeTag === 'TEXTAREA') return true
1973
+ if (node.isContentEditable === true) return true
1786
1974
  return nodeTag === 'INPUT' && nodeType !== 'checkbox' && nodeType !== 'radio' && nodeType !== 'file'
1787
1975
  }
1788
1976
  const wanted = els.find(matchesKind)
@@ -1904,33 +2092,78 @@ export class ElectronBrowserProvider implements BrowserProvider {
1904
2092
  const match = () => {
1905
2093
  const by = typeof spec.by === 'string' ? spec.by : 'css'
1906
2094
  const value = String(spec.value ?? '')
1907
- const index = typeof spec.index === 'number' ? spec.index : 0
2095
+ // Named matchIndex rather than index: this whole function is one host-side template
2096
+ // literal, so a bare name declared here can silently shadow a host binding of the
2097
+ // same name and turn an inner reference into a type error.
2098
+ const matchIndex = typeof spec.index === 'number' ? spec.index : 0
2099
+ // An index that cannot address an element can never match on a later poll, exactly
2100
+ // like a selector that fails to parse — so it must not look like a miss either. A
2101
+ // negative index, the common "last one" shorthand, otherwise burned the caller's
2102
+ // whole locate budget (10s) and then reported "element not found", which reads like a
2103
+ // slow page rather than a bad argument.
2104
+ // (No backticks in this comment: it lives inside a template literal.)
2105
+ if (!Number.isInteger(matchIndex) || matchIndex < 0) {
2106
+ throw new Error('target.index must be a non-negative integer, got ' + JSON.stringify(spec.index))
2107
+ }
1908
2108
  let els = []
2109
+ // Every root this page exposes: the top document plus its shadow trees (recursively)
2110
+ // and same-origin iframes. The snapshot and waitFor already walk all of them, so an
2111
+ // element inside one is shown to the model and can be waited for — resolving a click or
2112
+ // a fill used to look at the top document alone and answer "not found" for something
2113
+ // the model had just been handed a reference to. The two must agree about what exists.
2114
+ // (Kept in step with the collector in the snapshot script above.)
2115
+ const allRoots = () => {
2116
+ const roots = []
2117
+ const seen = new Set()
2118
+ const visit = doc => {
2119
+ if (seen.has(doc)) return
2120
+ seen.add(doc)
2121
+ roots.push(doc)
2122
+ for (const el of doc.querySelectorAll('*')) {
2123
+ if (el.shadowRoot) visit(el.shadowRoot)
2124
+ if (el.tagName === 'IFRAME') {
2125
+ try { const d = el.contentDocument; if (d) visit(d) } catch { /* cross-origin */ }
2126
+ }
2127
+ }
2128
+ }
2129
+ visit(document)
2130
+ return roots
2131
+ }
2132
+ const roots = allRoots()
1909
2133
  if (by === 'css') {
1910
2134
  // A selector that does not PARSE can never match on a later poll, so
1911
2135
  // it must not look like a miss: the old "return null" kept the caller
1912
2136
  // polling until the whole budget was gone, and "element not found
1913
2137
  // after 10s" reads like a slow page while the real cause is a
1914
2138
  // malformed selector (a label or plain text passed where one belongs).
1915
- try { els = Array.from(document.querySelectorAll(value)) }
2139
+ // Parse on the top document first: a selector that does not PARSE must still be
2140
+ // reported as a bad argument rather than as a miss, and that check is about syntax,
2141
+ // not about which root holds the element.
2142
+ try { document.querySelectorAll(value) }
1916
2143
  catch (error) { throw new Error('invalid CSS selector ' + JSON.stringify(value) + ' (' + String((error && error.message) || error) + ')') }
2144
+ for (const root of roots) els.push(...root.querySelectorAll(value))
1917
2145
  } else if (by === 'xpath') {
1918
2146
  try {
1919
- const snap = document.evaluate(value, document, null, XPathResult.ORDERED_NODE_SNAPSHOT_TYPE, null)
1920
- for (let i = 0; i < snap.snapshotLength; i++) {
1921
- const n = snap.snapshotItem(i)
1922
- if (n instanceof Element) els.push(n)
2147
+ for (const root of roots) {
2148
+ const snap = document.evaluate(value, root, null, XPathResult.ORDERED_NODE_SNAPSHOT_TYPE, null)
2149
+ for (let i = 0; i < snap.snapshotLength; i++) {
2150
+ const n = snap.snapshotItem(i)
2151
+ if (n instanceof Element) els.push(n)
2152
+ }
1923
2153
  }
1924
2154
  } catch (error) { throw new Error('invalid XPath ' + JSON.stringify(value) + ' (' + String((error && error.message) || error) + ')') }
1925
2155
  } else {
1926
- const all = Array.from(document.querySelectorAll('body *'))
2156
+ // '*' rather than 'body *': a shadow root and an iframe document have no body of
2157
+ // their own to anchor to, and the visibility and own-text predicates below are what
2158
+ // actually decide whether an element counts.
2159
+ const all = roots.flatMap(root => [...root.querySelectorAll('*')])
1927
2160
  const exact = all.filter(el => isVisible(el) && ownText(el) === value)
1928
2161
  // Exact first, then contains; deepest element preferred (so a
1929
2162
  // button inside a card wins over the card itself).
1930
2163
  const pool = exact.length > 0 ? exact : all.filter(el => isVisible(el) && ownText(el).includes(value))
1931
2164
  els = pool.sort((a, b) => depth(b) - depth(a))
1932
2165
  }
1933
- return els[index] ?? null
2166
+ return els[matchIndex] ?? null
1934
2167
  }
1935
2168
  const deadline = Date.now() + timeoutMs
1936
2169
  let el = null
@@ -2225,6 +2458,35 @@ export class ElectronBrowserProvider implements BrowserProvider {
2225
2458
  return { count: items.length, items }
2226
2459
  }
2227
2460
 
2461
+ /**
2462
+ * Refuse an action the settings have switched off.
2463
+ *
2464
+ * These switches belong to the OPERATOR. The settings document is written through the
2465
+ * plugin's own panel and no tool can reach it, so — unlike `browser_restrict`, which the
2466
+ * model owns and can lift at will — a refusal here stands for as long as the setting does.
2467
+ * That is the point of them: a deployment can take page-script execution, downloads, or
2468
+ * login-state writes off the table without relying on the model's cooperation.
2469
+ *
2470
+ * Read through `settingsSource` on every call rather than captured at construction, so
2471
+ * flipping a switch applies to the next command instead of the next restart — the same
2472
+ * rule the credentials gate follows.
2473
+ * @param action - which switch to consult.
2474
+ * @throws BrowserError when that switch is off.
2475
+ */
2476
+ private assertActionAllowed(action: 'execute' | 'download' | 'credentialWrite'): void {
2477
+ const settings = this.settingsSource?.()
2478
+ if (settings === undefined) return
2479
+ if (action === 'execute' && !settings.actions.allowExecute) {
2480
+ throw new BrowserError('browser: running page scripts is switched off in settings (Browser → actions)', 'BROWSER_EXECUTE_DISABLED')
2481
+ }
2482
+ if (action === 'download' && !settings.actions.allowDownload) {
2483
+ throw new BrowserError('browser: downloads are switched off in settings (Browser → actions)', 'BROWSER_DOWNLOAD_DISABLED')
2484
+ }
2485
+ if (action === 'credentialWrite' && !settings.actions.allowCredentialWrite) {
2486
+ throw new BrowserError('browser: writing cookies is switched off in settings (Browser → actions)', 'BROWSER_AUTH_WRITE_DISABLED')
2487
+ }
2488
+ }
2489
+
2228
2490
  /**
2229
2491
  * Admit a caller-supplied save path for a file the browser writes to disk.
2230
2492
  * ONE gate for both `browser_download` and `browser_screenshot`: the path
@@ -2245,19 +2507,55 @@ export class ElectronBrowserProvider implements BrowserProvider {
2245
2507
  const file = resolve(savePath)
2246
2508
  if (this.downloadDir !== undefined) {
2247
2509
  const dir = resolve(this.downloadDir)
2248
- // Case-insensitive comparison: Windows paths are case-insensitive, and
2249
- // resolve() does not normalize case. Without this, C:\Users\X\Downloads
2250
- // and c:\users\x\downloads would be treated as different roots.
2251
- const dirLower = dir.toLowerCase()
2252
- const fileLower = file.toLowerCase()
2253
- if (fileLower !== dirLower && !fileLower.startsWith(dirLower + sep.toLowerCase())) {
2510
+ // Windows paths are case-insensitive and resolve() does not normalize case, so there
2511
+ // the containment check folds case — otherwise C:\Users\X\Downloads and
2512
+ // c:\users\x\downloads would read as different roots. Everywhere else the fold is
2513
+ // REMOVED: a case-sensitive filesystem (Linux) treats /home/u/DOWNLOADS and
2514
+ // /home/u/Downloads as different directories, so folding case there let a savePath
2515
+ // that merely differed in case pass admission — a directory-confinement escape, and
2516
+ // one that mkdirSync below would then happily create.
2517
+ const fold = process.platform === 'win32' ? (value: string): string => value.toLowerCase() : (value: string): string => value
2518
+ const dirKey = fold(dir)
2519
+ const fileKey = fold(file)
2520
+ // The directory ITSELF is refused, not accepted: a save path that IS downloadDir names
2521
+ // a directory, so there is no file to write — and mkdirSync would have created the
2522
+ // directory when it did not exist. Two paths are "inside" only when the file path is
2523
+ // strictly below the directory, which the separator check below already expresses.
2524
+ if (!fileKey.startsWith(dirKey + fold(sep))) {
2254
2525
  throw new BrowserError(`browser: ${kind} savePath must be inside downloadDir "${dir}"`, code)
2255
2526
  }
2527
+ // The check above is textual, and a symlink defeats it: resolve() normalizes `..`
2528
+ // but never follows a link, so `<downloadDir>/out/x.png` passed every test above
2529
+ // while `out` pointed at a directory outside the gate — and the write then landed
2530
+ // there. Resolve the REAL location of the deepest existing ancestor (which is what
2531
+ // decides where the bytes go) and require it to be inside the real directory too.
2532
+ // A legitimate path is unaffected: its ancestor resolves to the directory itself.
2533
+ //
2534
+ // The directory is created first so that it can be resolved at all — mkdirSync is a
2535
+ // no-op for one that already exists — and a directory that cannot be created or
2536
+ // resolved falls back to its textual form, leaving the real filesystem error to the
2537
+ // write instead of inventing one here.
2538
+ let realDir: string
2539
+ try {
2540
+ mkdirSync(dir, { recursive: true })
2541
+ realDir = fold(realpathSync(dir))
2542
+ } catch {
2543
+ realDir = fold(dir)
2544
+ }
2545
+ const realAncestor = fold(realAncestorOf(file))
2546
+ if (realAncestor !== realDir && !realAncestor.startsWith(realDir + fold(sep))) {
2547
+ throw new BrowserError(`browser: ${kind} savePath resolves outside downloadDir "${dir}" (a symlink leaves it)`, code)
2548
+ }
2256
2549
  }
2257
2550
  // Never replace an existing file: its previous content is unrecoverable,
2258
2551
  // and the admitted directory may hold files the human put there. A new
2259
2552
  // name is one tool call away.
2260
- if (existsSync(file)) {
2553
+ //
2554
+ // Judged with lstat (see {@link entryExists}), not existsSync: existsSync FOLLOWS
2555
+ // links, so a DANGLING symlink read as "nothing here" and the write below created
2556
+ // its target — outside the directory whenever the link pointed out of it. It is the
2557
+ // directory entry that has to be free, whatever that entry points at.
2558
+ if (entryExists(file)) {
2261
2559
  throw new BrowserError(`browser: refusing to overwrite existing file "${file}" — use another name`, code)
2262
2560
  }
2263
2561
  return file
@@ -2276,6 +2574,9 @@ export class ElectronBrowserProvider implements BrowserProvider {
2276
2574
  * @returns the path the file was written to.
2277
2575
  */
2278
2576
  async download(session: BrowserSessionId, request: { readonly url: string; readonly savePath: string }, signal?: AbortSignal): Promise<{ readonly path: string }> {
2577
+ // Before admission, and before the session is even resolved: a switched-off action is
2578
+ // not a path problem, and the caller should hear the real reason.
2579
+ this.assertActionAllowed('download')
2279
2580
  const s = this.session(session)
2280
2581
  const { handle } = this.activeTab(s)
2281
2582
  signal?.throwIfAborted()
@@ -2340,8 +2641,29 @@ export class ElectronBrowserProvider implements BrowserProvider {
2340
2641
  // Every other carrier drives a real browser over CDP, and cookies are part of CDP —
2341
2642
  // the sidebar goes through the shell's webContents.debugger, an installed Chrome or
2342
2643
  // Edge is the browser itself. So this works everywhere rather than only self-hosted.
2644
+ // Scoped to the page this session is on. The jar itself is SHARED on purpose — every
2645
+ // window and every task drives the same browser, so signing in once works everywhere,
2646
+ // which is the whole point of "use your own browser". What must not be shared is the
2647
+ // HANDING OVER: an unfiltered read puts every domain every task has visited, plus
2648
+ // whatever the human is signed into, in front of one task's caller. Least exposure,
2649
+ // not isolation.
2650
+ //
2651
+ // Two corrections to the previous attempt. It called Storage.getCookies with a `urls`
2652
+ // parameter, but that method takes only browserContextId and ignores everything else, so
2653
+ // the filter did nothing and the whole jar was still exported; `urls` belongs to
2654
+ // Network.getCookies. And it silently fell back to an unfiltered read when the page URL
2655
+ // was unavailable — an error path that asked for MORE than the normal one. Both are
2656
+ // fixed: the method that honours the filter, and a failure that refuses instead of
2657
+ // widening.
2658
+ const cookiesUrl = await this.currentUrl(handle).catch(() => '')
2659
+ if (cookiesUrl === '') {
2660
+ throw new BrowserError(
2661
+ 'browser: cannot export cookies without knowing which page this session is on — the URL was unavailable, and reading the whole cookie jar would expose every other session\'s logins',
2662
+ 'BROWSER_AUTH_SCOPE_UNKNOWN',
2663
+ )
2664
+ }
2343
2665
  const result = await withTimeout(
2344
- handle.sendCommand('Storage.getCookies', {}),
2666
+ handle.sendCommand('Network.getCookies', { urls: [cookiesUrl] }),
2345
2667
  timeoutMs,
2346
2668
  undefined,
2347
2669
  `browser: auth export timed out after ${timeoutMs}ms`,
@@ -2353,12 +2675,12 @@ export class ElectronBrowserProvider implements BrowserProvider {
2353
2675
 
2354
2676
  /** Import cookies into the session (restore login state). Self-hosted only. */
2355
2677
  async restoreAuth(session: BrowserSessionId, cookies: readonly ExportedCookie[]): Promise<number> {
2356
- // The settings panel offers "allow the agent to read cookies / export login state".
2357
- // Until now nothing read that switch, so turning it off changed nothing at all —
2358
- // the tool exported every cookie regardless.
2359
- if (this.settingsSource?.().credentials.allowRead === false) {
2360
- throw new BrowserError('browser: reading cookies is switched off in settings (Browser → credentials)', 'BROWSER_AUTH_DISABLED')
2361
- }
2678
+ // Restore WRITES cookies, so it answers to the write switch — not to
2679
+ // `credentials.allowRead`. It used to consult the read switch and report "reading
2680
+ // cookies is switched off", which is both the wrong gate and the wrong sentence: an
2681
+ // operator who allowed reading but wanted logins left alone had no switch at all,
2682
+ // and one who turned reading off lost the restore path for the wrong reason.
2683
+ this.assertActionAllowed('credentialWrite')
2362
2684
  const s = this.session(session)
2363
2685
  const { handle } = this.activeTab(s)
2364
2686
  const timeoutMs = 30_000
@@ -2371,14 +2693,28 @@ export class ElectronBrowserProvider implements BrowserProvider {
2371
2693
  // CDP again, for the same reason as the export: any carrier that drives a real
2372
2694
  // browser can set cookies on it. Storage.setCookies needs the cookie fields spelled
2373
2695
  // the CDP way, including the sameSite spelling and seconds rather than milliseconds.
2696
+ //
2697
+ // toCdpCookie returns undefined for a cookie it cannot place (no usable domain), and
2698
+ // `cookies.map(...)` passed that undefined straight through: the array then held a
2699
+ // JSON `null`, CDP rejected the WHOLE call, and every valid cookie failed with an error
2700
+ // naming none of them. Filter the unusable ones out and say how many went.
2701
+ const mapped = cookies.map(toCdpCookie)
2702
+ const convertible = mapped.filter((cookie): cookie is Record<string, unknown> => cookie !== undefined)
2703
+ const dropped = mapped.length - convertible.length
2704
+ if (dropped > 0) {
2705
+ process.stderr.write(`[dsh-browser] restoreAuth: dropped ${dropped} of ${mapped.length} cookie(s) with no usable domain; the rest were still restored\n`)
2706
+ }
2374
2707
  const result = await withTimeout(
2375
- handle.sendCommand('Storage.setCookies', { cookies: cookies.map(toCdpCookie) }),
2708
+ handle.sendCommand('Storage.setCookies', { cookies: convertible }),
2376
2709
  timeoutMs,
2377
2710
  undefined,
2378
2711
  `browser: auth restore timed out after ${timeoutMs}ms`,
2379
2712
  )
2380
- const restored = toExportedCookies(result.cookies).length
2381
- this.record(s, 'restoreAuth', { count: cookies.length }, true, { result: `${restored} cookies (CDP)` })
2713
+ // Storage.setCookies returns nothing, so the count comes from what was actually
2714
+ // sent. Reading result.cookies here was always undefined, which rendered as
2715
+ // "Restored 0 cookies." while the write had succeeded.
2716
+ const restored = convertible.length
2717
+ this.record(s, 'restoreAuth', { count: cookies.length, dropped }, true, { result: `${restored} cookies (CDP)${dropped > 0 ? `; ${dropped} dropped` : ''}` })
2382
2718
  return restored
2383
2719
  }
2384
2720
 
@@ -2442,6 +2778,7 @@ export class ElectronBrowserProvider implements BrowserProvider {
2442
2778
  // to the region it names, so it has to cover the whole document. The page's own size
2443
2779
  // is read first; when that is unavailable the capture proceeds unscaled rather than
2444
2780
  // failing — an image the caller can shrink itself beats no image at all.
2781
+ let rendered: { width: number; height: number } | undefined
2445
2782
  if (request?.maxWidth !== undefined || request?.maxHeight !== undefined) {
2446
2783
  const metrics = await withTimeout(
2447
2784
  handle.sendCommand('Page.getLayoutMetrics', {}),
@@ -2455,6 +2792,9 @@ export class ElectronBrowserProvider implements BrowserProvider {
2455
2792
  const byWidth = request.maxWidth !== undefined && size.width > 0 ? request.maxWidth / size.width : 1
2456
2793
  const byHeight = request.maxHeight !== undefined && height > 0 ? request.maxHeight / height : 1
2457
2794
  const scale = Math.min(1, byWidth, byHeight)
2795
+ // What the caller will actually receive. Recorded before the branch below, because the
2796
+ // size is worth reporting even when no scaling happened (scale 1).
2797
+ rendered = { width: Math.round(size.width * scale), height: Math.round(height * scale) }
2458
2798
  if (scale < 1) {
2459
2799
  params.clip = { x: 0, y: 0, width: size.width, height, scale }
2460
2800
  // The clip defines the captured area; captureBeyondViewport must not also apply.
@@ -2472,7 +2812,12 @@ export class ElectronBrowserProvider implements BrowserProvider {
2472
2812
  if (typeof data !== 'string') {
2473
2813
  throw new BrowserError('browser: screenshot returned no image data', 'BROWSER_SCREENSHOT_FAILED')
2474
2814
  }
2475
- return this.saveScreenshot(data, request?.savePath, 'image/png')
2815
+ // The mime must name the format that was ACTUALLY requested (params.format), not PNG
2816
+ // unconditionally: a carrier that sets supportsCdpJpeg asks for `format: 'jpeg'`, so
2817
+ // hardcoding the PNG mime announced a JPEG body as a PNG — the data URL then carried a
2818
+ // mislabelled image, and a caller that decodes by mime wrote the wrong file.
2819
+ const mime = params.format === 'jpeg' ? 'image/jpeg' : 'image/png'
2820
+ return this.saveScreenshot(data, request?.savePath, mime, rendered)
2476
2821
  }
2477
2822
 
2478
2823
  /**
@@ -2482,7 +2827,12 @@ export class ElectronBrowserProvider implements BrowserProvider {
2482
2827
  * file — so a screenshot cannot be used to write to, or silently replace,
2483
2828
  * files anywhere the DSH process happens to have permission (issue #13).
2484
2829
  */
2485
- private saveScreenshot(base64: string, savePath: string | undefined, mime: string): { dataUrl: string; path?: string } {
2830
+ private saveScreenshot(
2831
+ base64: string,
2832
+ savePath: string | undefined,
2833
+ mime: string,
2834
+ rendered?: { width: number; height: number },
2835
+ ): { dataUrl: string; path?: string; width?: number; height?: number } {
2486
2836
  if (savePath !== undefined) {
2487
2837
  // Admission failures propagate unchanged: they describe the caller's
2488
2838
  // path, not a disk problem.
@@ -2490,13 +2840,13 @@ export class ElectronBrowserProvider implements BrowserProvider {
2490
2840
  try {
2491
2841
  mkdirSync(dirname(target), { recursive: true })
2492
2842
  writeFileSync(target, Buffer.from(base64, 'base64'))
2493
- return { dataUrl: `data:${mime};base64,${base64}`, path: target }
2843
+ return { dataUrl: `data:${mime};base64,${base64}`, path: target, ...rendered }
2494
2844
  } catch (error) {
2495
2845
  // Report the write problem but keep the capture usable.
2496
2846
  throw new BrowserError(`browser: screenshot save to "${target}" failed: ${String(error)}`, 'BROWSER_SCREENSHOT_SAVE_FAILED', { cause: error })
2497
2847
  }
2498
2848
  }
2499
- return { dataUrl: `data:${mime};base64,${base64}` }
2849
+ return { dataUrl: `data:${mime};base64,${base64}`, ...rendered }
2500
2850
  }
2501
2851
 
2502
2852
  /** Append one operation to the session's history. */