Remote session views stop receiving each streamed delta twice. open_session for a remote create, open, or fork now goes on the endpoint dialer (a fresh connection per call, closed once it settles), instead of sharing the long-lived connection the later session.attach rides on; opening on the shared connection had attached the host's delegated engine connection to the session as well, so the engine wrote every delta twice. A connection a caller lends is now named borrowed, so the CLI's one shared connection survives the open. (#13761, adopting #13075 by @surlymochan)
The first session on a newly enabled device now starts memory sync. Sync used to start before the memory repository existed, fail to write its space binding, log "memory sync failed to start" and never retry; it now waits for the repository and starts on its own once the first memory write creates it. (#13493, fixes #13486) (#13627)
Memory sync drops a replaced desktop account cleanly. When the desktop moved to another account without announcing a retirement, the engine opened the new account but left the old one half alive: it kept listening on the desktop channel and its open sync sessions did not end. The old account now ends as soon as the engine sees it is no longer current, which also ends its sessions. In a session Worker, an operation started just before the desktop retired its account can no longer hand back a handle from the old account. (#13334, follow-up to #13327 and #13329) (#13627)
OmO is ready to load on the tool side of a split session: its first-paint startup gate is declared for the agent-loop side, so the tool side never registers those handlers, and on the tool side deferred startup work runs right away instead of waiting for a paint that never comes; local sessions register exactly as before. (#12329) (#13627)
Computer use is ready for the tool side of a split session: foreground control is revoked, and the engine session ended, when the run of the session holding it ends or that session closes, through that session's own lifecycle instead of loop events; local sessions behave exactly as before. (#12328) (#13627)
Adds the executable proof that computer use returns one screenshot per tool call, under today's capture caps and unchanged across the tool side of a split session; no behavior change. (#12337) (#13627)
Computer use is ready for the tool side of a split session: there, foreground control follows the device's own exec setting (refused until exec is enabled on that device, then granted to any session without a per-session confirm); local sessions keep today's confirm exactly. (#12332) (#13627)
Computer use is ready for the tool side of a split session: sessions sharing one device take its foreground one at a time, and only the holder can release it; local sessions behave exactly as before. (#12339) (#13627)
Computer use is ready for the tool side of a split session: a missing OS permission (Screen Recording, Accessibility) raised there is reported to the session that owns the call, once per permission, without a session marker on that side; local sessions behave exactly as before. (#12450) (#13627)
Computer use v3: the result contract refuses a receipt under an unverified delivery and types refusal reasons. An ActionReport that carries a consumption receipt under not_delivered, delivered_unverified or unknown/unknown (#13147) is now refused, both when it is built and when it is parsed from the wire. A backend refusal's reason is the closed Reason type, so a mistyped reason is a compile error. (#13137)
- CI: a computer-use bench regression job compares the desktop bench against a base and fails on a p50 regression above 30%; it runs nightly against the newest release tag and on manual dispatch, not on pull requests. (#12876)
A question or approval a remote session asks now reaches every omo remote attach on it, not only the one that sent the prompt. The terminal UI shows the request and sends your answer back by its request id, as it already did for prompts you sent yourself, and dismisses it when another device answers it first. A secret input or a grant approval is still shown only on the terminal whose prompt raised it. A request already waiting when you attach, or raised while the relay link was down, is shown too, and one that was settled during a relay gap is dismissed after the reconnect (#13474). (#12750)
Changed
- The ulw-plan plan template and reference, ulw-execute's decomposition step, and the Codex edition's ulw-loop reference and plan agent no longer say "target 5-8 per wave; fewer than 3 means under-splitting". A wave is sized to the tasks that are independent of each other. (#13851)
- The mass-ulw skill's listing description now says a graph is sized to the request with one run per phase, matching the rewritten skill body; it no longer says each phase's stages are defined up front in one run. (#13848)
- ulw-loop no longer tells HEAVY goals to target 5-8 tasks per wave or treats fewer than 3 as under-splitting; a wave is sized to the tasks that are independent of each other. Mass research gives lead nodes the same handoff file and source tagging as harvest nodes, and both ulw-research editions make slice reducers conditional like the mass-ulw planning reference does. (#13847)
- mass-ulw now sizes the workflow graph to the request: a small graph by default, a verification node only when the final check is too long to run inline, per-item chains instead of a review node between waves, and mass research that starts from the angles the topic has instead of a 60-angle floor. The skill and its planning reference were rewritten and are about a quarter shorter. (#13834)
- Internal: the desktop service's fake engine and test harness are now the
./testingpackage export ofsenpi-desktop-service, and the computer-use tests inomo-senpiandsenpi-desktop-toolimport them from there instead of a relative path into its test tree. Nothing changes for users. (fixes #13614) (#13712) - Memory sync: a test now covers a remote head that arrives on the stream while a sync cycle is running and is not applied by it, so the engine must run exactly one more cycle to pick it up; a lost wakeup there now fails the test. No behavior change. (#13656, fixes #11778)
- Remote sessions are now on by default. The moment
omo logincompletes it switches them on for this device and says so in one line ("Remote is on for this device: other signed-in devices can reach it. Turn off: omo remote disable"), and every omo start keeps the device set up.omo remote disableturns them off and remembers that, so neither a lateromo loginnor an omo start turns them on again;omo remote enableturns them back on (#13594). (#13602) (#13627) - device-service-core's local-API tests fail when omo starts sending a request type that has no reply row in the LA21 decoding test or that the conformance test owner still lists as unserved (#13491). (#13491)
omo remote, the thread tools and the memory sync use the device service's own local-API client, vendored byte for byte from omo-relay-daemon (main after #1115, plus the read-side tolerance of omo-relay-daemon#1130), instead of a hand-edited copy; a guard test keeps that client and the connection factories behind the device-service layer (#13475). (#13475)
Fixed
- A fanout that fails as a whole (an unknown or ambiguous target, invalid options, a duplicate target under
dedup: "reject") and a target that runs pasttimeoutMsget the same typed error, and a failedresultOfnames the request instead of treating its op id as a device. (#13813) - The shipped device account, device service, memory sync, wake, thread client and bunshin bundles, and the omo-ai launcher's sign-in module (
omo login/logout/whoami), are rebuilt from their current sources, so the released plugin and launcher no longer lag the code merged into dev. (#13810) - The shipped thread client and thread SDK bundles now carry the session-only model change from #13494, so switching a remote session's model no longer rewrites the target device's default model in the released build. (#13786)
- Memory: a contender recovering a lock from a dead owner no longer fails when that lock file is already gone by the time it removes it (seen on Windows, where memory reflection supervisors then failed admission). Any other removal error still surfaces. (#13745, fixes #11644)
- Device operations run through the bundled fleet SDK (
bunshinin eval cells) now fail with a typed error,{ code, message, next_action }like the thread tools, instead of an internal transport sentence. The message names the operation and the device, and the code tells a wrong argument (invalid_arguments) apart from an offline device (device_offline), a policy refusal (scope_denied) and a lost connection (transport_closed). The relay's own op-result codes pass through unchanged and carryuncertainwhen the device may have run the op. The "no transport" error and a failed connect now say what to do next. (#13744) - Computer use on X11 no longer reports a background click, scroll, drag, typing or key chord as delivered when the target app drops synthetic input (GTK 3/4, Qt, LibreOffice, Chromium, Electron and Firefox apps whose window class names no toolkit, and xterm, whose
allowSendEventssetting no other app can read). Background input now goes only to apps whose toolkit is known to accept it (Xt and Motif apps other than xterm, Tk); any other app is refused before anything is sent, with the element action and foreground delivery offered as the next step. (#13736, addresses #9109) - On Windows, background keyboard input that holds Ctrl, Alt, Shift or Win while another key goes down is now refused with
BackgroundUnavailableand nothing sent, instead of being reported delivered while the application received only the bare key. This covers key chords such asctrl+k, bounded key and mouse holds with such keys, and a single character the keyboard layout types with Shift or AltGr (for exampleK, or@on a German layout). Send it withdelivery: "foreground"or use accessibility actions. (#13734, fixes #13297) - On Linux, the computer tool's
desktop.elementAt(x, y)now returns the element under the point, such as a GTK window's Press button, instead of the window's untitled rootfillerat every point. (#13732, fixes #13296) - With real hit-test answers, the check that an element is not covered before an accessibility action now works on Linux too: another element of the same window across the target's centre refuses the action with
occluded, as on macOS. The target's own descendants, such as a list row wider than the list's viewport, never count as covering it, on any platform where the backend can compare elements. (#13732) - Memory that two devices already held as the same text, such as the default files every memory starts with, no longer appears twice after the devices start syncing one space. A device joining a space now merges its own copy of a file into the shared one line by line, keeping every line of both and the lines they share once. (#13730, fixes #13725)
- A launch spec that lists no extensions now gives its host the same runtime build id as
omo host ensurepredicts. The compiled binary still added its own plugin folder to that host's command line, so the two ids disagreed for such a spec. (#13729) - Memory in a folder with no git remote, and memory from a session outside any project, now syncs between your signed-in devices through one account default space instead of staying on the device that wrote it. A repository that had its own per-device space moves to the account space once and sends its memory there. (#13726, fixes #13720)
- A host the compiled
omobinary starts now reports the same runtime build id as theomo host ensurethat started it. The binary added its own plugin folder to the host's command line on top of the launch spec, so with no spec, or with a copy of the spec, the host and the CLI disagreed: the desktop app judged its own engine a build mismatch and every handover's replacement host exited. (#13700) - The omo-native trust test suites no longer fail when the test runner's
TMPDIRis group-writable: each suite now runs with a private temporary directory the trust stage accepts (#13697). (#13698) omo remote open --modelnow sets the model on the session it just created even when its durable ID differs from its relay listing key. Send, model, interrupt, close and attach accept both the durable ID returned byopen --jsonand<device>/<session>; the durable ID is the documented canonical identity. (#13666)omo remotenow prints the credential store's own message when it cannot read your sign-in, such as theomo account forgethint for an invalid stored sign-in or the note that the OS credential store is locked or unreachable, instead of only the generic notice. (#13644, fixes #13638)omo remote wake --helpnow gives the switch's real default: waking is on unless you turn it off. It said "off unless turned on", which never matched the device service. (#13637)- A session host start no longer times out when its supervisor or host child has to re-run the engine trust step: the re-run keeps the startup pipe the launcher reads, instead of losing it and reporting
supervisor_bootstrap/host_bootstraptimeouts for a healthy host (#13622). (#13632) - Thread gateway: the omo daemon registers this device's agent dir at start-up even when senpi reports its endpoint registry as a relative or unnormalized path. It falls back to the launcher's agent dir with one warning instead of registering nothing (#13604). (#13609) (#13627)
- Memory keeps working after an extension hot reload, for example when a watched
SKILL.mdchanges (#13098). The reloaded extension no longer fails to register memory withsession worker data "omo.memory-sync.v1" is already provided. (#13101) (#13627) - omob dev builds bundle one senpi engine again: the build linked the source-built engine only at the root while omo-native, omo-senpi, senpi-task and omo-opencode kept the registry copy, so the trust hand-off was lost and every coordinated session host start timed out in
supervisor_bootstrap(#13620). (#13621) thread_set_model,omo remote modelandomo remote open --modelchanged the target device's default model along with the session's. The engine record now carriesscope: "session", so only the addressed session switches models.thread_set_reasoninglikewise no longer rewrites the level the device remembers for the model; it changes only that session's level. (#13494)- A waiting
omo remote sendno longer reports its turn as interrupted because an earlier run in the same session was interrupted. (#13481) - A waiting
omo remote sendno longer prints the turn's answer twice when the answer is both streamed and already stored in the transcript: the transcript read is not printed once the stream printed this turn's text. (#13481) - A waiting
omo remote sendno longer prints an earlier identical message's answer as its own when that earlier run finished just before the send was answered: only a finish that follows this send's own prompt ends its wait. (#13481) - A waiting
omo remote sendno longer prints an earlier identical message's answer when that earlier message carried no client message id (a send without--wait, text typed in the TUI), and a send whose interrupted turn was followed by a later message still prints its partial answer after the later turn's text. (#13481) - A waiting
omo remote sendto an engine that stores no client message id prints a turn that finished before the send was answered and exits, instead of waiting until the session closes; an interrupted turn whose partial answer only partly streamed prints the rest of it. (#13481) - A first waiting
omo remote sendinto a session whose earlier identical message was typed in the TUI or sent without--waitno longer prints that earlier message's answer when the engine stores no client message id: a message already in the session before the send is never taken as the send's own (an identical message stored between the send starting and its prompt landing is still indistinguishable, oh-my-openagent#13580). (#13481) - While OmO accounts are closed,
omo logout --helpandomo whoami --helpprint their own help and exit 0, asomo login --helpdoes, instead of the login refusal; any other argument is still refused. (#13466, fixes #13461) (#13466) - In the standalone omo binary, the kill guard's check (
omo __kill-guard check ..., run by the guard before akill,pkillorkillall) now reaches the engine's judge instead of being read as a chat prompt, so the guard allows or refuses the signal as it does in the npm install. (#13460, fixes #13449) (#13460) thread_listandomo thread listnow list threads on all of the account's devices in every call, not only withall_scope. An ended thread on another device shows as resumable instead of live. The result includes the account's devices: name, platform, kind, online state, last seen, thread count, and whether this listing reached them, so a device that did not answer is named instead of silently missing.omo thread list --jsonnow prints that whole result as one object (threads,devices,remote_unavailable) instead of a bare array of threads, and the plain listing names each other device's thread's device and prints one line per device. (#13394, fixes #13392)- Computer use v3 (fake backend):
inject()now refuses a scripted failure whose code and reason have no row in the error table, the same check the scenario loader applies, instead of failing later. A capture whose extent is under one whole pixel (an empty size or region, or any length that scales to less than 1 px), or whose scale is zero or negative, or whose size or region is negative, now refusesCaptureFailedbefore allocating instead of returning a 0x0 frame or, for a negative size times a negative scale, a positive one. The fake's source guard now also bans environment reads, including a groupeduse std::{...}import of the module (on one line or split across lines) and any path into it. (#13133) (#13120) - cu-eval live runs now restore and tear down every task on every exit path, close the apps before deleting their scratch files, and list any restore step that failed in the row's
restoreErrors. (#12849)
Added
-
The relay transport re-vendored in #13744 also brought in new public surface:
createRelayTransporttakesretryUnreachableMs(how long a device whose describe failed is listed unsupported before it is asked again, default 30 s), and the transport it returns hasuploadArtifact(bytes)andfetchArtifact, which carrypushTree/pullTreearchives over the relay. (#13813) -
A new thread can be started with its model, thinking level, approval mode and first message in one step, on this device or on another of your devices:
thread_createtakesmodel,reasoning,permission_preset,promptanddelivery, andomo remote opentakes--model,--thinking,--worktree(with--branchand--base) and--prompt. A model the device cannot run now refuses the create and leaves no session behind, instead of opening one on the default model; an agent can keep or narrow its own approval mode for the thread it creates, never widen it. (#13748) -
The session id
omo remote openprints now works with every otheromo remoteverb right away.send,model,interrupt,closeandattachaccept the durablethread_idon its own (its device is found in the account listing) or as<device>/<session>, and a session opened a moment ago is no longer answered "no session". A thread created with a first message gets that message even before its device lists the new session. (#13748) -
An agent can be woken by a session on another device of your account:
monitor({ type: "rpc", target: "<device>/<session>" })reports its status changes and turn ends as they happen, and after a relay drop, a device-service restart or an engine restart it reports what changed meanwhile, once.<session>can be the thread idthread_listshows, an alias, or the session's key;<device>/*follows every session on that device. (#13743, #12719) -
thread_readtakeswait_ms: for a thread on another device it first waits, up to that bound, until it sees the thread's turn end (not a momentary idle), then returns the transcript with how the wait ended (settledwith theturn_idit saw end,timeout,cancelledorended). (#13743, #13417) -
/remoteinside the TUI: list this account's devices and sessions, open a session on another device, send it a message and see the reply, interrupt or close it, change its model, and turn remote on or off, without leaving OmO. A bare/remotewalks you from a device to a session to what to do with it./remote attachnames the command to run in a new terminal. (#13741) -
Memory sync can now parse, derive and redact recovery-phrase keys: a 24-word BIP39 English encoding of 32 random bytes, keys derived with HKDF-SHA-256 bound to the account, and fixed test vectors. Nothing creates a phrase key yet; turning on sync with a recovery phrase lands in a later change. (#13330) (#13627)
-
Memory sync can create a recovery-phrase first device: turning sync on takes a method (passkey or recovery phrase), shows a void notice that survives crashes when a shown phrase was not saved, and sends a sync capability header. Creating a phrase stays off behind a client flag and a service floor check until the service side is ready. (#13335) (#13627)
-
Memory sync can recover with a recovery phrase and enroll, rotate or remove phrase keys from a member device; creating a phrase stays behind the client flag and service floor check. (#13336) (#13627)
-
macOS live QA: a Chrome two-window typing scenario checks that background typing lands in the second window and leaves the first one untouched and key. (#12823) (#13627)
-
macOS live QA: a dense-window observe scenario checks that a page with 5,000+ accessibility nodes returns a partial snapshot at the deadline, drills into a cut, and keeps the default skeleton within its token budget. (#12823) (#13627)
-
The live macOS QA driver's
--listflag prints every scenario name on any host. (#12823) (#13627) -
The live macOS QA driver's
--dense-rows <n>flag sets how many rows the dense-window page renders (default 2,000). (#12823) (#13627) -
Desktop engine: a request size cap, a frame cap and an idle flush keep long sessions bounded (P12b). (#12807) (#13627)
-
Computer use v3: spike SP-B ships a probe ledger that classifies the ten bunshin hosting gaps as closed, open with a plan row, or needing an installer change, and records which host owes each condition it cannot measure. No engine code changes and nothing user-visible changes. (#13156)
-
Computer use v3: the SP-T spike probe for macOS lands under
crates/omo-computer-backend-macos/probes/. It is a Bun probe with shell helpers and a ledger, and it asks which process TCC charges for the computer-use launcher, whichclient_typethe grant records, whether a grant survives a generation swap and a rebuild, and which launchd session the launcher lands in. Every ledger row is still unmeasured, waiting for a macOS 26 host with a QA account and a person at its session. No engine code and nothing user-visible changes. (#13143) -
Computer use v3: the scriptable fake backend lands in
omo-computer-backend-fake. It provides one scenario format with typed, JSON-pointer refusals for an unknown key, a non-object root, a dangling window and an error-code pair the table does not have; a recording sink that writes every OS-level side effect as one line; a fake clock driven only byOMO_COMPUTER_FAKE_CLOCK; timelines for loading indicators, observer events and user input; a fault FIFO whose refusals resolve through theCOMPUTER_*table; andmacos,x11,waylandandwin32personas. Every scenario field has a fixture and a null control. Nothing user-visible changes yet; the existing desktop engine is untouched. (#13120) -
A latency bench and the 7-task cu-eval for computer-use v2 (P13b). (#12849)
-
7fcf9db ci(publish): gate-reuse accepts a maintainer CI-basis receipt for the prepared release head (#13886)
-
402ffaa Merge pull request #13800 from code-yeongyu/fix/12643-secret-service-none-bus
-
aca6ed8 Merge pull request #13848 from code-yeongyu/docs/13842-mass-ulw-listing-description
-
10a7e36 Merge pull request #13832 from code-yeongyu/rt-u4/s3-on-dev
-
8b7fc88 Merge pull request #13813 from code-yeongyu/fix/13406-review-lows
-
7b81414 Merge pull request #13826 from code-yeongyu/fix/13484-parity-pair-test-code-only
-
7bc6f43 Merge pull request #13730 from code-yeongyu/fix/memory-sync-adopt-history-less-13725
-
96ecdd6 Merge pull request #13791 from code-yeongyu/rt-u4/13771-lows
-
9ed7fbc test(remote): S3 lists dev's two new thread surfaces (attach-pool.ts, remote-watch.ts) for U5
-
ab6d027 Merge remote-tracking branch 'origin/dev' into rt-u4/s3-on-dev
-
7e0c6da Merge pull request #13743 from code-yeongyu/feat/remote-session-wait
-
b24d167 test(remote): S3 holds the TUI's /remote (components/remote/tui) as a surface; imports between surface files are not the layer; the rule states its scope (#13832 review M1, L1, L2)
-
7e14985 Merge branch 'dev' into fix/13484-parity-pair-test-code-only
-
085b855 Merge pull request #13852 from code-yeongyu/ci/pr-bots-review-lows
-
db18c69 Merge pull request #13843 from code-yeongyu/fix/13801-review-lows
-
b0825e9 test(eq): carry the mass-ulw listing sentence into the darwin-arm64 EQ.omo goldens
-
fa6c467 test(eq): re-record the darwin-arm64 EQ.omo goldens at dev 9af8d30 with a physical temp dir
-
871a808 Merge pull request #13120 from code-yeongyu/feat/cu-v3-row14-fake-backend
-
5a7c887 Merge pull request #13833 from code-yeongyu/ci/platform-red-writer-no-drop
-
97c5194 test(eq): carry the darwin-arm64 EQ.omo goldens to the merged computer.md skill digest
-
d6a7e0d test(eq): record the linux-x64 EQ.omo goldens for the mass-ulw listing description
-
b69f5a2 Merge remote-tracking branch 'origin/dev' into docs/13842-mass-ulw-listing-description
-
18e650b Merge remote-tracking branch 'origin/dev' into ci/platform-red-writer-no-drop
-
90a79b7 chore(fmt): keep four generate-assets.ts outputs out of bun fmt (fmt:check red on dev b9be6e5)
-
9af8d30 Merge pull request #13851 from code-yeongyu/docs/13846-wave-size-floors
-
d1f028e ci(pr-bots): fold in the review LOWs from #13841 and #13845 (#13840)
-
1cf026a test(device-service-core): keep this PR's send patterns through the dev merge (the merge took dev's older sent-request-types.ts)
-
0160ca6 test(eq): record the darwin-arm64 EQ.omo goldens for the mass-ulw listing description
-
b747ebb Merge remote-tracking branch 'origin/dev' into local/13791-dev
-
e96f0c2 Merge pull request #13845 from code-yeongyu/ci/fold-pr-bots
-
577f3a4 Merge pull request #13841 from code-yeongyu/ci/bot-runs-per-pr-concurrency
-
0533b37 Merge pull request #13771 from code-yeongyu/rt-u4/13491-lows
-
b775c1e test(eq): re-record the darwin-arm64 EQ.omo goldens at dev 0543add
-
8b7d29c docs(mass-ulw): number harvest files by the node's index in the run
-
214e0d3 Merge remote-tracking branch 'origin/dev' into docs/13846-wave-size-floors
-
5622d72 Merge pull request #13847 from code-yeongyu/docs/13834-review-lows
-
46bcd4e docs(changes): name the changelog fragment after #13851
-
b8208c1 docs(changes): add the changelog fragment for #13850
-
5901f88 Merge origin/dev into fix/13801-review-lows
-
88f0eac Merge origin/dev (b9be6e5, 5.1.73 #13627) into feat/remote-session-wait
-
b82f67d docs(skills): size plan waves to the work in ulw-plan, ulw-execute and the Codex ulw-loop
-
a9b332f chore(memory-sync): regenerate the omo-memory-sync bundle for this PR's own delta
-
6430114 test(device-service-core): the sent-request scan sorts its unscanned list, so the result does not depend on readdir order
-
22fa153 Merge origin/dev into fix/memory-sync-adopt-history-less-13725
-
885c276 Merge remote-tracking branch 'origin/dev' into fix/13406-review-lows
-
b9be6e5 Merge pull request #13627 from code-yeongyu/chore/omo-5173-train
-
24cb52c docs(changes): add the changelog fragment for #13848
-
9decf4f docs(mass-ulw): make the skill listing description match request-sized graphs
-
30283b8 docs(changes): add the changelog fragment for #13847
-
160f150 docs(skills): fold the #13834 review findings into mass-ulw, ulw-loop, ulw-research and frontend
-
bf84999 Merge dev 0543add (includes 5775869, #13801 G1 remote-import-guard fix) into row 14 (merge only)
-
0543add Merge pull request #13834 from code-yeongyu/docs/mass-ulw-request-sized-graphs
-
afb4c34 revert(remote): drop the remote-session-verbs.ts comment edit from this follow-up (#13843 review M1)
-
ea718d2 Revert the runtime-dependency-resolution.test.ts format fix; #13843 owns it
-
23b700c Merge remote-tracking branch 'origin/dev' into w135/omo-5173-train
-
d579ab1 chore(release): regenerate bundles on dev d9ccd64 for 5.1.73
-
bddac67 Merge remote-tracking branch 'origin/dev' into local/13771-dev
-
fd4b11b Merge dev d9ccd64 (wave 2: #13810 makes dev generated-clean) into row 14 (merge only)
-
e6249a4 Merge branch 'dev' into fix/13484-parity-pair-test-code-only
-
ad6d439 Merge remote-tracking branch 'origin/dev' into local/13832-dev
-
8d31716 Merge remote-tracking branch 'origin/dev' into feat/remote-session-wait
-
b321aa9 Merge origin/dev (d9ccd64, wave 2 #13810) into feat/remote-session-wait
-
2bd6f4b chore(remote): review LOWs of #13801 and #13761 (duplicate attach test, comment, dead const)
-
b8a3e63 Merge pull request #13801 from code-yeongyu/fix/13761-remote-cli-dial
-
63c6bab Merge remote-tracking branch 'origin/dev' into fix/13406-review-lows
-
1a7c81e Merge origin/dev into chore/omo-5173-train
-
d9ccd64 Merge pull request #13810 from code-yeongyu/chore/regen-bundles-wave2
-
2ecce78 docs(mass-ulw): describe start warnings without naming which ones exist
-
cf0deed docs(changes): point the #13834 fragment at its own pull request
-
92d48de Merge remote-tracking branch 'origin/dev' into docs/mass-ulw-request-sized-graphs
-
e0fe89e docs(changes): name the changelog fragment after #13834
-
7d4d078 chore(bundle): set the omo.js budget from the remerged train-top regen (1,497,747 B + 4 KiB)
-
ea791d5 ci(platform-red-tracker): no concurrency group can drop a red run's details (#13802)
-
70bb6cb chore(release): regenerate bundles on dev 050c187 for 5.1.73
-
d84da7a docs(changes): add the changelog fragment for #13833
-
7b073ca fix(memory-sync): drop a never-sent first batch when its key joins
-
2a3b6ca docs(skills): align ulw-execute, ulw-loop and ulw-research with request-sized mass-ulw graphs
-
6cbf0c6 docs(mass-ulw): size the workflow graph to the request and make verification a check, not a node
-
d5cb453 Merge pull request #13811 from code-yeongyu/ci/pr-linux-only-13802
-
038c469 Merge origin/dev into chore/omo-5173-train
-
c731edd test(remote): S3 rows and S2 as on dev: the thread layout has not moved yet, so S2 keeps its rows and S3 lists dev's 15 surfaces (U5/U6 remove them)
-
7862411 Merge pull request #13809 from code-yeongyu/fix/localpath-test-isolation
-
89df41d fix(rpc): match parity pair tests on the parsed TypeScript AST, not on scanned text (#13826 review L1-L3)
-
36667ee test(remote): G1 rule S3, the import boundary of #13414: surfaces reach the remote layer only through components/remote/index.ts
-
f8a6b6c Merge commit '4ebb1a250a44dd3a5b05617ea7f7d3d3177f3447^2' into rt-u4/13771-reworded
-
050c187 Merge pull request #13491 from code-yeongyu/test/device-service-core-13475-lows
-
5775869 test(remote): the attach pin test waits for its endpoint's removal and names what it proves (#13801 review M1, L1)
-
c747b84 Merge pull request #12876 from code-yeongyu/cu-v2/task-38-on-e
-
e202a3f Merge pull request #13732 from code-yeongyu/fix/is6-x11-hit-test
-
c208512 fix(release): drop the dead raiser-only state filter and the duplicate #13482 release notes (review L1, L2)
-
e47b5d5 Merge origin/dev (2bc17a2) into feat/remote-session-wait
-
0978963 Merge remote-tracking branch 'origin/dev' into chore/regen-bundles-wave2
-
94f3dcf fix(rpc): a pair test counts only where the record type appears in code, never inside a comment, a string or on $type (#13484 review L1)
-
216a9bf test(remote-import-guard): freeze the hand-browser-placement fixture's import() like its sibling fixture
-
e60b3a7 test(device-service-core): knip export-checks sent-request-types.ts as an explicit entry instead of listing it unchecked (#13491 review M1)
-
2a028f7 Merge origin/dev into fix/13761-remote-cli-dial
-
6e1a29f Merge remote-tracking branch 'origin/dev' into fix/13406-review-lows
-
70e4e49 Merge pull request #13484 from code-yeongyu/docs/13478-rpc-relay-parity
-
1c122fc Merge pull request #13734 from code-yeongyu/fix/is6-win-bg-chord-truth
-
2bc17a2 Merge pull request #13807 from code-yeongyu/fix/omob-provision-race
-
c3b41df Merge pull request #13748 from code-yeongyu/rt/create-options-model-prompt
-
ec3f9be docs(changes): name the launcher sign-in module and bunshin bundle in #13810's fragment (review LOW)
-
6577e6e Merge pull request #13745 from code-yeongyu/fix/11644-stale-owner-enoent
-
b1a9f01 chore(bundles): regenerate bunshin-ts sdk after merging dev 17c48ba
-
8d4fba2 Merge remote-tracking branch 'origin/dev' into chore/regen-bundles-wave2
-
17c48ba Merge pull request #13736 from code-yeongyu/fix/is6-x11-bg-input-truth
-
bbbc6ed Merge pull request #13806 from code-yeongyu/fix/thread-list-qa-message-on-dev
-
7972966 fix(memory-sync): a joining file's adoption goes out in the same cycle
-
7c87052 chore(release): regenerate bundles on dev e478744 for 5.1.73
-
535173a Merge remote-tracking branch 'origin/dev' into fix/13406-review-lows
-
0841949 fix(guard): G1 lists remote-cli.ts's createDeviceServiceSessions; cu-eval imports the desktop SDK by a literal path
-
4670625 Merge pull request #13744 from code-yeongyu/fix/13406-typed-device-errors
-
c4d1305 test(device-service-core): the entry-export check imports through a non-literal path, so knip keeps export-checking the package entry
-
6f6e94f chore(knip): REMOTE_COMMAND_USAGE is file-local (dev #13412 left it exported and unused; knip:check red on dev 2e5a021 too)
-
11d84d8 docs(changes): fragment for dev regen wave 2 (#13810)
-
8888fcf chore(bundles): regenerate every generated file on dev 2e5a021 (regen wave 2)
-
9a637cb Merge origin/dev into rt/create-options-model-prompt (second)
-
1b7b429 chore(release): keep dev's changelog fragments on the 5.1.73 train
-
2e02d2c Merge origin/dev (e478744) into chore/omo-5173-train
-
291e0c3 Merge origin/dev into feat/remote-session-wait
-
2e5a021 Merge pull request #13741 from code-yeongyu/feat/13412-remote-tui
-
bd2f725 Merge branch 'dev' into docs/13478-rpc-relay-parity
-
ea7631e test(device-service-core): the spec-proto export pin also reads an export after a ; on the same line
-
504d4fb test(device-service-core): the send scan reports every non-literal first argument (M1) and nested type arguments (L2); the spec-proto pin sees declare and export-list forms (L1)
-
f9b1a7f fix(qa): held-then-exit's failure message counts the listing's threads (#13789 / #13394 review LOW)
-
e478744 Merge pull request #13786 from code-yeongyu/chore/regen-bundles-after-13494
-
75f073a Merge pull request #13789 from code-yeongyu/fix/thread-list-merged-devices-on-dev
-
b8739ed Merge pull request #13143 from code-yeongyu/spike/cu-v3-row9-tcc
-
150cf33 fix(omo-native): isolate concurrent executable provisioning (#13795)
-
1938378 Merge pull request #12750 from code-yeongyu/feat/remote-approvals
-
c09a616 fix: keep dev's probe-test records and type the thinking level as its literals after the dev merge
-
a2f2868 fix(omo-native): remote CLI attach uses the new DeviceServiceDial shape (dev typecheck red after #13761)
-
0bf8752 test(remote): attach by durable ID borrows the CLI's connection and leaves it open (#13761)
-
4cb21d9 fix(computer): recognise xterm by its /proc exe link too, so a renamed xterm is refused in the background (review LOW at c57bfd6)
-
16b6d03 test(computer): an xterm known only by its exe link must be refused in the background (review LOW at c57bfd6)
-
138d77a Merge pull request #13475 from code-yeongyu/feat/device-service-core-vendor-local-api
-
034c589 Merge commit '43233d01a3d22ff92f83c3f66f98026a5ffaa8e6' into rt-u4/13491
-
9e7b4e1 Merge pull request #13726 from code-yeongyu/fix/memory-sync-account-default-space-13720
-
80d5dec test(device-service-core): spec-proto export pin covers every export form; the send scan reads typed request<...>() calls (#13771 LOWs)
-
8505581 Merge remote-tracking branch 'origin/dev' into chore/regen-bundles-after-13494
-
5b1e88f Merge origin/dev into fix/thread-list-merged-devices
-
e58ce21 Merge pull request #13761 from code-yeongyu/adopt/12777-remote-view-double-delta
-
0b1843a Merge origin/dev into rt/create-options-model-prompt
-
c799c37 Merge remote-tracking branch 'origin/dev' into chore/regen-bundles-after-13494
-
43233d0 chore(device-service-core): re-pin the vendored local-API client to omo-relay-daemon main edd2ee49 (relay-daemon#1354 merged as a squash; bytes identical to 04288f85)
-
a7de4fc Merge pull request #13135 from code-yeongyu/fix/13099-guards-lstat-parent
-
7471c3d Reapply "chore(bundles): carry service-auth.generated.js here (settled with #13766)"
-
90a9f78 Merge pull request #13156 from code-yeongyu/spike/cu-v3-row11-bunshin
-
c8b0754 Revert "chore(bundles): carry service-auth.generated.js here (settled with #13766)"
-
2c9de05 chore(bundles): carry service-auth.generated.js here (settled with #13766)
-
3cc0c3a fix(memory-sync): rewind only a true first capture, and recover a crashed rewind
-
996064c chore(bundles): leave service-auth.generated.js to #13766
-
c50ceb2 docs(changes): fragment for the post-#13494 bundle regen (#13786)
-
41cdefb Merge commit '8327954099b6c4969f5afe2b4420eeb8cfd342d8' into rt-u4/13491-lows
-
43f443b Merge commit '8327954099b6c4969f5afe2b4420eeb8cfd342d8' into rt-u4/13491-lows
-
6bf353f Merge remote-tracking branch 'origin/fix/13406-typed-device-errors' into fix/13406-review-lows
-
8327954 fix(device-service-core): re-vendor local-api at omo-relay-daemon 04288f85 (relay-daemon#1354): no Bun global, so the client runs under Node
-
be1ef0b Merge pull request #13494 from code-yeongyu/fix/thread-set-model-session-scope
-
f261b92 Merge pull request #13729 from code-yeongyu/fix/13700-lows
-
67c3f2e Merge pull request #13385 from code-yeongyu/t1x/test-audit-omo-2
-
c57bfd6 test(computer): the xterm capture-click scenario clicks in the foreground (background is refused for xterm)
-
c6de204 fix(remote): only run-content records show a run; handle watch rejections on both paths (#13743 review M1, L1)
-
945ec25 test(remote): RED - idle-session records are not a run; a late watch rejection is handled (#13743 review at 89c5463)
-
0602e9c Merge remote-tracking branch 'origin/dev' into cf-13135
-
a32254e chore(bundle): set the omo.js budget from the merge-result measurement (+7,946 B for /remote)
-
3dc6604 Merge pull request #13466 from code-yeongyu/fix/closed-auth-logout-whoami-help
-
cbe1afb Merge pull request #13460 from code-yeongyu/fix/13449-compiled-kill-guard-dispatch
-
4c03fcc Merge pull request #13637 from code-yeongyu/fix/remote-wake-default-wording
-
3af4577 Merge pull request #13632 from code-yeongyu/fix/trust-exec-keeps-startup-pipe
-
8d6a1ac Merge pull request #13666 from code-yeongyu/fix/remote-open-model-session-id
-
2c4c023 fix(computer): the fake backend builds its refusals from the typed Reason that dev #13137 introduced (Refusal::new takes a Reason; no .wire() at the call site; tests compare against Some(Reason::...)); no behaviour change
-
568ef3c Merge pull request #13656 from code-yeongyu/test/11778-mid-cycle-rerun
-
b8c0ad6 Merge pull request #13698 from code-yeongyu/fix/trust-tests-private-tmpdir
-
ce7e624 Merge pull request #13481 from code-yeongyu/fix/13272-stale-stop-reason
-
c1a446a Merge pull request #13712 from code-yeongyu/fix/13614-desktop-service-testing-export
-
ef9ec6c fix(computer): refuse X11 background input to xterm, whose allowSendEvents no client can read (review LOW at 1878d94)
-
df09373 Merge dev (a6901c2, with #13137 typed Reason and #13259) into row 14 (W2 D1: one more take of dev at row 14, merge only)
-
fc17de8 test(computer): xterm background input is refused whatever its visible resources say (review LOW, 1878d94)
-
4a226a8 fix(device-service-core): re-pin the vendored local-API client to the relay daemon main 42078b7c; pin spec-proto.ts's exports (#13475 LOW)
-
228cb11 fix(device-service-core): re-pin the vendored local-API client to omo-relay-daemon main 42078b7c; pin spec-proto.ts's exports (#13475 LOW)
-
a6901c2 Merge pull request #13259 from code-yeongyu/ci/computer-live-macos-tier
-
f1a6ccc fix(desktop): an unreadable parent never makes a descendant hit count as covering (#13296)
-
6446bbe fix(remote): a record addressed by a durable id names the key its session is attached on
-
62235fe test(device-service-core): the sent-request scan reports bare request() calls too; the entry check says what it checks (#13491 LOWs)
-
3a91f3c fix(remote): any mid-run record shows a run; only status events override the baseline (#13743 review M1, L1)
-
24403f4 test(remote): RED - a run that started before the wait attached is not settled by the quiet window (#13743 review M1)
-
55d1b6e docs(changes): name holds and layout-shifted characters in the #13734 fragment
-
b439831 fix(computer): refuse a background mouse hold's modifier chord before mapping its point
-
7ec816b test(computer): a background mouse hold with ctrl+k is refused as a modifier chord, and modifier-only chords pass
-
e7457dd fix(remote): a durable id whose locate fails keeps that failure, never unknown_session (review M1/L1 of 6ce7397)
-
3b8aa04 fix(remote): start the wait deadline and abort before the attach; stop a watch that opens late (#13743 review M1, L1)
-
71fe91a test(remote): RED - a wait's deadline and abort cover the watch attach (#13743 review M1, L1)
-
3fbbc17 Merge pull request #13137 from code-yeongyu/fix/cu-v3-13118-contract-hardening
-
b9ecc90 fix(service-auth): the private-bus probe tests use a bus that cannot activate services (#12643)
-
6e56cb7 test(service-auth): run the private-bus probes beside an activatable Secret Service that never answers (#12643)
-
ff68e56 chore(bundle): raise the omo.js budget for /remote in the TUI (+6.8 KB, #13412)
-
11ba0c9 fix(remote): trim the wait_ms strings so omo.js stays inside its byte budget (#13743)
-
6ce7397 chore(bundle): raise the omo.js budget for thread_create's launch options (+4.0 KB)
-
b383ba6 fix(remote): never send a durable id to session.attach; locate it first (relay-daemon#1350)
-
25e22cb docs(changelog): the #12777 fragment names #13761, which adopts #13075
-
920c16f chore(omo-native): regenerate the service auth bundle for the wait fix (#13743)
-
0b14d45 Merge commit '83d42fed42d148abc46550ce8ed71c547860fffa' into rt-u4/13491
-
83d42fe test(script): the knip coverage floor stops counting the vendored local-API client (reviewed lowering)
-
786bbd9 build(remote): regenerate service-auth.generated.js for the borrowed CLI connection and the open_session dialer (#12777)
-
d4fae1b style(thread): format the merged lookup-answers test fakes
-
ddd7614 fix(thread): only open_session rides its own short-lived connection; every other endpoint record reuses the shared one (#12777)
-
6bd4dc0 test(script): the shared-core guard allows omo's moved Windows exact-DACL writer by exact path
-
8ed7ba3 Revert the service-auth.generated.js regen; the base fix lands in its own dev PR (#13484)
-
1a7cb5a fix(remote): thread_read wait_ms settles on an observed turn end; a snapshot idle is not one (#13417)
-
69cd9c7 test(remote): RED - a wait settles on an observed turn end, never on a snapshot's idle (#13417)
-
dbc45b4 chore(service-auth): regenerate service-auth.generated.js; dev 3975913 committed it stale against script/build-service-auth.ts
-
80a5e9b Merge dev into cu-v2/task-38-on-e
-
587f207 Merge pull request #12849 from code-yeongyu/cu-v2/task-35-p13b-latency-bench-and-cu-eval-7-tasks-3
-
0571de8 fix(remote): every remote verb accepts the session id open printed (durable or listing, bare or device-qualified)
-
f4c3a53 fix(remote): stamp forwarded records with the monitor target label (#13743 live proof)
-
73336b4 test(remote): RED - a monitored remote record carries the target label, not the engine's routing id (#13743 live proof)
-
ec675da fix(thread): a create's first prompt addresses the new remote session by the relay key its open named
-
aa9d674 Merge #13475 head 3c38fd7 (dev + re-vendor at relay-daemon eda08c78) into #13491
-
3c38fd7 Merge origin/dev into feat/device-service-core-vendor-local-api: re-vendor local-api at omo-relay-daemon eda08c78 (relay-daemon#1345)
-
68d6294 chore(changes): name the fragment by its PR (#13748)
-
a6da1d8 fix(omo-native): omo thread list --json prints threads, devices and remote_unavailable; the plain listing names each device (#13392)
-
8ab4ed7 test(omo-native): omo thread list prints the whole listing with the account's devices (#13392; RED)
-
30ff3e8 chore(omo-native): regenerate the service auth bundle (#13743)
-
b152d6e fix(computer): the background key hold takes the same modifier refusal as the chord
-
36b2e4d test(remote): open --model owner tests follow the one-open contract (#13409)
-
e9c9d28 docs(changes): name the fragment after the PR's real number (#13745)
-
7c52f52 feat(thread): create a thread with its model, reasoning, approval mode, worktree and first prompt in one call
-
d168af9 test(thread): a non-open endpoint command rides the shared connection; only open_session dials its own (#12777)
-
3659b4e Merge origin/dev into fix/12777-remote-view-double-delta
-
1878d94 fix(computer): trust an X-Resource pid only when the server sees this engine under its own pid (review L2)
-
16429ed feat(remote): wake on and wait for a session on another device - watch v2, monitor rpc /, thread_read wait_ms (#12719, #13417)
-
34985e9 test(remote): RED - watch v2, monitor rpc remote targets, thread_read wait_ms (#12719, #13417)
-
e317d29 test(computer): an X-Resource pid from another pid namespace must not be trusted (review L2)
-
7e1ed14 fix(computer): check the client of the window a background pointer event lands in, not only the top-level (review L3)
-
678d313 fix(remote): a dialog pending before the attach or across a relay gap is answerable from the attach (fixes #13474)
-
8700cf0 fix(remote): the lazy surface passes the sessions device filter through
-
0de89fd test(remote): a one-device listing through the lazy surface stays one-device (RED)
-
8754256 fix(thread): thread_list degrades to local threads when this device's own id read fails or is empty (#13610)
-
d4b7e78 fix(memory-sync): rewind an unsent first capture that meets the space's copy
-
61ef113 fix(computer): resolve xterm allowSendEvents per Xrm for the target client, accepting only when every matching entry is true (review L1)
-
f571da3 docs(rpc): --check fails on a broken pair-test row; a pair test must use its method as a record type (#13484 review L1, L2)
-
1c9fa67 Merge remote-tracking branch 'origin/dev' into feat/remote-approvals
-
32ef161 test(thread): thread_list when this device's own id read fails or is empty (#13394 review L1, L2; RED)
-
b572386 Merge origin/dev into fix/thread-list-merged-devices
-
cd8c1ca test(computer): xterm allowSendEvents must follow Xrm for the target client (review L1)
-
2ea5aa7 fix(desktop): a target's own descendant never covers it in the occlusion check (#13296)
-
cd01ad2 Merge branch 'dev' into docs/13478-rpc-relay-parity
-
69361aa Merge remote-tracking branch 'origin/dev' into fix/thread-set-model-session-scope
-
307bafa test(service-auth): granted closed-device logins skip the post-login remote setup (a real device service was installed into the test HOME and recreated it after removal on macOS CI)
-
fc47a5f test(desktop): compare the live hit with the target's own reported role
-
edbc640 fix(memory-core): treat ENOENT on the fenced unlink of a dead owner's lock as reclaimed (#11644)
-
448a66c test(memory-core): a dead owner's lock that vanishes before the fenced unlink is still reclaimed (#11644)
-
32fe58b Merge dev 3975913 into feat/cu-v3-row14-fake-backend (take dev once)
-
18d623d test(computer): name each toolkit family by its library; changelog fragment for #13736
-
83acd27 fix(computer): read the last allowSendEvents setting from the back (clippy)
-
e4d6fbf fix(computer): X11 background input is sent only to clients known to accept synthetic events (#9109)
-
cf6b72e refactor(computer): return the modifier-chord refusal before resolving the focus window
-
ca7df9c fix(computer): refuse a Windows background chord that holds a modifier
-
fe9f1f9 fix(desktop): X11 elementAt descends to the deepest element under the point (#13296)
-
2d2871c test(desktop): X11 elementAt must name the deepest element under the point (#13296)
-
d7d15c1 test(computer): a background modifier chord on Windows must be refused before anything is posted
-
8f709e2 test(computer): X11 background input with an unknown receipt must not report delivered (#9109)
-
c047f62 Merge remote-tracking branch 'origin/dev' into cf-13137
-
46bda80 Merge remote-tracking branch 'origin/dev' into cf-13135
-
a7472f4 fix(memory-sync): a device joining a space adopts its history-less files insert-only
-
421b8f4 style(omo-native): format the new host runtime id tests
-
08fdc0f Merge remote-tracking branch 'origin/dev' into fix/13700-lows
-
3975913 Merge pull request #13700 from code-yeongyu/fix/7604-compiled-host-launch-profile
-
13bd127 fix(omo-native): a spec that lists no extensions keeps its host's id; cover every host command the wrapper gives a spec (#13700 review LOWs)
-
6f90994 fix(memory-sync): every unnamed remote-less memory syncs through one account default space
-
07cca16 fix(service-auth): a test omo login never installs or links a device service on the suite's host (cli-roundtrip leaked a real launchd service on macOS CI); regenerate service-auth.generated.js
-
52fdce2 Merge dev into cu-v2/task-35-p13b-latency-bench-and-cu-eval-7-tasks-3
-
6047c76 fix(remote): skip the post-login device setup on Windows for 5.1.73
-
5a921fe chore(knip): export-check the trust suites' private-TMPDIR helper (#13698 CI)
-
5ce3752 test(desktop-service): expose the fake engine and harness through a ./testing export (Fixes #13614)
-
818f408 test(omo-native): guard the launch-spec argument instead of asserting it
-
5359da0 test(omo-native): the private-TMPDIR helper judges ancestors with the product's namespace view (#13698 review L1)
-
095bf45 test(remote): the opt-out record's 0600 mode check applies on POSIX only (Windows has no mode bits; #13602 on Windows)
-
951c683 chore(changes): name the fragment after its pull request (#13700)
-
f2cde59 fix(omo-native): a host the compiled binary starts reports the CLI's runtime id
-
a44c5bd docs(changes): trust suites run with a private TMPDIR (#13698)
-
6eb4bfa test(omo-native): trust suites run with a private TMPDIR (#13697)
-
e61752b Merge pull request #13644 from code-yeongyu/fix/13638-remote-account-error-lines
-
22a629e chore(release): regenerate bundles at the omo-5173 train top (dev d84b9c2 + budget + quiet suppressed install)
-
9d2a7c8 fix(remote): a suppressed device-service install leaves omo login quiet (#13602 follow-up)
-
f79a7c3 Merge remote-tracking branch 'origin/dev' into chore/omo-5173-train
-
0af4ef3 test(memory-sync): take the rerun before any other await in the #11778 case (review LOW)
-
d84b9c2 Merge pull request #13630 from code-yeongyu/release/v5.1.72-source-state
-
c342c71 train(omo-5173): omo.js budget -> 1,489,433 (train-top regen measures 1,485,337 B; measured + 4 KiB, lead rule)
-
9a492ad docs(changelog): record remote session ID contract fix (#13666)
-
33783d3 fix(remote): accept durable session IDs across CLI verbs (#13010)
-
45a05f5 train(omo-5173): sort the A3 knip rows, drop the now-used FIXTURES row, fix the stale ignoreFiles entry and row comments (review L1-L3), from t3fn
-
d898650 train(omo-5173): knip rows for the A3 phrase stack (fixture as entry; new barrel/exports listed), from t3fn (#13336 owner)
-
5e77f8c chore(release): regenerate bundles at the omo-5173 train top (22 members)
-
c65eb8e test(memory-sync): pin the rerun for a head advertised while a cycle runs (#11778)
-
3b81111 fix(omo-native): omo remote reports the account store's error lines on stderr (#13638)
-
a630ef2 test(omo-native): the remote account read reports the store's error lines (#13638)
-
e6c13ff fix(omo-native): the waiting launcher relays SIGUSR1; test private fds with the startup pipe (#13632 review L1, L2)
-
849e2fd fix(omo-native): omo remote wake's usage gives the device service's real default (on)
-
d722ec2 ci(desktop): print the missing-base error when no release tag resolves
-
8a7f802 train(omo-5173): move #13334's changes-log block after the CHANGELOG text and drop its stale regeneration sentence (review H1, L1)
-
a1e1fda train(omo-5173): drop #12779's lines from the fold (withdrawn per Q, no QuickJS computer-run)
-
5b3d289 docs(changes): trust re-exec keeps the startup pipe (#13632)
-
d5da11a fix(omo-native): a trust re-exec of a coordinated engine start keeps its startup pipe (#13622)
-
def19ac Revert "Merge remote-tracking branch 'origin/m-12779' into chore/omo-5173-train"
-
edfb085 ci(desktop): run the bench nightly and on dispatch, not on every pull request
-
9b4c309 train(omo-5173): fold the member fragments into 13627.md; drop the stale regeneration sentence from #13609's line (review L2, #13618)
-
b2cc84d Merge remote-tracking branch 'origin/dev' into chore/omo-5173-train
-
84464fc Merge remote-tracking branch 'origin/m-12807' into chore/omo-5173-train
-
6abb083 Merge remote-tracking branch 'origin/m-12823' into chore/omo-5173-train
-
4cbe4fe Merge remote-tracking branch 'origin/m-12849' into chore/omo-5173-train
-
8490d35 Merge remote-tracking branch 'origin/m-12779' into chore/omo-5173-train
-
42022b1 Merge remote-tracking branch 'origin/m-10519' into chore/omo-5173-train
-
ad1f281 Merge remote-tracking branch 'origin/m-12450' into chore/omo-5173-train
-
5a07569 Merge remote-tracking branch 'origin/m-12339' into chore/omo-5173-train
-
90e3194 Merge remote-tracking branch 'origin/m-12332' into chore/omo-5173-train
-
60a6fc3 Merge remote-tracking branch 'origin/m-12337' into chore/omo-5173-train
-
641a1db Merge remote-tracking branch 'origin/m-12328' into chore/omo-5173-train
-
e10e54f Merge remote-tracking branch 'origin/m-12329' into chore/omo-5173-train
-
e336118 Merge remote-tracking branch 'origin/m-13101' into chore/omo-5173-train
-
8e2ab5f Merge remote-tracking branch 'origin/m-13334' into chore/omo-5173-train
-
3e987f4 Merge remote-tracking branch 'origin/m-13493' into chore/omo-5173-train
-
e23d79e Merge remote-tracking branch 'origin/m-13336' into chore/omo-5173-train
-
3e49d16 Merge remote-tracking branch 'origin/m-13335' into chore/omo-5173-train
-
a517f21 Merge remote-tracking branch 'origin/m-13330' into chore/omo-5173-train
-
ea96476 Merge remote-tracking branch 'origin/m-13481' into chore/omo-5173-train
-
3be8e7d Merge remote-tracking branch 'origin/m-13482' into chore/omo-5173-train
-
3382e5a Merge remote-tracking branch 'origin/m-12750' into chore/omo-5173-train
-
945abea Merge remote-tracking branch 'origin/m-13609' into chore/omo-5173-train
-
38f318c Merge remote-tracking branch 'origin/m-13602' into chore/omo-5173-train
-
1ecf612 Merge remote-tracking branch 'origin/m-13621' into chore/omo-5173-train
-
c52e5ee fix(memory-sync): wait for the memory repo's first commit and stop re-reading the account while waiting (#13486)
-
cac5ce1 Merge dev facd9fa into cu-v2/task-33-p12b-resource-caps-and-idle-flush
-
c57d249 Merge dev facd9fa into cu-v2/task-34-macos-live-qa-scenarios-chrome-two-windo
-
468351f Merge dev facd9fa into fix/docs-chromium-background-input
-
e1c1a29 Merge dev facd9fa into cu-v2/task-32-p5b-ii-computer-assert-host-bridge-insid
-
12989c3 Merge dev facd9fa into cu-v2/task-35-p13b-latency-bench-and-cu-eval-7-tasks-3
-
d41e5bc style: bun fmt the O14 files
-
c7b9685 Merge dev facd9fa into feat/o1d-computer-use-per-call-screenshot
-
3fb668f Merge remote-tracking branch 'origin/dev' into fix/13098-memory-sync-provider-reload
-
4cf8cba test(memory): load account-runtime-loader generations through one typed helper
-
f19272f Merge dev facd9fa into feat/o14-loopless-composition
-
742ee97 Merge feat/a3-phrase-enable (restacked on dev facd9fa) into feat/a3-phrase-recovery
-
28d748f Merge feat/a3-phrase-parse-derive (7e801b9, dev facd9fa) into feat/a3-phrase-enable
-
eb79fd0 style: bun fmt the O1a files after merging dev
-
3560a10 Merge dev facd9fa into feat/o1a-computer-use-scope-revoke
-
d921f31 style: bun fmt the O12 files after merging dev
-
a53fae5 style: bun fmt the O1c files after merging dev
-
7af6065 style: bun fmt the O1b files after merging dev
-
033e19e Merge remote-tracking branch 'origin/dev' into fix/13325-rollover-lows
-
ed7c5a6 style(memory-sync): format the device-session resolver (#13486)
-
7e801b9 Merge origin/dev (facd9fa) into feat/a3-phrase-parse-derive
-
1ab85ef Merge origin/dev into fix/memory-sync-first-session-repo
-
7f36912 Merge dev facd9fa into feat/o12-computer-use-hand-permission-report
-
8802e8d Merge dev facd9fa into feat/o1c-computer-use-foreground-lease
-
a370552 Merge dev facd9fa into feat/o1b-computer-use-device-exec-policy
-
e1751c3 Merge remote-tracking branch 'origin/dev' into feat/remote-approvals
-
2509616 Merge remote-tracking branch 'origin/dev' into fix/13604-review-lows
-
6d337d4 fix(remote): disable's unsaved-choice notice says remote is off only when it is (#13594)
-
ea342c7 fix(remote): disable survives an opt-out it cannot save; a user-linked ~/.omo is honored (#13594)
-
8cefe72 fix(remote): write the opt-out record without following symlinks (#13594)
-
9698c0c fix(remote): omo login enforces a recorded opt-out instead of only printing it (#13594)
-
7fb3e20 fix(remote): disable records the opt-out even when the device service cannot be reached (#13594)
-
0059493 feat(remote): remote sessions on by default; omo login switches them on and says so (#13594)
-
c1eb98c fix(device-service,changes): the #13599 review LOWs (#13604)
-
336d2ba docs(remote): state the true scope of the no-id guarantee (#13481 review LOW)
-
9ce5c61 fix(remote): a waiting send never takes a prompt stored before it was sent as its own (#13481 delta)
-
81e33dc fix(remote): an id-carrying send still prints a turn an engine without ids finished before the aim (#13481 review)
-
b0b9a91 fix(remote): a send with an id owns only its own prompt, and an interrupted superseded send keeps its answer (#13481 review)
-
26bf15f Merge dev into feat/remote-approvals, and say what the dialog note really does
-
3105102 test(attach): cover grantSelect and pendingQuestions through get_state (#13482 review LOW)
-
93ac3e6 fix(attach): keep raiser-only frames out of the state the TUI reads (#13482 review)
-
e948afe fix(remote): an earlier run's settle never ends this send's wait (#13481 review)
-
eb3ae5a fix(attach): present the snapshot's pending dialogs and questions as mirror records (#13482 review)
-
98cfbc9 fix(attach): show a dialog already pending when an attach starts (#13482)
-
af079a7 fix(remote): a waiting send no longer prints the turn's answer twice (#13481)
-
ec841f3 docs(changes): fragment for the device-service-core sent-request-type checks (#13491)
-
01bbe97 test(omo-native): prove the agent is still alive; scope the dispatch comment to this route (#13449)
-
dad35e2 test(device-service-core): compare the sent-type scan against the list with the wider type received (#13487)
-
b52aaca fix(memory-sync): defer the first session's sync until the memory repository exists (#13486)
-
709748e docs(changes): fragment for scoped thread set_model (#9999999)
-
c86ee46 fix(thread): scope set_model to the session and reasoning to session-only
-
ce1c218 test(device-service-core): LA21 reply map exhaustive over the request types omo sends; unserved list drop-out check
-
76671b2 docs(rpc): check curated pair tests and the rows file shape; mark local as not measured (#13484 review)
-
2fd3986 fix(thread): one device.info read per thread_list; test titles name an absent field (#13394 review LOWs)
-
6c9bb0f docs(rpc): generate a local-versus-relay parity table from the engine method table (#13478)
-
142bee6 test(remote): a dialog raised before the attach or during a relay gap is shown (RED, #13474)
-
d8f211d fix(remote): an earlier run's stop reason does not mark this send's run (#13272 review)
-
8f9a362 device-service-core: re-pin local-api to omo-relay-daemon main 4ed58920
-
7d3686a fix(service-auth): one closed-help map; user-facing fragment wording (#13466)
-
952e4b6 Merge origin/dev into feat/a3-phrase-parse-derive
-
234b584 test(device-service-core): LA21 matches the vendored client's defined reply fields
-
52e5525 docs(changes): fragment for the vendored local-api client and guard G1 (#13475)
-
0e9c09f device-service-core: re-pin local-api to relay-daemon main + #1130; add remote import guard G1
-
0d065ee test(omo-native): run the refuse case on Linux and macOS (#13449)
-
c67256c docs(changes): fragment for the closed logout/whoami --help fix (#13466)
-
aeca522 fix(service-auth): logout and whoami answer --help while accounts are closed (#13461)
-
2ffe162 Merge dev 83373dc into feat/cu-v3-row14-fake-backend
-
58d2f5c docs(changes): fragment for the compiled kill-guard dispatch fix (#13460)
-
3179eab fix(omo-native): answer the kill guard's __kill-guard check before the compiled boot (#13449)
-
3ba81e7 refactor(device-service-core): vendor the relay-daemon local-api client byte for byte
-
e38406b fix(thread): fold #13394 review LOWs; list this device in devices[]
-
0cabdf1 fix(thread): merge other devices into every thread_list, report ended rows resumable, name devices and unanswered ones
-
726cab5 chore(knip): drop the stale pre-existing-unused row for derive.ts#x25519PublicKey
-
901fc94 fix(script): --residue names every removal path it cannot check, beside its other findings, and fails closed (review L1 of #13135)
-
453941a test(script): RED - --residue reports every unreadable removal path beside its other findings, by its own classification (review L1 and L2 of #13135)
-
5691c83 test(omo-opencode): drop hooks/shared prompt-async-gate replays of the utils tests
-
711960b Merge commit '83373dcbb545ae2336e58abb46af028e9ad29a9e' into feat/remote-approvals
-
13b1bd0 Merge branch 'a3-pr2a' into a3-pr2b
-
638868a fix(memory-core): fold the #13330 checker (format salt-inputs, async nonce refusal, zero keys on a failed derive)
-
eef7bbd Merge branch 'a3-pr2a' into a3-pr2b
-
30a426a Merge branch 'a3-pr1' into a3-pr2a
-
7f3471c Merge branch 'a3-pr2a' into a3-pr2b
-
a0a21f1 chore(memory-core): re-pin the vendored phrase vectors to omo-service main after #1068 merged
-
a3cb9cd Merge branch 'a3-pr2a' into a3-pr2b
-
e52d8e7 test(memory-core): pin that an undelivered void notice stops enrollment before the client flag is read
-
a5e0aba fix(memory-core): fold checker findings on #13336 (unverifiable acknowledge, fmt, lint)
-
96d8660 style(script): bun fmt the computer-v3 guards test (fmt:check on #13135)
-
080df28 Merge dev into fix/cu-v3-13118-contract-hardening (take dev, merge only)
-
8ed322c Merge dev into fix/13099-guards-lstat-parent (take dev, merge only)
-
dbf96de fix(computer): the fake's source guard reads grouped std imports across lines; the capture comment names which refusal halves are load-bearing
-
856b732 test(computer): the fake's source guard must catch a grouped env import split across lines
-
781dd7d Merge remote-tracking branch 'origin/dev' into omo-13120-take-dev
-
4b4b6eb fix(memory-core): fold LOWs on the service-refusal switch path (#13335 @ 9b1f3c9, #13337 @ 6e72603)
-
cb93057 Merge branch 'a3-pr2a' into a3-pr2b
-
b9c13ac Merge branch 'a3-pr2a' into a3-pr2b
-
9b1f3c9 test(memory-core): drive the admission-refusal row through the phrase bridge
-
4c88635 fix(memory-core): fold #13335 review LOWs (account switch only on an empty chain, admission refusal, test await)
-
91cabff Merge branch 'a3-pr2a' into a3-pr2b
-
9f9734a fix(memory-core): fold #13336 review (acknowledgeable enroll/rotate void, rotation id check, abortable bridge)
-
0125449 Merge branch 'a3-pr2a' into a3-pr2b
-
75f47e1 Merge commit 'b4f98dadf7863e6ac62cbff028c7365ec3845baa' into a3-13336
-
b4f98da fix(memory-core): read the service floor from accepts_phrase_keys (omo-service #1070 rename)
-
7aff562 Merge commit '364bbf7cced5d1b6494647b138158acf761a7a9d' into a3-pr2a
-
364bbf7 refactor(memory-core): fold #13330 review LOWs (single rp_id, shared reserved prefix)
-
8027ad8 Merge remote-tracking branch 'origin/fix/13325-refusal-reads-signed-out' into fix/13325-rollover-lows
-
663531d Merge remote-tracking branch 'origin/fix/13325-account-rollover-live-channel' into fix/13325-rollover-lows
-
9e9325b feat(memory-core): phrase recovery, enrollment, rotation and removal (A3 part 2b)
-
3441e5d docs(changes): move the #13334 package-log entry into its fragment's changes-log block
-
1ae227f style(memory-sync): bun fmt the rollover test (#13327 review LOWs)
-
71c0338 Merge remote-tracking branch 'origin/feat/a3-phrase-parse-derive' into feat/memsync-phrase-genesis-core
-
6702c4c test(memory-core): leak canaries match whole words; old-member refusal read on the sync chain (A3 part 2a)
-
b3b06aa fix(memory-sync): a replaced desktop account ends; a Worker begin that crosses a retirement owns nothing (#13327 review LOWs)
-
75fa8dc test(memory-sync): RED - refused describe is null on both threads, a replaced account ends, a begin crossing retirement owns nothing (#13327 review LOWs)
-
ad2ee67 Merge remote-tracking branch 'origin/fix/13325-refusal-reads-signed-out' into fix/13325-rollover-lows
-
08ec33a test(memory-core): use the canonical PHRASE_RECOVERY_RP_ID in phrase test support (knip)
-
e4149fc feat(memory-core): recovery-phrase genesis in enable, with a durable void marker (A3 part 2a)
-
5e64862 feat(memory-core): parse, derive and redact recovery-phrase keys (A3 part 1)
-
2cf7e2d test(remote): an answer to a never-shown or settled dialog carries no uiRequestId (security review of #12750)
-
bbafd82 Merge remote-tracking branch 'origin/dev' into t5-13101
-
51a4e1b Merge remote-tracking branch 'origin/dev' into feat/remote-approvals
-
7edae09 Merge dev into ci/computer-live-macos-tier (take-dev, 97 behind; knip.jsonc keeps both sibling entry lines)
-
20c8781 docs(computer): record the guest-label jobs of the macOS live tier as a drop for now
-
88db1cb ci(computer): owned macOS guest pool for the live tier (part 1)
-
58a7df3 Merge remote-tracking branch 'origin/dev' into feat/remote-approvals
-
4f85a44 test(eq): re-record linux-x64 E21-E24 baseline after merging dev 8b2916e (row 33, #12807)
-
ad767f4 Revert "test(eq): re-record linux-x64 E21-E24 after merging dev 8b2916e (skill digest)"
-
de4952a style: apply bun fmt to files the dev merge left unformatted
-
71590d0 style: apply bun fmt to files the dev merge left unformatted
-
3f88fc3 row 38: merge base #12849 (fmt fold) so the stack carries its head
-
31eb4a8 style: apply bun fmt to files the dev merge left unformatted
-
99fd196 style: apply bun fmt to files the dev merge left unformatted
-
6843059 fix(computer): SP-T every exit value comes from EXIT, checked structurally; README documents --measure/--row exit codes
-
8c790ee test(computer): SP-T exit-code mutants for every form and README --measure/--row exit codes (RED)
-
e44cf6d fix(computer): SP-T tap lines must be JSON objects; README documents the --cleanup exit codes, checked against the code
-
4378bf0 test(computer): SP-T class-level self-test for non-object tap lines and README exit codes (RED)
-
a3647cf test(computer): SP-B full --write regenerates; bunshinSrc and helper timeout hardened
-
bdb3103 fix(computer): SP-T reads a tap file only when complete, and --cleanup exits 1 on a failed receipt
-
1759ea2 test(computer): SP-T class-level self-test for incomplete tap files and the --cleanup exit code (RED)
-
b19c743 test(computer): SP-B probe proves findings only from returned values
-
6cc55f9 fix(computer): SP-T unread tap files, reportless grant checks and failed cleanup reads are never negatives
-
b5735ec test(computer): SP-T self-test cases for unread tap files, reportless grant checks and cleanup that stops early (RED)
-
bee8023 fix(computer): SP-T readers never turn a failed command into a negative (tccLog, tccDb, pgrep, helpers)
-
f9152d7 test(computer): SP-T self-test cases for swallowed tccLog, tccDb and pgrep errors (RED)
-
c7ef529 test(computer): SP-B probe never turns an error into a verdict
-
4b38bd5 fix(computer): SP-T delta-review fixes - prompt-probe conditions, separate verify evidence, checked tccutil resets, route before person, QA-owned results dir, no eval in the GUI leg, bounded prompt poll and --self-test
-
340ce84 test(computer): SP-B probe measures each host-death watch alone and keeps verdicts in the repo
-
a7cbf72 test(computer): sweep unknown/unknown receipts; fragment states the full rule (#13147)
-
a16f1f8 row 38: merge base #12849 (car 35) - keep its restore/teardown fix and run.mjs split with row 38's additions
-
18fdac8 chore(changelog): align the #12849 fragment with its base branch (stack gate)
-
be5ade7 test(computer): run the SP-B probe as a knip entry
-
2d1295a fix(computer): refuse a receipt under unknown/unknown (LEAD ruling R13, #13147)
-
6187dd9 test(computer): a receipt under unknown/unknown is refused (RED, LEAD ruling R13, #13147)
-
93d305a test(computer): fold the #13137 delta-review nits; pin unknown/unknown receipts pending #13147
-
ff923f8 fix(computer): SP-T review fixes - person-gate and by-path launch of the tap, keep the T3 keychain off the search list, reset prompt rows in finally
-
7ddad08 docs(threat-model): the worker-crash run error reason is crashed, matching the code (pre-flight CI finding)
-
2fb19d1 fix(computer): capture refuses a non-positive scale or negative size before multiplying; the source guard catches grouped env imports
-
21ab5ba test(computer): RED - sub-pixel and negative-scale capture extents, grouped env imports in the source guard
-
cc95f96 test(computer): SP-T ledger rows await a macOS 26 host with a QA account
-
a3df36a Merge pull request #13133 from code-yeongyu/fix/cu-v3-13121-fake-followups
-
ce12ee0 fix(memory): name the in-place update and restart remedy when an older omo build holds the provider (#13101 review LOW)
-
bf434f1 Merge origin/dev into fix/12777-remote-view-double-delta
-
a48ab68 test(computer): SP-T rows that need a person read WAITING-PERSON
-
bb846f3 test(computer): probe ledger for SP-T
-
474a2fe test(script): an unreadable removal path (ELOOP) is reported as a finding naming it (#13135 review LOW)
-
07fe926 Merge remote-tracking branch 'origin/dev' into feat/cu-v3-row14-fake-backend
-
c2b8990 test(computer): probe ledger for SP-B
-
9ff759e test(computer): landed-unit receipts stay under an uncertain partial report (LEAD ruling 2026-11-11, #13118)
-
cfa1027 docs(changes): #13133 fragment covers the empty-extent refusal
-
e09b1c0 test(computer): move the two new fake tests out of the ported-test module
-
5bbc9b3 fix(computer): fake capture refuses a zero or negative extent, like a non-finite one
-
894fccc test(computer): fake capture must refuse a zero or negative extent
-
17532e7 fix(script): --residue treats a removal path under a file parent as absent (#13099 follow-up)
-
a9d7bce fix(computer): type Refusal.reason with the closed Reason vocabulary (#13118)
-
a141eb5 Merge remote-tracking branch 'origin/feat/cu-v3-row14-fake-backend' into fix/cu-v3-13121-fake-followups
-
ceea65b test(computer): a refusal reason outside the closed vocabulary does not compile (RED, #13118)
-
f6cfcd0 fix(computer): refuse a receipt under not_delivered or delivered_unverified (#13118)
-
0848acd fix(computer): the fake's source guard bans environment reads
-
f0e0c4a test(computer): the fake's source guard must catch environment reads
-
8cef02f fix(computer): fake inject() refuses a fail pair with no error-table row, as the loader does
-
13a3b37 test(computer): a receipt under not_delivered or delivered_unverified is refused (RED, #13118)
-
f0fb135 test(computer): inject must refuse a fail pair with no error-table row
-
832f58d row 34: merge dev 04e22c0 (post bun-fmt; car sources re-formatted)
-
10e2c94 row 33: merge dev 04e22c0 (post bun-fmt; car sources re-formatted)
-
a8bc42e test(desktop-service): the bridged call-cap case is ended by the cap, not the 5 s run clock
-
55b9054 row 32: merge dev 04e22c0 (post bun-fmt; car sources re-formatted)
-
f6a7f12 Merge remote-tracking branch 'origin/dev' into feat/remote-approvals
-
fb4689a fix(computer): refuse a non-finite fake capture and keep focus-record ids
-
e183279 fix(ci): keep the attach test rig's privateDir unexported
-
8646008 Merge remote-tracking branch 'origin/dev' into feat/remote-approvals
-
81daf81 fix(ci): export-check the attach dialog test rig and write the native-contract job summary
-
614b640 test(computer): v2 values cover identity 101 and the full effects list; focus-read recording superseded by AC-70
-
a577202 test(computer): the oversized-capture test reports only the frame size
-
3bc219c fix(computer): fake capture refuses an over-cap composite before allocating
-
6a2dfbb test(computer): ported parity lines cite tests that fail for the old behaviour
-
6f66e5d docs(changes): add the row 14 fragment
-
868372c Merge remote-tracking branch 'origin/dev' into feat/cu-v3-row14-fake-backend
-
51dc257 test(eq): re-record linux-x64 E21-E24 after merging dev 8b2916e (skill digest)
-
308944e Merge remote-tracking branch 'origin/dev' into feat/cu-v3-row14-fake-backend
-
fb767bb refactor(computer): fake uses the core Refusal::bare and row(); drop the bare-reason workaround
-
b7ae338 Merge branch 'feat/cu-v3-row13-core-contract' into feat/cu-v3-row14-fake-backend
-
832b596 feat(computer): fake refusals through the closed EngineCode table; unanswered AX writes are AxUnconfirmed
-
0aa2360 Merge remote-tracking branch 'origin/dev' into fix/10519-assets
-
607b336 Merge branch 'feat/cu-v3-row13-core-contract' into feat/cu-v3-row14-fake-backend
-
65ef2b3 Merge branch 'feat/cu-v3-row12-core-protocol' into feat/cu-v3-row14-fake-backend
-
6fdbd8c chore(omo-native): regenerate the service-auth bundle for the attach dialog mirror
-
3ee57c1 test(computer): port the old fake-backend tests in the parity manifest; pin reference refusals
-
628d018 Merge branch 'feat/cu-v3-row12-core-protocol' into feat/cu-v3-row14-fake-backend
-
c020493 test(eq): re-record linux-x64 E21-E24 after merging dev 89deb53 (skill digest)
-
39f0169 docs(changes): note remote dialogs for every attach (#12750)
-
c4f1cb9 feat(remote): omo remote attach shows a dialog another client raised and answers it by uiRequestId (#12749)
-
c58907b test(remote): a viewer that did not raise a dialog sees and answers it in omo remote attach (RED)
-
af80df8 docs(changelog): describe only the duplicate-delta fix in #13075 fragment
-
a1658f2 fix(computer): fake scenario root must be an object; owe conformance_replay to row 28
-
71366a0 chore(changelog): drop the #12737 fragment already shipped in v5.1.69 (train F)
-
ebb6bc9 docs(changes): note the memory hot-reload registration fix (#13101)
-
a393fd9 fix(memory): keep the memory-sync provider guard process-wide across hot reloads (#13098)
-
4647820 chore(computer): drop the re-deferred C1 leg line (row 12 owns it)
-
558ee04 feat(computer): v3 scriptable fake backend
-
a49f693 test(computer): RED - v3 fake backend tests (signature stubs only)
-
aca0e27 fix(cu-eval): restore and tear down live tasks on every exit path (#12849 review)
-
9070368 fix(desktop-service): drop trailing undefined arguments at the sandbox boundary so quickjs matches vm (#12779 review)
-
60262c8 fix(remote): keep the CLI's shared device-service connection open across open_session
-
c550354 fix(thread): send remote open_session on a short-lived connection (#12777)
-
10afe96 chore(changes): drop 4 fragments the v5.1.66 release already consumed on dev
-
51ca5a1 row 35: align 4 paths with the reviewed head 1e4cf94e56 (generated + goldens + fragment)
-
686a4e2 row 35: regenerate bundles on the genuine dev base
-
e8efe10 test(eq): re-record linux-x64 E21-E24 after merging dev a96a51d (skill digest)
-
97a4faf fix(cu-eval): prove realStateUntouched with a sha256 digest, not a length (row 35)
-
f9b30a0 fix(cu-eval): observe the oracle state before the teardown restores the surface (row 35)
-
b8eb516 fix(qa): route the cu-eval launcher and engine spawns through withTelemetryOptOut
-
3ff545d docs(changes): fragment for plan row 35 (placeholder number; renamed to the PR number once it exists)
-
9833415 test(eq): re-record linux-x64 E21-E24 after merging dev a96a51d (skill digest)
-
1dbaf6c Merge remote-tracking branch 'origin/dev' into HEAD
-
4da9324 row 38: merge train E 7461ea5 (GNOME extension move, placement fix, fixture placeholders)
-
d8226b5 row 33: regen generated bundles after train E merge
-
95cb265 row 33: merge train E 7461ea5 (GNOME extension move, placement fix, fixture placeholders)
-
4716a95 row 32: merge train E 7461ea5 (GNOME extension move, placement fix, fixture placeholders)
-
62a570e row 34: merge train E 7461ea5 (GNOME extension move, placement fix, fixture placeholders)
-
6de9a53 row 35: merge train E 7461ea5 (GNOME extension move, placement fix, fixture placeholders)
-
0b40660 chore(regen): rebuild generated bundles for the row-32 review fold
-
e337d08 test(desktop-service): cover the remaining row-32 bridge cases (row 32)
-
87e2400 fix(desktop-service): bridged errors carry code, engineRestarted and method into the quickjs body (row 32)
-
c48e801 docs(sandbox): threat model names resultSize and the per-run handle table (row 32)
-
bcfdf3e fix(desktop-service): a bridged call that returns undefined reaches the quickjs body as undefined (row 32)
-
a7b3c26 fix(desktop-session): exempt v1 sessions from the frame cap, as the idle flush does (row 33)
-
66e50eb fix(desktop-engine): answer request_too_large with the request's id so clients see it (row 33)
-
818e86d fix(cu-eval): prove realStateUntouched with a sha256 digest, not a length (row 35)
-
f27a4e6 fix(cu-eval): observe the oracle state before the teardown restores the surface (row 35)
-
f353203 docs(changes): add the row 38 fragment (new-PR placeholder)
-
efc076c ci(desktop): computer-use bench regression workflow (path-gated, base vs head)
-
30e38b3 docs(changes): fragment for P12b resource caps and idle flush
-
64ea701 feat(desktop-engine): request, frame and ref caps with idle flush
-
7d180ea chore(regen): rebuild generated bundles on train E for P5b-ii
-
3749333 docs(changes): fragment for P5b-ii computer/assert host bridge
-
aae2e9c fix(desktop-service): SANDBOX_MAX_HANDLES stays module-private
-
02b378c feat(desktop-service): bridge computer and assert into the QuickJS run cell
-
7e43522 docs(changes): name the row 35 fragment with the new-PR placeholder
-
40fb3e8 fix(qa): declare the desktop sdk beside its sibling desktop packages (row 34)
-
1d5bc73 Merge remote-tracking branch 'origin/dev' 8af3fae into the row 35 stack
-
a50a256 Merge remote-tracking branch 'origin/dev' 8af3fae into the row 34 stack
-
b99fe53 Merge remote-tracking branch 'origin/dev' 1954eab into the row 34 stack
-
a0b9783 fix(qa): fold review nits on the macOS live scenarios (row 34)
-
29157a5 Merge remote-tracking branch 'origin/dev' into cu-v2/task-35-p13b-latency-bench-and-cu-eval-7-tasks-3
-
9801654 chore(knip): drop the computer-control-live-macos-scenarios rows now used by this car
-
712d113 Merge remote-tracking branch 'origin/dev' into cu-v2/task-34-macos-live-qa-scenarios-chrome-two-windo
-
bea1bb7 Merge remote-tracking branch 'origin/dev' into cu-v2/task-34-macos-live-qa-scenarios-chrome-two-windo
-
bd69cae docs(computer-use): correct X11 refusal signature and macOS right-click coercion
-
49f6b88 test(eq): re-record darwin-arm64 E21-E24 goldens for the moved skill digest
-
95371e8 fix(desktop-macos): bound and classify priming's focus reads, log layer 1
-
974ba0e test(eq): re-record the EQ.omo linux goldens for the computer-use skill hash
-
318f73e Merge remote-tracking branch 'origin/dev' into cu-v2/task-16-p7b-exact-window-key-priming-steps-1-6-8
-
797769c fix(qa): route the cu-eval launcher and engine spawns through withTelemetryOptOut
-
0abc2cf fix(qa): route the capture-consent probe and the off-Space scenario env through withTelemetryOptOut
-
40af05d fix(qa): route the capture-consent probe and the off-Space scenario env through withTelemetryOptOut
-
83cc52e fix(qa): route the capture-consent probe and the off-Space scenario env through withTelemetryOptOut
-
0485d70 docs(changes): fragment for plan row 35 (placeholder number; renamed to the PR number once it exists)
-
fb14208 docs(changes): fragment for plan row 34 (placeholder number; renamed to the PR number once it exists)
-
7d0dcfa docs(changes): fragment for plan row 16 (placeholder number; renamed to the PR number once it exists)
-
5945057 docs(computer-use): add Windows/X11 platform scope, note macOS right-click refusal, and regenerate assets
-
0aee1df test(computer-use): update COMPUTER_DOCUMENTATION sha256 fixture for #10410
-
be3d2a9 build(computer-use): sync staged plugin assets and remove duplicate shared skill
-
c846a31 docs(computer-use): document Chromium background input refusal (#10410)
-
c732f00 chore(bundle): regenerate O12 extension after the dev sync
-
eff2db9 dev sync: merge dev f2a520f into feat/o12-computer-use-hand-permission-report
-
0d0a023 chore(bundle): regenerate O12 extension after the #12598 dev sync
-
5e2fcd4 dev sync: merge dev efd4734 into feat/o12-computer-use-hand-permission-report
-
65ee581 chore(bundle): regenerate O12 extension after dev sync
-
061ea8d dev sync: merge dev b4e373c into feat/o12-computer-use-hand-permission-report
-
29576a3 Merge origin/dev into feat/o14-loopless-composition
-
8e85cfe Merge origin/dev into feat/o14-loopless-composition
-
9f92946 test(audit): allowlist the hand-browser-placement child-process fixture (#12329)
-
16c00f3 Merge origin/dev into feat/o14-loopless-composition
-
366b1e9 chore(bundle): regenerate O12 extension after dev sync
-
e8cbee7 Merge remote-tracking branch 'origin/dev' into feat/o12-computer-use-hand-permission-report
-
1fd4562 feat(computer-use): report OS-permission denials from a hand's execution scope (O12)
-
2dd0429 test(computer-use): OS-permission denials reported from a hand's execution scope (O12, RED)
-
91c3e2f test(desktop-service): fake engine screen marker that changes between captures (O1d, W3)
-
0253d5a test(computer-use): one screenshot per call survives the hand wire, under today's caps (O1d, RED)
-
7bf1f3a feat(computer-use): foreground lease keyed by (issuer, scope) for hand-placed sessions (O1c)
-
93b6d5f test(computer-use): one foreground lease between sessions on a device (O1c, RED)
-
816b21e feat(computer-use): device exec policy decides foreground control on a hand (O1b, C3(b))
-
9267973 test(computer-use): a hand's foreground control follows the device exec setting (O1b, RED)
-
5601b71 feat(computer-use): revoke and close through the holding session's lifecycle source (O1a)
-
323ffbd test(computer-use): grant revocation follows the holding session's lifecycle, hand scopes included (O1a, RED)
bun add -g omo-ai