chromeflow 0.7.1 → 0.9.8

package/package.json

@@ -1,36 +1,50 @@
 {
   "name": "chromeflow",
-  "version": "0.7.1",
-  "description": "Browser guidance MCP server for Claude Code — highlights, clicks, fills, and captures from the web so you don't have to.",
+  "version": "0.9.8",
+  "description": "MCP server for chromeflow — lets Claude Code or Codex CLI drive your real Chrome browser with sessions intact. Plugin install recommended; npx chromeflow for manual MCP wiring.",
   "type": "module",
+  "main": "./bin/chromeflow.mjs",
   "bin": {
-    "chromeflow": "dist/index.js"
+    "chromeflow": "./bin/chromeflow.mjs"
   },
   "files": [
-    "dist",
-    "CLAUDE.md",
-    "README.md"
+    "bin/",
+    "README.md",
+    "LICENSE"
   ],
-  "repository": {
-    "type": "git",
-    "url": "git+https://github.com/NeoDrew/chromeflow.git"
+  "engines": {
+    "node": ">=20"
+  },
+  "scripts": {
+    "build": "../plugin/scripts/build-server.sh && mkdir -p bin && cp ../plugin/server/chromeflow.mjs bin/chromeflow.mjs",
+    "dev": "tsc --watch"
   },
-  "homepage": "https://github.com/NeoDrew/chromeflow",
-  "license": "MIT",
   "keywords": [
     "mcp",
-    "claude",
+    "model-context-protocol",
     "claude-code",
-    "browser",
-    "automation",
-    "chrome"
+    "codex-cli",
+    "browser-automation",
+    "chrome-extension",
+    "ai-agent",
+    "agent-tooling",
+    "playwright-alternative",
+    "browser-use-alternative",
+    "puppeteer-alternative",
+    "web-automation",
+    "stripe",
+    "canvas",
+    "real-chrome-session"
   ],
-  "scripts": {
-    "prebuild": "cp ../../CLAUDE.md ./CLAUDE.md && cp ../../README.md ./README.md",
-    "build": "tsc",
-    "postbuild": "node --input-type=commonjs -e \"const fs=require('fs');const f='dist/index.js';fs.writeFileSync(f,'#!/usr/bin/env node\\n'+fs.readFileSync(f,'utf8'));\"",
-    "dev": "tsc --watch"
+  "repository": {
+    "type": "git",
+    "url": "git+https://gitlab.com/NeoDrew/chromeflow.git",
+    "directory": "packages/mcp-server"
   },
+  "homepage": "https://chromeflow.run",
+  "bugs": "https://gitlab.com/NeoDrew/chromeflow/-/issues",
+  "license": "MIT",
+  "author": "Andrew Robertson (NeoDrewX)",
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.0.0",
     "ws": "^8.18.0",
@@ -39,6 +53,7 @@
   "devDependencies": {
     "@types/node": "^22.0.0",
     "@types/ws": "^8.5.0",
+    "esbuild": "^0.24.0",
     "typescript": "^5.5.0"
   }
 }

package/CLAUDE.md

@@ -1,356 +0,0 @@
-# Chromeflow — Claude Instructions
-
-## What chromeflow is
-Chromeflow is a browser guidance tool. When a task requires the user to interact with a
-website (create accounts, set up billing, retrieve API keys, configure third-party services),
-use chromeflow to guide them through it visually instead of giving text instructions.
-
-## When to use chromeflow (be proactive)
-Use chromeflow automatically whenever a task requires:
-- Creating or configuring a third-party account (Stripe, SendGrid, Supabase, Vercel, etc.)
-- Retrieving API keys, secrets, or credentials to place in `.env`
-- Setting up pricing tiers, webhooks, or service configuration in a web UI
-- Any browser-based step that is blocking code work
-
-Do NOT ask "should I open the browser?" — just do it. The user expects seamless handoff.
-
-**Never end a response with a "you still need to" list of browser tasks.** If code changes are done and browser steps remain (e.g. creating a Stripe product, adding an env var), continue immediately with chromeflow — don't hand them back to the user.
-
-## HARD RULES — never break these
-
-1. **Never use Bash as a fallback for browser tasks.** If `click_element` fails, use
-   `scroll_page` then retry, or use `highlight_region` to show the user. Never use
-   `osascript`, `applescript`, or any shell command to control the browser.
-
-2. **Never use `take_screenshot` to read page content.** After `scroll_page`, after
-   `click_element`, after navigation — always call `get_page_text`, not `take_screenshot`.
-   `get_page_text` returns up to 10,000 characters; if truncated it tells you the next
-   `startIndex` to paginate. When you only need to confirm a specific phrase is present,
-   prefer `find_text("phrase")` — it returns matches with context and selectors instead of
-   dumping the whole page. Screenshots are only for locating an element's pixel position
-   when DOM queries have already failed. Never take more than 1–2 screenshots in a row.
-
-3. **Use `wait_for_selector` to wait for async page changes** (build completion, modals,
-   toasts). Never poll with repeated `take_screenshot` calls.
-
-## Guided flow pattern
-
-```
-1. open_page(url)                            — navigate to the right page (add new_tab=true to keep current tab open; add background=true to keep the current tab focused if its form auto-saves on blur)
-2. For each step:
-   a. Claude acts directly:
-        click_element("Save")               — press buttons/links Claude can press
-        click_element("Save", until_selector=".success-toast")  — when synthetic clicks may silently no-op on a React-heavy site, require an observable post-click condition (or until_url_contains / until_text_contains)
-        get_page_text() or wait_for_selector(".success") — confirm after click without an until-clause; click_element returns after 600ms regardless of outcome unless until_* was used
-        fill_form([{label, value}, ...], exact=true)  — fill multiple fields in one call; pass exact=true on dense forms to refuse fuzzy text-walk matches
-        fill_input("Product name", "Pro")   — fill a single field (works on React, CodeMirror, and contenteditable). Always check the response — it names the matched element so you can spot wrong-field matches
-        fill_input("Rate", "5", exact=true) — exact-match mode for short generic labels that may collide with neighbouring fields
-        react_set_input("input[name=email]", "x@y") — for inputs where fill_input fails (or for iframe-hosted inputs via frame=...) — handles the prototype-from-instance gotcha automatically
-        type_text("hello world")            — type via trusted keyboard events (use when fill_input fails isTrusted checks)
-        type_text("description", frame="iframe.se-rte")  — type into a same-origin iframe's contenteditable (eBay description editor pattern)
-        set_file_input("Upload", "/abs/path/to/file.zip") — upload a file; returns success only after the upload is observably committed (no manual sleep needed between rapid uploads)
-        clear_overlays()                    — call this immediately after fill_input/fill_form succeeds
-        scroll_to_element("label text")     — jump directly to a known field; prefer this over scroll_page when the target is known
-        scroll_page("down")                 — reveal off-screen content when target location is unknown
-   b. Check results with text, not vision:
-        get_page_text()                     — read errors/status after actions
-        wait_for_selector(".success")       — wait for a new element to appear
-        wait_for_change(".toast")          — wait for an existing element's content to mutate, then read it (uses MutationObserver, cheaper than polling)
-        execute_script("return await fetch('/api/x').then(r => r.json())")  — top-level await is supported, no window.__variable + sleep dance needed
-   c. When an element can't be found or clicked:
-        scroll_page("down") and retry      — always try this first
-        get_elements()                      — get EXACT DOM coords when needed
-        highlight_region(selector,msg)      — highlight by CSS selector (preferred; scrolls element into view automatically)
-        highlight_region(x,y,w,h,msg)       — highlight by coords only if no selector available (coords go stale on scroll)
-        [absolute last resort] take_screenshot() — only if you genuinely can't identify the element from DOM
-   d. Pause for the user when needed:
-        find_and_highlight(text, msg)        — show the user what to do
-        wait_for_click()                    — wait for user interaction
-        [after fill_input] clear_overlays() — always clear after filling
-3. clear_overlays()                          — clean up when done
-```
-
-**Default to automation.** Only pause for human input when the step genuinely requires
-personal data or a human decision.
-
-## What to do automatically vs pause for the user
-
-**Claude acts directly** (`click_element` / `fill_input`):
-- Any button: Save, Continue, Create, Add, Confirm, Next, Submit, Update
-- Product names, descriptions, feature lists
-- Prices and amounts specified in the task
-- URLs, redirect URIs, webhook endpoints
-- Selecting billing period, currency, or other known options
-- Dismissing cookie banners, cookie dialogs, "not now" prompts
-
-**Pause for the user** (`find_and_highlight` + `wait_for_click`):
-- Email address / username / login
-- Password or passphrase
-- Payment method / billing / card details
-- Phone number / 2FA / OTP codes
-- Any legal consent the user must personally accept
-- Choices that depend on user preference Claude wasn't told
-
-## Capturing credentials
-After a secret key or API key is revealed:
-1. `read_element(hint)` — capture the value
-2. `write_to_env(KEY_NAME, value, envPath)` — write to `.env`
-3. Tell the user what was written
-
-Use the absolute path for `envPath` — it's the Claude Code working directory + `/.env`.
-
-To capture and share a screenshot (e.g. for uploading to a form or pasting into a chat),
-use `take_screenshot(copy_to_clipboard=true, save_to="downloads")` — saves a PNG to ~/Downloads
-and copies it to the clipboard. The defaults (`copy_to_clipboard=false, save_to="none"`) return
-the image to Claude only.
-
-## Working with complex forms
-- Before filling a large or unfamiliar form, call `get_form_fields()` to get a full inventory
-  of every field (type, label, current value, vertical position, and section heading). Use
-  `get_elements()` when you need pixel coordinates of visible elements; use `get_form_fields()`
-  when you need to understand the full structure of a form including fields below the fold.
-  If you only need one or two specific fields, use `find_input("hint")` instead — targeted
-  lookup is much cheaper than the full inventory and returns labels you can pipe straight
-  into `fill_input`.
-- `get_form_fields()` includes `[type=file]` fields even when they are visually hidden behind
-  custom drag-and-drop zones. Use `set_file_input(hint, filePath)` to upload a file — provide
-  the label/hint text and the absolute path to the file on disk.
-- For forms with multiple fields, use `fill_form([{label, value}, ...])` to fill them all
-  in a single call. It returns a per-field success/failure report so you can immediately see
-  which fields weren't found. Use `fill_input` only for a single field.
-- `fill_input` and `fill_form` work on React-controlled inputs, contenteditable (Stripe,
-  Notion), and **CodeMirror 6 editors** — auto-detected. After filling, the value is read
-  back and a warning is shown if React did not accept it.
-- **Monaco editors** (VS Code-style code editors on DataAnnotation, etc.) appear in
-  `get_form_fields()` as type "monaco". They cannot be filled via `fill_input` — use
-  `execute_script` with the Monaco API instead:
-  ```js
-  // Read content from the first Monaco model
-  monaco.editor.getModels()[0].getValue()
-  // Write content to the first Monaco model
-  monaco.editor.getModels()[0].setValue('new content here')
-  ```
-- `set_file_input` accepts CSS selectors as the hint (e.g. `#import-problem-file`,
-  `.upload-input`) in addition to label text. Use selectors when file inputs are hidden
-  behind custom UIs and have no visible label.
-- **Replacing an already-uploaded file**: after `set_file_input` succeeds, the input
-  becomes invisible and a "Remove" span/button typically appears near the upload area.
-  To replace the file: `click_element("Remove", nth=N)` (the right `nth` if there are
-  multiple), then call `set_file_input(hint, newPath)` again — the same hidden input is
-  recycled and accepts the new file. Verify with `get_form_fields()` between the two
-  steps so you're sure the input has reappeared.
-- **Forcing auto-save on idempotent text edits** (e.g. keep-alive loop on an
-  auto-saving DataAnnotation form): some auto-save logic diffs against the last-saved
-  value and skips no-op writes. To force a real save on each tick without changing
-  visible content, toggle a trailing space — add when absent, remove when present.
-  `fill_input` value comparison handles both directions transparently.
-- After any radio/checkbox click that reveals new fields, call `get_form_fields()` again —
-  the inventory will include the new fields and warn if more hidden ones still exist.
-- If a form has collapsible sections, expand them all before calling `get_form_fields()` so
-  the field list is complete. Use the `[under: "section name"]` context in each field's entry
-  to identify fields by section rather than by index — indices shift when sections expand.
-- Prefer `scroll_to_element("label text or #selector")` over `scroll_page` whenever you know
-  which field or section you need — it scrolls precisely and confirms the matched element.
-- For multi-session tasks (long forms that may exceed context), call `save_page_state()` as a
-  checkpoint. A future session can call `restore_page_state()` to reload all field values.
-
-## Discovery — find without dumping the whole page
-
-Three lightweight tools save tokens vs `get_page_text` / `get_form_fields` when you don't need the full content:
-
-- `find_text("Saved successfully")` — grep the DOM. Returns surrounding context, a CSS selector, and a `clickable` flag for each match. Use this instead of `get_page_text` when you're checking whether a specific phrase is present, or to locate a button by its visible text. If `clickable=true`, pipe the matched text straight into `click_element`.
-- `find_input("Email")` — fuzzy form-field lookup, top-N. Returns labels you can pipe straight into `fill_input(label, value)` — both tools share the same match ranks (`aria-eq` → `placeholder-eq` → `label-text-eq` → `name-eq` → `id-eq` → `*-includes` → `fuzzy-text-walk`). Cheaper than `get_form_fields` when you just need a couple of specific fields. Pass `type_filter="email"` to restrict to a specific input type.
-- `wait_for_text("Saved")` — wait for text to appear without knowing the selector ahead of time. Complements `wait_for_selector` for the case where you only know the post-action message.
-
-All three pierce open shadow roots and accept `frame="iframe.selector"` for same-origin iframes. Pass `regex=true` on `find_text` / `wait_for_text` for case-insensitive regex matching. Pass `exact=true` on `find_input` to refuse fuzzy text-walk matches.
-
-```
-find_text("Build complete", scope_selector=".log-output")     — only check the build log section
-find_input("Card number", type_filter="text")                  — find Stripe's card-number field
-wait_for_text("Deploy successful", timeout_ms=30000)           — wait up to 30s after clicking Deploy
-```
-
-Reach for these BEFORE `get_page_text` / `get_form_fields` when the goal is "is X here?" or "where is X?". Reserve `get_page_text` for reading actual content, and `get_form_fields` for understanding a whole form's structure.
-
-## Working with multiple tabs
-- Before opening a new tab, call `list_tabs()` to check if the target URL is already open —
-  use `switch_to_tab` to return to it instead of opening a duplicate.
-- `open_page(url, new_tab=true)` opens a URL without losing the current tab. Use sparingly —
-  prefer switching to an existing tab over opening a new one.
-- `switch_to_tab("1")` switches by tab number; `switch_to_tab("form")` matches by URL or title substring.
-- Before navigating away from a partially-filled form, call `save_page_state()` so the form
-  can be restored if the tab reloads or the page loses its state on return.
-- **In long-lived self-rescheduling loops**, the active tab can silently drift mid-session
-  (the user navigates manually while AFK, or another tab steals focus). At the start of
-  every loop iteration, call `list_tabs` and verify the active tab's URL matches your
-  expected target — if not, `switch_to_tab(<URL or title substring>)` before running
-  `execute_script` or any other tab-scoped tool. Without this guard, scripts run on the
-  wrong tab and fail with confusing "undefined" errors that look like page bugs.
-
-## Error handling
-
-**After any action**, confirm with `get_page_text()` or `wait_for_selector` — never take a
-screenshot to check what happened.
-
-**`click_element` not found:**
-1. `scroll_page("down")` then retry `click_element`
-2. `get_elements()` to get exact coords → `highlight_region(x,y,w,h,msg)`
-3. `take_screenshot()` only if you still can't identify the element from DOM queries
-
-**Multiple elements with the same label** (e.g. many "Remove" buttons):
-`click_element("Remove", nth=3)` — use `nth` (1-based) to target the specific one by order top-to-bottom. Check `get_form_fields` or `get_page_text` first to determine which index corresponds to the right section.
-
-**`fill_input` matched the wrong field** (always read the response — it names the matched element):
-- If you wanted "Ad rate" and got back `<input name="title">`, the fuzzy text walker latched onto a neighbour. Retry with `exact=true` and a more specific hint, or use `react_set_input(selector, value)` with a precise CSS selector.
-- The match-strength is reported as `aria-eq`, `placeholder-eq`, `name-eq`, `id-eq`, `label-text-eq`, or fuzzier kinds. Anything labeled `fuzzy-text-walk` or `*-includes` is the lowest-confidence kind — verify the matched element really was what you wanted.
-
-**`fill_input` not found or rejected by the page:**
-1. `click_element(hint)` to focus the field, then retry `fill_input`
-2. `react_set_input("input[name=...]", value)` — uses the input's own prototype to set the value, dispatches input/change. Handles the "Illegal invocation" iframe gotcha and the prototype-from-instance ceremony for you.
-3. If the site rejects programmatic input (isTrusted check, shadow DOM, custom editors):
-   - `click_element(hint)` to focus the field
-   - `execute_script("document.execCommand('selectAll')")` to clear existing content
-   - `type_text("new value")` — uses CDP trusted keyboard events that pass isTrusted checks
-4. For iframe-hosted contenteditable rich-text editors (eBay's description, etc.):
-   - `type_text("body content", frame="iframe.selector")` — same-origin only. Focuses the iframe's contenteditable, types via CDP, dispatches input/change in the iframe's context so React reads the new value.
-5. `find_and_highlight(hint, "Click here — I'll fill it in")` (no `valueToType`) then
-   `wait_for_click()` — the user's click focuses the field and `fill_input`'s active-element
-   fallback fills it automatically
-6. Call `clear_overlays()` after `fill_input` succeeds
-7. Only use `valueToType` when the user must personally type the value (password, personal data)
-
-**`click_element` returned success but the page didn't change** (common on React-heavy sites where synthetic clicks no-op):
-Pass an `until_*` clause to require an observable post-click condition. `click_element` returns success=false if the condition isn't met within `until_timeout_ms` (default 5000):
-```
-click_element("List with displayed fees", until_url_contains="/listing-published")
-click_element("Save", until_selector=".success-toast")
-click_element("Confirm", until_text_contains="Order placed")
-```
-If success=false: try `react_set_input` to fire the click via the page's own React handler, or use `execute_script("document.querySelector(...).click()")` directly.
-
-**`set_file_input` not committing on rapid back-to-back uploads:**
-The default 3000ms commit-wait is enough for most uploaders. For batch photo uploads on slow react file handlers (eBay's 25-photo carousel, Stripe Connect document upload), increase `wait_ms` to 6000–8000 OR pass `verify_selector` pointing at the thumbnail/Remove-button that should appear:
-```
-set_file_input("Photos", "/path/1.jpg", verify_selector=".photo-thumbnail:nth-of-type(1)")
-set_file_input("Photos", "/path/2.jpg", verify_selector=".photo-thumbnail:nth-of-type(2)")
-```
-The page-level file count is reported in the response — use it to spot uploaders that consume-and-reset the input vs uploaders that keep the file there.
-
-**Waiting for async results** (build, save, deploy): `wait_for_selector(selector, timeout)` — never poll with screenshots. `wait_for_selector` pierces open shadow roots, so a selector inside a web component (Outlier task UI, Lit/Stencil widget) matches without ceremony.
-
-**Waiting for a shadow host's tree to attach** (e.g. SPA route flips where `<my-host>` appears 10s before its shadow content hydrates, and `wait_for_selector("my-host")` resolves while `host.shadowRoot` is still null): pass `shadow_root=true`. The wait then requires the matched element's `.shadowRoot` to be non-null, not just for the host element to exist.
-```
-wait_for_selector("iframe", shadow_root=true)   — wait until the iframe both exists AND has an attached shadowRoot
-```
-
-**Waiting for an existing region to update** (e.g. click Save, then get the confirmation toast; send a chat message, then get the reply): `wait_for_change(selector)` uses a MutationObserver on the element's subtree and returns its new text content as soon as the mutation settles. Prefer this over `wait_for_selector` + `get_page_text` when the element already exists and you just need its next state — one call instead of two, no polling.
-
-**Pre-filling `prompt()` and `confirm()` dialogs**: When a page action will trigger a JS
-dialog (e.g. "Save As" calling `prompt()`), call `set_dialog_response` BEFORE the action:
-```
-set_dialog_response(type="prompt", value="my-filename")   — next prompt() returns "my-filename"
-set_dialog_response(type="confirm", value="true")          — next confirm() returns true
-```
-Then trigger the action (e.g. `click_element("Save As")`). The response is consumed once.
-
-**React Select / custom styled dropdowns** (e.g. "Select..." components on DataAnnotation):
-`click_element` and `fill_input` do NOT work on these — they intercept native events. The cleanest path is `react_set_input` (which handles the prototype-from-instance setter for you) followed by a click on the filtered option:
-
-```
-1. react_set_input('input[id*="react-select-3-input"]', "Target Option")
-   — sets the hidden combobox input via its own prototype's value-setter and dispatches the input event React's onChange listens for
-2. (300ms pause for the dropdown to filter)
-3. execute_script("document.querySelector('[id*=\"react-select-3-option-0\"]').click()")
-4. Verify the control shows the selected value:
-   execute_script("document.querySelector('[class*=\"singleValue\"]').textContent.trim()")
-```
-
-If you must hand-roll this with `execute_script` (older React-Select versions, weird custom wrappers), prefer reading the prototype FROM the instance to avoid "Illegal invocation" inside iframes:
-
-```js
-var input = document.querySelector('input[id*="react-select-3-input"]');
-input.focus();
-var setter = Object.getOwnPropertyDescriptor(Object.getPrototypeOf(input), 'value').set;
-setter.call(input, 'Target Option');
-input.dispatchEvent(new Event('input', { bubbles: true }));
-```
-
-Fallback if the combobox approach doesn't work (older React Select versions):
-```js
-var controls = document.querySelectorAll('[class*="control"]');
-controls[N].click();
-var allEls = document.querySelectorAll('*');
-for (var i = 0; i < allEls.length; i++) {
-  if (allEls[i].textContent.trim() === 'Target Option' && allEls[i].children.length === 0) {
-    allEls[i].dispatchEvent(new MouseEvent('mousedown', {bubbles: true}));
-    allEls[i].click();
-    break;
-  }
-}
-```
-
-**Page text with large embedded content** (e.g. uploaded log files previewed inline): full-page `get_page_text()` pagination becomes unwieldy. Scope to a specific section instead:
-```
-get_page_text(selector=".section-3")   — scope to a CSS selector
-get_page_text(selector="#upload-form") — scope to an id
-```
-Use `execute_script("document.querySelectorAll('section').length")` to find structural selectors first.
-
-**Page content rendered as images** (e.g. qualification "Examples" tabs that show PNG screenshots
-instead of DOM text): `get_page_text()` returns nothing useful. Zoom out and screenshot instead:
-
-```js
-// Shrink to fit wide content, then screenshot
-document.body.style.zoom = '0.4';
-// use take_screenshot() to read it
-// restore afterward:
-document.body.style.zoom = '1';
-```
-
-**Downloads via `execute_script`**: Creating a Blob URL and clicking an anchor via
-`execute_script` sometimes fails due to CSP or timing. If a download doesn't trigger:
-1. Retry the exact same `execute_script` call
-2. If still failing, use `find_and_highlight` to show the user a download button to click manually
-
-**React-controlled native radios/checkboxes that don't update `checked`**: `click_element`
-auto-handles this for native `<input type=radio>` and `<input type=checkbox>` inputs
-(including labels that wrap or `for=` reference them). The flow:
-- If a radio is already `checked=true`, `click_element` skips the click — re-clicking can
-  toggle it OFF on React forms whose `onChange` interprets the click as a deselect. The
-  response says `"X — radio already checked, click skipped"`.
-- If the standard click fires but the input's `checked` state didn't change as expected
-  (radio still unchecked, or checkbox didn't toggle), `click_element` automatically
-  dispatches the full pointer-event chain (`pointerdown → mousedown → pointerup → mouseup
-  → click`) on the input. The response says `"now checked (after pointer-chain fallback)"`.
-
-You only need to drop into `execute_script` for the no-native-input case below.
-
-**Shadow DOM `[role=radio]` / role-only custom radios silently no-op**: On sites like
-Outlier where the radio is a `[role=radio]` div with no underlying `<input>`,
-`click_element`'s native-input fallback can't help — the click target has no `.checked`
-property to verify. Two things must be true: (a) the element must be scrolled into view
-FIRST (`scrollIntoView({block:'center'})`), and (b) the full pointer-event chain must
-fire — not just `click()`:
-```js
-['pointerdown','mousedown','pointerup','mouseup','click'].forEach(t =>
-  el.dispatchEvent(new MouseEvent(t, {bubbles: true, cancelable: true}))
-);
-```
-After scroll, re-query the radio list — its length may change as more content becomes
-visible. Then verify `aria-checked === "true"` before moving on.
-
-**Visibility-detection overlays** (e.g. Multimango's "Content Hidden" black overlay):
-Some sites render a full-screen overlay when the tab loses focus, triggered by
-`document.visibilityState` / `document.hidden`. Chromeflow tab-switching triggers it.
-Workaround — remove the overlay and patch the APIs:
-```js
-document.querySelectorAll('[style*="z-index: 99999"]').forEach(el => el.remove());
-Object.defineProperty(document, 'hidden', { get: () => false, configurable: true });
-Object.defineProperty(document, 'visibilityState', { get: () => 'visible', configurable: true });
-['visibilitychange','blur'].forEach(t =>
-  document.addEventListener(t, e => e.stopImmediatePropagation(), true)
-);
-```
-Re-apply after every navigation.
-
-**Never use Bash to work around a stuck browser interaction.**

package/dist/index.js

@@ -1,92 +0,0 @@
-#!/usr/bin/env node
-import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
-import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
-import { WsBridge } from "./ws-bridge.js";
-import { registerBrowserTools } from "./tools/browser.js";
-import { registerHighlightTools } from "./tools/highlight.js";
-import { registerCaptureTools } from "./tools/capture.js";
-import { registerFlowTools } from "./tools/flow.js";
-import { runSetup, runUpdate, runUninstall } from "./setup.js";
-if (process.argv[2] === "setup") {
-  runSetup().catch((err) => {
-    console.error(err);
-    process.exit(1);
-  });
-} else if (process.argv[2] === "update") {
-  runUpdate().catch((err) => {
-    console.error(err);
-    process.exit(1);
-  });
-} else if (process.argv[2] === "uninstall") {
-  runUninstall().catch((err) => {
-    console.error(err);
-    process.exit(1);
-  });
-} else {
-  main().catch((err) => {
-    console.error("[chromeflow] Fatal error:", err);
-    process.exit(1);
-  });
-}
-async function main() {
-  const bridge = new WsBridge();
-  const server = new McpServer({
-    name: "chromeflow",
-    version: "0.1.14"
-  });
-  registerBrowserTools(server, bridge);
-  registerHighlightTools(server, bridge);
-  registerCaptureTools(server, bridge);
-  registerFlowTools(server, bridge);
-  server.prompt(
-    "chromeflow-status",
-    "Check if the chromeflow Chrome extension is connected and which tab is active",
-    async () => {
-      const connected = bridge.isConnected();
-      if (!connected) {
-        return {
-          messages: [{
-            role: "user",
-            content: {
-              type: "text",
-              text: "Check chromeflow status. The Chrome extension is NOT connected. Tell the user to reload the chromeflow extension in chrome://extensions."
-            }
-          }]
-        };
-      }
-      try {
-        const response = await bridge.request({ type: "list_tabs" }, 3e3);
-        const tabs = response.tabs;
-        const active = tabs.find((t) => t.active);
-        const tabList = tabs.map((t) => `${t.index}. ${t.active ? "[active] " : ""}${t.title} \u2014 ${t.url}`).join("\n");
-        return {
-          messages: [{
-            role: "user",
-            content: {
-              type: "text",
-              text: `Check chromeflow status.
-
-Extension: Connected
-Active tab: ${active?.title ?? "none"} \u2014 ${active?.url ?? ""}
-All tabs:
-${tabList}`
-            }
-          }]
-        };
-      } catch {
-        return {
-          messages: [{
-            role: "user",
-            content: {
-              type: "text",
-              text: "Check chromeflow status. Extension is connected but not responding. The user may need to reload it."
-            }
-          }]
-        };
-      }
-    }
-  );
-  const transport = new StdioServerTransport();
-  await server.connect(transport);
-  console.error("[chromeflow] MCP server running. Waiting for Claude...");
-}