- quick_tools indexed serde_json Values mutably, which inserted
actions.items.properties = null into click/typing/scroll; xAI refused
the whole tool list, so since ba1fce5 every step fell back to the main
model. It now uses get_mut and the test checks nothing but ref changes.
- The front app's menu bar titles and any open menu (submenus, right-click
menus) are listed with numbers and their shortcuts; an option-key
alternate in the same place is noted on the visible item.
- Disclosure triangles, colour wells and date fields get numbers; an
icon-only control without a name goes by its description.
- When the front window can't be read the list says so (with the reason)
instead of 「」 above the menu bar alone.
- Computer facts add each system's conventions (Mac: cmd+shift+G, save
panel, default format, /bin/sh echo -n; Linux: GTK ctrl+L).
- `anybot --help` and unknown -- flags print the usage and exit instead of
opening the window on the real data folder.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uirha5F5gJt93t3yEx3XcS
Measured on the bot's own box (8 tasks x 2, grok-4.7): 14/16 done, 120 steps, 1114 s before;
16/16, 102 steps, 574 s after (model time 1015 -> 471 s). A replayed skill: 49 s -> 15 s.
Knowing the machine
- facts.rs: 電腦現況 per computer (system, desktop, systemd, sudo, network, package manager,
tools, apps, a fixed TZ), read once by a small read-only probe, kept 30 min, in the
instructions at the start of a turn.
- The box's sudo is the bot's: only what follows sudo is checked against the always-denied
list (it was refused outright, while the instructions said it could sudo), and commands with
sudo run as the box's root (setuid does nothing under gVisor).
- Box image v8 drops ENV TZ=UTC (it overrode /etc/localtime). ANYBOT_BOX_IMAGE /
ANYBOT_BOX_VOLUME give a throwaway box its own image and volume, never adopting the real one.
- Mac shell sandbox: App Store app CLIs (they carry their own app sandbox) and ps/top run
outside it; plain `osascript -e` too, asked the first time even in direct mode, refusing do
shell script, terminals, keystrokes, file reads, ObjC and the secret folders. A command the
sandbox broke says so, instead of looking like a broken program.
Reaching what it sees
- The Mac menu bar's status icons are numbered (each app's AXExtrasMenuBar, named from inside
for the system's own), computer_use_zoom looks closer at full sharpness.
- Numbers stay with their controls from look to look (they were reassigned every look, so a
step's second click landed elsewhere); the box's front window is listed first; a control can
be named instead of numbered (name "4").
- Web page numbering walks open shadow DOM (Chromium settings, web components).
- A box on a Docker host elsewhere retries the next port when the usual one is taken (a retry
without it came up with no network).
Working like a person
- Instructions: pick the route, know what done looks like, one route, check once, say what's
stuck instead of faking it with a substitute, ask before cutting the connection.
- StepPlan: the bot's model takes the first step, the step after trouble (failed, refused,
repeated, nothing changed on screen, the quick model switching routes), one in four, and the
final check; a quick model (Bot.step_model, auto: grok-4.20-0309-non-reasoning on xAI) does
the planned steps in between, falling back to the bot's model if it can't be used.
- Skills keep the steps that worked (with what each acted on, not numbers); replay_skill does
them again with no model in between, finding each control again, handing back on a mismatch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uirha5F5gJt93t3yEx3XcS
From real runs on the cloud brain with the person's Mac (Threads patrol, a post draft, "show me my
screen"):
- computer_use_show: a plain picture of any computer's screen goes into the chat (the bot sees it
too). The bot used to describe a screenshot in words and fail to send it.
- The phone shows the bot's attachments: pictures in place, other files download on a tap (api/file,
only from the chats' file folder). It showed nothing for the bot and names only for the person.
- send_file works from the person's other computers (fetch_file over the relay, 20 MB, secrets stay).
- The same call with the same answer again and again is warned the 3rd time and stopped from the
4th (a blank tab was read 11 times, then the bot said Threads was logged out without opening it);
a blank tab says so; checking a login means opening the site.
- Screenshot lists: a nameless row around a named button isn't numbered, a nameless row takes the
words inside it, empty spacer rows aren't numbered (System Settings: 63 numbers -> 33, all named).
- Writing "in my voice" reads the person's own recent posts first.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uirha2RVfSYmrhqNnqx9Z4
Web
- web_fetch kept pages with a <header> only up to it (taken for <head>): articles came back as their menus.
- browser_open waits for the page to load and returns it like browser_read (one step fewer per page);
actions wait for the page to hold still instead of fixed 1-2.5 s sleeps; one way to reach the tab.
The bot's computer (box v6)
- Numbers only on what can be seen: clipped to the window and the page's visible area, none under
another app's window (a link scrolled under the toolbar was numbered on Back).
- press_key with a typed character no longer swallows the keys after it (xdotool type --args 1).
- The person's input through noVNC/TigerVNC is noticed again (fence, continuous updates, QEMU keys).
- "Nothing changed" only after clicks, scrolls and drags; long text is pasted; one screen step at a
time; one docker inspect per step when the box is known ready; poppler-utils for long PDFs.
- One typing path and fewer round trips (stdin variant, unused picture path, extra batch execs gone).
The person's computers
- Linux: the window system's front app when it has one, and the reason when there is none (was 「」);
hidden and off-screen controls get no number; wording no longer names a tool that doesn't exist.
The loop
- Without Docker, the bot's own computer runs no commands where there is no sandbox (Linux, Windows):
it was a plain shell on the brain itself, keys in reach, asked nothing.
- The bot's own computer is re-checked each step; unreachable computers are failures, not safety
blocks; no direct-mode note on every result; an empty reply is asked again; scrolling and keys
aren't counted as going in circles; steps aren't moved to a computer that can't do them.
- One finish() for every step's result, keep_grant(), dead appctl list_computers/SERVING and the
impossible server branches removed; duplicated prompt instructions merged.
Tests: 162 (Mac) / 166 (Linux) unit; real box drills 6/6 on box v6; a brain without Chrome browses
on its box; remote E2E 28/28.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uirha2RVfSYmrhqNnqx9Z4
A bot asked to open a Threads profile clicked an empty strip ten times: it guessed coordinates
from a picture, and the web tools' description told it they didn't work on its own computer.
- marks.rs: the picture a bot gets has every button, link, field and menu item boxed and
numbered (a tiny built-in digit font, pure Rust), with the same numbers in a list. click /
typing / scroll / mouse_move take the number (ref) and land on the control's centre.
- Where the controls come from: macOS AX and Linux AT-SPI through agent-desktop on a person's
computer (desktop::elements); AT-SPI in the bots' own computer (box/a11y, after XBot's
box-a11y), web pages included. On a Mac a control whose spot another app's window covers gets
no number (topmost window at its centre must be the app's; the Dock's invisible full-screen
window doesn't count). open_app waits until the app is really in front.
- No going round in circles: a step that changed nothing says so; the same step (clicks within
16 px) a third time in a turn comes with a warning, from the fourth time it isn't run.
- The box (version 4): one TigerVNC server instead of Xvfb + x11vnc, the AT-SPI bus and registry
at boot, one browser launcher (box/browser, after XBot's box-chrome: same profile and CDP,
renderer accessibility, a launch lock, stale profile locks cleared). One docker exec per look
(list + picture, ~0.4 s). A brain updates an older box by itself when it has been idle a
minute (files and logins stay); a new box is never made from an older image.
- Gone: computer_use_read_app (the list is in every picture), computer_use_mouse_button, the
per-call picture for the phone (it gets a plain one), the stale browser_read note; the prompt
says web work goes through browser_* by number, and never to read browser cookie or history
files.
Tests: marks (numbering, drawing, fingerprints), box_numbers_are_exact on a box from the new
image (a Threads-like page: the icon links get numbers, Profile clicked by number, Chinese typed
into the numbered field, a click on nothing called out), box_screen_is_exact, VNC e2e 11/11 on
TigerVNC, Calculator 7×8=56 by numbers through the signed app on this Mac, remote_e2e 28/28,
153 unit tests; Linux build + 157 tests + x11_live.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The core: the bot works the computer; a person watches it like a colleague, and takes over
when it needs them. VNC is made for that (only what changed, every mouse move and key, touch on
phones), so the pictures sent one by one go, and with them their own input path.
- vnc.rs: a VNC proxy to the box's x11vnc through one docker exec (a Python bridge that ends
with either side). It signs in to the box with the box's own password (its VNC_PASSWORD) and
offers the viewer no password, so the password never leaves the brain. The person's clicks
and keys make the bot's steps wait until 8 s after they stop; text typed in the viewer's box
(any language) is pasted in the box.
- Account server: /screen (the viewer page) and /novnc.js (noVNC 1.6.0, MPL-2.0, bundled);
/relay/vnc/<host> takes the browser's WebSocket (account session, or a 10-minute ticket the
app asks for with POST /relay/screen-ticket; owners and admins) and carries it over that
AnyBot's lasting connection (vopen / vdata / vtext / vclose). Both ends of that connection
wait 2 ms instead of 40 ms to read while things move: a request and its answer took up to
160 ms more than needed.
- Phone page: 看 Bot 的電腦 and 接管 open the viewer page (also on the local network: /api/vnc on
the AnyBot's own phone server). This Mac's own screen moved to the ⋯ menu.
- Mac app: 看畫面 opens the viewer in the browser (a screen of a brain: the brain's computer by
ticket; else this computer's box, noVNC straight from it). The in-app picture viewer, the
live JPEG stream, the person's click/move/scroll/key endpoints and the input shell are gone.
- Opening a Mac app uses LaunchServices (`open -a`): agent-desktop's launch waited on the window
through accessibility and failed without 輔助使用 now and then.
Measured on this Mac (made-up account, local relay + headless brain + the real box): viewer
signed in in 0.4–0.7 s, a request and its answer through relay + brain + box 25–58 ms median;
headless Chrome shows the desktop live in /screen. remote_e2e 28/28 ×2, 149 unit tests, relay 11.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Acting by an app's number (read_app's e7) is now a real click at that control's spot, the
same path as screenshot coordinates. agent-desktop only reads; its own actions (a second set
with its own key names and failures: no typing in terminals, "/" and "`" failing off the main
thread) are gone, with computer_use_select, background clicks and press_key's app.
- Every action comes back with the screen after it (or, with no picture allowed, the front
app's numbered controls), so the bot no longer spends a step on a screenshot to check.
- Coordinates map without remembering a screenshot: the picture's size depends only on the
screen. The phone's taps go straight to the screen as the person's, in the pixels of the
picture it was sent; they used to borrow the bot's last screenshot scale and landed 1.25x off
on the bots' computer, or failed with "screenshot first".
- Mac typing pastes (⌘V, clipboard put back): with 注音 on, text sent as key events vanished
while return arrived, so commands typed into a terminal ran empty. Bare printing keys
("slash", "r") are typed the same way; keys like "slash" and "grave" have names.
- The bots' Docker computer: one docker exec per step (action, settle, picture), never
`mousemove --sync` (it waits forever when the pointer is already there: a second click on
the same spot hung two minutes), open_app (chromium = the browser browser_* drive, xterm).
Tests: box_screen_is_exact (real Docker: bot and phone coordinates exact, same spot twice in
under 2 s, ASCII + Chinese + named key typed), Calculator by numbers and Terminal typing under
注音 through the signed app, remote_e2e 28/28, Linux build + 151 tests + x11_live.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Read apps as numbered element lists (computer_use_read_app), act by ref
(click/typing/select/scroll/press_key/open_app) via agent-desktop v1.1.0.
Verified on macOS (Calculator, Finder) and headless Linux (Xvfb+openbox+AT-SPI).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Uirha3oEtWeHh58zA7ErHa
王性驊