#818 read a gitActor variant the control plane never shipped, so the line never showed.
- Source control and the people tooltip read githubActing / you from the workspace record
(cloudWorkspaces.get, and the team live list for the people chip); no feature header.
- Reasons: window_expired -> 'no one has sent in the last hour'; no_sender_yet -> 'whoever sends
next, or Orca bot'; sender_not_connected / reconnect -> 'Orca bot'; off / not_github -> nothing.
Connect GitHub when you.connected is false (bot cases only).
- Removed #818's gitActor schema, client method, IPC and the git-actor header (the server doesn't
read it).
Neil: hold share links this release, keep the code. Every way to produce a link is behind the
account capability flag `cloudShareLinks` (capabilities.flags, the existing server-driven pattern),
so it's off unless the control plane turns it on:
- sidebar right-click Copy › "Link to W-n" / "Workspace link": hidden;
- the cloud workspace list's ⋯ "Copy link to W-n" and "Copy link to this chat/terminal": hidden;
- ⌘⌥⇧L: does nothing (global and in the list row), and is gone from Settings › Shortcuts and its
search;
- copyCloudWorkspaceLink itself returns false with no clipboard write or toast, so the
private-link toast (#815) can't be reached from anywhere.
Unchanged: opening incoming orca://cloud links, and Share with team / Make private (access, not
the link).
A cloud workspace's git acts as its solo owner, else the latest sender within the hour, else Orca's
bot; the control plane decides (#817's gitActor) and the desktop only says it, quietly:
- Source control, under Commit/Push: "Pushes and PRs go out as @login" / "as Orca bot" / "as Orca
bot (no one has sent in the last hour)", with the approvals/merges caveat in its tooltip. The
only place with an action: Connect GitHub, when it's the bot and the viewer isn't connected.
- The people chip: the same words as a second line in the acting member's avatar tooltip (the
first avatar's when it's the bot). No new header chip.
- Read with feature `git-actor` from GET /v1/team/workspaces/:id (owner and members alike), one
read per workspace shared by both places; re-read on focus, every 2 min, and when a shared
workspace's hour lapses (`until`). Absent field (not GitHub, identities off, older control
plane): nothing shown.
Copy link (menu, ⌘⌥⇧L, the list) on a private cloud workspace used to say "Workspace link copied"
while a teammate opening it is refused. Now, still copied:
- private, you can share it (owner): "Link copied, but teammates can't open it yet" (warning
tone) with Share with team, the same visibility change as the menu's Share with team, then the
usual "Shared with your team" toast;
- private, someone else's: the same title, "Only its owner can share it with the team", no action;
- team-visible, or visibility unknown (no answer within the copy's lookup): unchanged, "Workspace
link copied". A chat link says "Chat link copied, but…".
Visibility comes from the workspace identity cache (now keeps it) or the copy's own lookup.
- The cloud workspace's W-number chip (CloudWorkspaceNumberChip) is gone from the workspace tab
bar and the titlebar strip; the component, its fixture and its two chip-only strings are
deleted. Copying the workspace link stays on the sidebar's right-click Copy › Workspace link and
⌘⌥⇧L (cloud-workspace-links.ts), neither of which used the chip.
- Delete-workspace dialog buttons: "Archive instead" -> "Archive", "Delete workspace" -> "Delete"
(same string in the cloud workspace list's confirm dialog).
stablyai/orca#42960 (Neil, dev) was authored by orca-cloud-dev[bot]. Neil had connected GitHub; member
identities only reached git: gh ran on the workspace placeholder (the App's installation token), and
the Worker's member route on api.github.com allowed only REST under the repo while gh pr create is
GraphQL.
- VM: a gh wrapper runs gh with the attested sender's own placeholder (orcad attests, as for git);
plain gh otherwise. Behind GH_AS_MEMBER=on (dev on).
- Worker, api.github.com with a member placeholder: the member's own token only for REST under the
workspace repo and for allowlisted authorship mutations whose every target resolves to it
(repositoryId equals its node id; other ids looked up with the installation token, cached per repo,
failures not cached). graphql-js parse; variable defaults, owner-level projectIds, duplicate
top-level keys and anything unexpected are not member requests. Everything else goes out on the
installation token as before; an out-of-repo pinned mutation is refused with a GraphQL error.
- Shared workspaces never approve or merge as a member (GraphQL and REST): those go out on the
installation token, so branch protection still applies.
- Identity rule: the only person who can use a workspace always acts; in a shared one the latest
sender acts until another member sends, for at most MEMBER_GIT_SHARED_MAX_MINUTES (default 480).
- A GitHub connect or disconnect mid-session reaches the running VM's identities now.
- Tokens never enter the VM.
GH-AS-MEMBER (#812): a connected member's PRs, comments and pushes from cloud workspaces now show as
them, not only commits. The settings row's connect and disconnect copy says so.
Once the GitHub access step settles (ok, granted or skipped), a person who hasn't connected their
own GitHub sees "Connect your GitHub so PRs show as you." with Connect GitHub and Not now, in the
step's own style. It never holds the create.
- Reads GET /v1/github/user through the existing Settings hook (useGitHubIdentity): focus re-read
and a light poll while the browser OAuth is open; "PRs from this workspace show as @login" once
connected here. Already connected, or no answer (signed out, older control plane): nothing shown.
- Not now is remembered on this device (it's about the person, not the repo); Settings › Orca
account still has Connect.
Neil (n75, dev): the browser pane in a cloud workspace showed 'Connection isn't secure'
(ERR_CERT_AUTHORITY_INVALID) on github.com. A paired runtime's browser route sends every connection
through the VM (SOCKS on the desktop, dialed by orcad in the sandbox). An Orca Cloud sandbox's egress
terminates TLS for hosts that carry credential rules (github.com always does: the App token and member
routes), so Chromium on the Mac got a certificate from the sandbox provider's CA. It must never trust
that CA, and a public site never needs the sandbox anyway.
For Orca Cloud environments the route now sends only the workspace's own loopback through the sandbox
(localhost, *.localhost, 127.0.0.0/8, ::1: its dev servers and forwarded ports); every other site is
dialed from this desktop exactly like a local workspace's browser (the native dialer, honouring the
desktop's proxy settings and destination policy). Other paired runtimes and SSH/WSL routes are
unchanged.
(A PAC script can't express this: Chromium never consults PAC for loopback, measured on this Electron.)
Picking "Run on: Orca Cloud" then another project reset Run on to this computer.
useEphemeralVmRecipeOptions reloads the project's recipes on every repo change with
load(resetSelection=true), which cleared the selection to null (Orca Cloud is the same for every
project, unlike a project's own orca.yaml recipes).
- A project change now carries Orca Cloud: it stays selected while the new project's options load
(no flash to Local) and after, whenever the new project offers it.
- Only when the new project can't use it does Run on fall back, and a line under Run on says why:
the project lives on another machine (SSH / remote server / not loadable here), Orca Cloud didn't
answer, or it isn't offered for this project. The next pick clears it.
- A project's own recipe still doesn't carry to another project.
Tests: 5 new cases in useEphemeralVmRecipeOptions.test.tsx; 4 of them fail on the old hook.
Puts the token-free compile on a probe account's LIVE session, so the probe can see whether a
header-matched respond rule answers at all (separating 'respond not enforced' from 'resume restored
something else').
egress-docs' A4 run: after resume-raw the echo to the custom base-URL host answered 200 where the
token-free record should answer the placeholder 403. The Worker's token-free compile is right (the
host keeps its respond rules: checked in the harness), so the probe needs to see what Vercel holds
per host and what status came back:
- the audit adds domains: {host: {injection, forward, response}} (hostnames only);
- echo-probe returns the HTTP status again. The echoed body stays in a shell variable, never a file.
A Cloud Run cold start cost Codex's first chatgpt.com call ~10.8 s on dev (seen twice on
2026-11-15, each right after an idle period). min_instance_count 0 -> 1 on staging (applied).