Repository navigation
refactor(i18n): restore the Crew vocabulary for remote crews - #12697
Conversation
Design Review (Fable 5) — 🟡 CONCERNSDesign-level review of Design-Verdict: CONCERNS "Crew" now names both the remote machine and the agent group; disambiguation hangs entirely on the open, unmerged #11869. Watch
[DESIGN-REVIEWED] 2c2dd36 |
GPT 5.6 Review — ✅ no blocking findingsGPT 5.6 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
UX Review (Fable 5) — 🟡 CONCERNSUX-level review of UX-Verdict: CONCERNS "Crew" now means two things — the agent team (shot-02, "your crew runs only on this computer") and a connected machine (shots 03–06) — and the blind reader couldn't connect them. Watch
Suggestions
[UX-REVIEWED] 2c2dd36 |
First Principles Review (Fable 5) — 🟡 CONCERNSPremise-level review of First-Principles-Verdict: CONCERNS The relabel's only ground is "restores the original vocabulary" — nobody is named as failed by "Remote Instances", and it reverses #6950's pinned rationale. Not justified as shipped
What this change shipsInventory (10 items) — 8 justifiedIntent: give the remote-gateway feature back its "Crew" name across UI, catalogs, docs and pins — a RELABEL (framed as a restore), not a fix or addition.
[FIRST-PRINCIPLES-REVIEWED] 2c2dd36 |
Opus 4.8 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
6ef9895 to
708cf0b
Compare
1cc4090 to
e2ed916
Compare
GPT 5.6 Review —
|
e2ed916 to
f1bc1ea
Compare
Design (Fable 5) — PASS, authorization itemDesign returned PASS with no item sections, so this item carries no finding span. Answering without one rather than inventing an identity.
|
Design (Fable 5) — PASS, follow-up-filing suggestionDesign's suggestion section carries no finding span; answering without one.
|
First Principles — Item 2, the re-authored naming note (span=6f66c85ed6e3)
|
UX — Watch, collision at first contact (span=76d8ef03a1f2)
|
The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750
PR #6950 renamed the remote-crew feature's user-visible text to "Remote Instances" and #7378 swept the docs to match. This restores the original "Crew" vocabulary across every catalog, the docs, and the tests and harnesses that pin it. Reverting #6950's key set alone would have been incomplete: nine strings adopted "Remote Instances" in the three weeks after it merged and were never in its key set. Four of those are cross-references that tell the reader to open a panel this change renames, so they move too. The word "instance" stays wherever it means a machine, an EC2 box, a gateway process, or a code identifier -- backend keys and routes, the `instances.json` registry, the Mochi multi-instance namespace, the gui_user `remote-instances` slug, and the filed Fargate RFC. CHANGELOG history is not rewritten.
8be8656 to
2c2dd36
Compare
The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750
The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750
The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750
The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750
The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750
) The Remote Crew feature's page kept its remote-instances filename after the #12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes #12750 Co-authored-by: Wael Dimassi <ydimawae@amazon.com>
…dev#12697) PR kirodotdev#6950 renamed the remote-crew feature's user-visible text to "Remote Instances" and kirodotdev#7378 swept the docs to match. This restores the original "Crew" vocabulary across every catalog, the docs, and the tests and harnesses that pin it. Reverting kirodotdev#6950's key set alone would have been incomplete: nine strings adopted "Remote Instances" in the three weeks after it merged and were never in its key set. Four of those are cross-references that tell the reader to open a panel this change renames, so they move too. The word "instance" stays wherever it means a machine, an EC2 box, a gateway process, or a code identifier -- backend keys and routes, the `instances.json` registry, the Mochi multi-instance namespace, the gui_user `remote-instances` slug, and the filed Fargate RFC. CHANGELOG history is not rewritten.
…odotdev#12752) The Remote Crew feature's page kept its remote-instances filename after the kirodotdev#12697 vocabulary pass because the filename is a published URL. Rename it to remote-crew.md and leave a stub at the old path so existing links keep resolving. There is no docs site generator in this repository — these pages ship inside the Python package and are linked externally as GitHub blob URLs — so no true redirect mechanism exists; the stub page is the fallback the issue sanctions for that case. Referring links in index.md and README.md now target remote-crew.md, and test_remote_instances_doc.py moves to test_remote_crew_doc.py with its pinned path updated. The test/gui_user scenario slug and backend 'instance' vocabulary are deliberately untouched. Fixes kirodotdev#12750 Co-authored-by: Wael Dimassi <ydimawae@amazon.com>
What changed
The remote-crew feature's user-visible text goes back to Crew. PR #6950 renamed it to "Remote Instances" and #7378 swept the docs to match; this restores the original vocabulary across all 14 i18n catalogs, the guides and system specs, the shipped user docs, and the tests and capture harnesses that pin the strings.
Reverting #6950's key set alone would have shipped a broken product. Nine strings adopted "Remote Instances" in the three weeks after it merged and were never in its key set, and four of those are cross-references: they tell the reader to open "Remote Instances in Settings", a panel this change renames. A user following that instruction would look for a panel that no longer exists. Those nine move too, and each locale's reference now names that locale's own panel label rather than a translation of the English one.
"Instance" stays wherever it means a machine, an EC2 box, a gateway process or a code identifier: backend keys and routes, the
instances.jsonregistry, theapps.mochi.instances.*namespace (internally consistent machine vocabulary — "every Kiro Crew runs its own", "This computer"), the gui_userremote-instancesslug, and the filed Fargate RFC. Catalog keys are untouched; only values move, exactly as #6950 did. CHANGELOG history is not rewritten, and the changelog-history gate confirms all nine shipped sections are intact.One derived identifier does move, and it is not a catalog key. Settings-registry ids are generated from the label, so relabelling the preview card shifted
developer.remote-instance-sessionstodeveloper.remote-crew-sessions— which would silently break a saved?highlight=developer.remote-instance-sessionsbookmark. ALEGACY_ID_EXACTentry inwebsite/src/hooks/useSettingHighlight.tsnow maps the old id onto the new one, the same mechanism the Slack-tab collapse and the Polly relabel use, with a revert-verified test inwebsite/src/test/settingHighlightAliases.test.ts(removing the alias fails 2 of its 8 tests). Caught by the First Principles lane, not by me; my earlier claim that nothing but values moved was wrong.The mechanical revert restored a pre-existing inconsistency, so the per-locale terms were normalized against each catalog's own precedent rather than shipped as found.
Two passages are authored, not restored
docs/system-specs/modules/crew-mode.mdexisted to say "a crew is not a Remote Instance". With both labels reading "crew", a literal swap disambiguates nothing, and the file post-dates both merges so there is no earlier text to recover — it now carries a short gloss naming the feature.docs/system-specs/modules/instances.mdcarried a note from #6950 stating "Remote Crew" was retired in favour of "instance". A faithful revert deletes it, but the UI/code mismatch it documents is real and is more confusing inverted, so it states the restored direction instead.Both are flagged for review as prose I wrote rather than text I restored.
Deliberate divergence
src/kiro_crew/docs/remote-instances.md(shipped last week by #12581) has its content and title restored to "Remote Crew" but keeps its filename, because that filename is a published URL and renaming it breaks inbound links. Its glossary row now carries both names. Renaming the file with redirects is filed as #12750 rather than buried here, at the Design lane's request, so the documented displayed/stored split does not quietly become permanent.Files touched
Catalogs and the pseudolocale:
website/src/i18n/locales/en.json,en.manual.json, the 11 translations and the regenerateden-XA.json, pluswebsite/src/i18n/instancesDisconnectHint.test.ts.Product surfaces, all panel-name references in copy or comments:
website/src/pages/settings/RemoteCrewPanel.tsx,website/src/pages/settings/InstancesPanel.tsx,website/src/pages/members/DeployMyCrew.tsx,website/src/pages/members/FargateCrewView.tsx,website/src/pages/members/MembersPage.tsx,website/src/pages/ChatSidebar.tsx,website/src/components/InstanceTabBar.tsx,website/src/surfaces/builtins.tsx,website/src/utils/previewFlags.ts.Tests re-pinned to the restored wording:
website/src/test/InstancesPanel.test.tsx,InstanceTabBar.test.tsx,InstancesViewport.test.tsx,InstancesViewport.winElectron.test.tsx,RemoteCrewPanel.test.tsx,RemoteCrewPanelSigninRecovery.test.tsx,SettingsRemoteCrewPanelCoverage.test.tsx,settingsLinkMigrations.test.tsx,EmbeddedSwitcher.test.tsx,App.mobileCrewSwitcher.test.tsx, pluswebsite/src/pages/members/DeployMyCrew.test.tsxandFargateCrewView.test.tsx.Harnesses and specs whose selectors and assertions track the restored strings:
website/scripts/capture-deploy-my-crew.mjs,capture-browser-tabs-demo.mjs,capture-crew-pin-chips.mjs,capture-crew-chip-shrink.mjs,capture-crew-tab-stable-order.mjs,capture-chat-on-crew-toggle-move.mjs,record-crew-pin-chips.mjs, andwebsite/playwright/settings.spec.ts. Five of these were already broken on this branch — they waited on[aria-label^="Switch instance"]while the catalog renders "Switch crew".Docs and data:
website/docs/extension-seams.md,website/docs/narrow-viewport.md,docs/build/desktop-app.md,docs/guides/README.md,docs/guides/remote-crew-on-ec2.md,docs/guides/cloud-instance-ssm-vs-ssh.md,docs/guides/adding-a-remote-provisioner.md,docs/system-specs/modules/platform-context.md,docs/feature-map/README.md,src/kiro_crew/data/tips_curated.json, and thetest/gui_usercatalog (features.json,features_catalog.py,scenarios.py, regeneratedFEATURES.md) whose display label moved while itsremote-instancesslug stayed.The revert also erased 99 real translations
Nine
pages.settings.remoteCrewPanel.*keys (changed_outside_this_form,crews_you_can_switch_to,launch_done,new_cloud_crew,no_crews,remove_warning,sign_in_unconfirmed,use_my_edits_anyway,your_crews) were untranslated English placeholders before #6950. #6950 translated them. A faithful revert therefore restores the placeholders — 99 values across 11 locales that would have shipped in English inside otherwise-translated panels.Those are restored to #6949-era translations with the instance-term swapped to each locale's crew term, rather than re-authored. This is a deliberate departure from a pure revert and the largest judgment call in the change. The
[changed-passthrough]gate — zero tolerance — now reports 0 changed values still reading as English, which is the check that would have caught the regression had it shipped.Per-locale renderings follow each catalog's own untouched precedent rather than the English:
de Remote-Crew,es crew remoto(masc.),fr crew distante(fem.),it crew remota,pt crew remota,ru удалённая команда,ja リモートクルー,ko 원격 크루,zh-CN 远程 Crew,bn রিমোট ক্রু,hi रिमोट क्रू. Where the revert restored a split — fr used three renderings, ko three, ru two — the locale was normalized to one, and article, gender and case agreement moved with it.Two fixes to pre-existing defects
launch_done_unsignedtold users the crew "appears in Your instances" while the list it names is labelledyour_crews— "Your crews". #6950 never touched that key, so neither the rename nor this revert fixed it; it has pointed at a non-existent list the whole time. All 12 locales now name their own label.ru.transport_locked_notewas the one value still carryingэкипаж, the term the rest of the ru catalog abandoned. The English source is "matches this crew to its AWS resources", so the possessive points at the crew itself; moving the noun to feminineкомандаcarriesэтот → этуandего → её. A Russian speaker should confirm that agreement.Known residual splits, left alone
ru components.kiroPrerequisiteGate.your_crew_is_almost_readystill saysВаш экипаж; it is unchanged by this branch, belongs to a different component and is not a remote-crew string.pt/es/zh-CNdisagree with other features' precedent for "crew" (Issue Radar usesequipo,小队, and masculinecrewin pt) — each catalog is internally consistent here, and picking one gender product-wide is a separate decision.demixesSieanddu,kostill shows원격 크루above untouched리모트 호스트field labels, andbn/himix register in these panels. All predate this change. Three more surfaced while translating the row-entity keys and are also left alone:esdisagrees with itself on the gender of "crew" (the row/list cluster is masculineel crew, the fiveidentity*keys are femininela crew);hidoes the same (translated values treatक्रूas masculine, the values that leave the English tokencrewuntranslated treat it as feminine); and every locale mixes the latin tokencrewwith its localized form inauto_connect,identity,identity_center_hint,identity_chooseandidentity_region_hint. That last one is pre-#6950 text this revert restored faithfully rather than a defect it introduced — the same reason the tworuauto_connectvalues were fixed only once they appeared inside an evidence frame. Settling any of them means picking one gender or one rendering product-wide, which is a vocabulary decision rather than part of this revert.Fixed after the UX lane's Watch item
UX caught a sentence this change had made self-contradictory. The remote-tab hint read "disconnect the instance in Settings → Remote Crew" — I renamed the panel inside that string and left the entity, so one sentence used both vocabularies while the tab above it says "Switch crew" and the destination panel lists "Your crews". Every remote-tab user would read a term the destination never uses for that thing. Fixed in all 12 locales, each keeping its own article, gender and case (
ruaccusative feminineкоманду,fr/itla crew,esmasculineel crew), plus the test and the capture harness that pinned the old English.While capturing the Russian spot-check UX asked for, the frame showed
Автоподключение crew— latin "crew" sitting directly under theУдалённая командаheading. Pre-existing on main and not caused by this change, but it is the same incoherence inside the very screenshot this PR submits as evidence, so those tworuvalues now readкоманд/каждую команду … к нейlike the rest of the panel.UX's second pass found one more of the dangling-pointer class, and it was a genuine miss:
pages.membersPage.deploy_state_gonesaid the machine "is no longer in your instances list" while the list it points at is labelledyour_crews.launch_done_unsignedgot that fix in the earlier round and this sibling did not, because my sweep matched the panel's name rather than the list it names. Fixed in all 12 locales, each referencing its own label exactly aslaunch_donealready does (de„Deine Crews“,ja「あなたのクルー」,ruв твоём списке командkeeping that value's own informal register), plus the two test assertions and three capture-harness assertions that pinned the old English. The harness was re-run against a fresh build, so the new sentence is a verified render rather than a string swap.The row entity is a crew again, not an "instance"
UX's third pass named the real shape of that class:
sign_in_unconfirmedsaid "then connect the crew" while its neighbours said "keep the instance" and "This instance exists" — one panel calling one row two things. The scope question is settled by the baseline rather than by taste:sign_in_unconfirmedis byte-identical to its pre-#6950 text, while 13 of its neighbours did not exist pre-#6950. They were authored during the three weeks "instance" was the shipped word, which makes them the same category as the residual keys already fixed here.A sweep of all 43 bare-"instance" values split them: 17 postdate #6950, and 15 of those are
remoteCrewPanelrow-entity strings (the sign-in, cancel/remove and provisioner clusters) now reading "crew" in English and all 11 locales. The other 26 predate the rename and stay — including the wholeapps.opsMissionControl.*namespace, where "this instance" means the gateway itself, andstamped_ec2_note, where "Its instance may still be running and billing" genuinely names the EC2 box.Two tests asserted the opposite premise and had to be inverted rather than deleted: both a unit test and a capture harness required the Cancel label to not contain "crew", each with a comment explaining that "instance" was the word the panel headers used. Those comments now state the deliberate split — the row is a crew,
instanceis reserved for the machine — and the inverted assertion is mutation-verified (reverting the English to "instance" fails it).Backend copy was in this class too:
src/kiro_crew/cloud/launch_job.pyhard-codes"Added to your instances.", rendered verbatim as a launch step's detail, so it pointed at the renamed list from Python where no catalog sweep could see it. Its siblings in that file ("Create the instance", "Could not reach the instance") name the machine and are unchanged. One curated tip was the same shape from a third direction:tips_curated.json's Remote Crew tip told users to "switch between instances from the header" while the control it names now reads "Switch crew". Its"route": "/settings/instances"deliberately stays — routes are on the stored side of the split.Screenshots
Every frame below postdates the
dist/build that carries this change.Settings, three renames in one frame: the nav item reads Remote Crew, the preview card reads Remote crew sessions, and the "Chat on a crew" description now points at "Settings > Remote Crew" — that last one is the cross-reference that used to send readers to a panel this change renames.
The dangling reference, fixed and rendered. This button read "Open Remote Instances in Settings" while the panel it opens is now "Remote Crew".
The scope boundary holding in the product: the list is Crews you can switch to, while the row below still says "Its instance may still be running and billing" for the EC2 box. Crew for the thing you talk to, instance for the machine.
The panel's empty state and its buttons, which UX listed as an uncovered surface — "No crews yet. Set one up, or add a machine you already have." above Add remote crew.
One non-English locale — the spot-check UX asked for against the 99 re-authored translations. Russian renders the feature as
Удалённая командаin the nav, the heading, the tab (Ваши команды), the list (Команды, на которые можно переключиться) and the button (Добавить удалённую команду): one term throughout, noэкипаж. The row beneath still readsЕго экземпляр может всё ещё работатьfor the EC2 machine, so the crew/instance split survives translation. The remaining English words here (Set up a new one,Disconnected,Added by you) are pre-existing untranslated passthrough thati18n:checkcounts at baseline, not values this PR touched.The top-bar crew switcher, whose control carries the restored
Switch crewlabel. Five committed capture harnesses were waiting on[aria-label^="Switch instance"]and were therefore already broken on this branch until this change fixed their selectors.Verification
tsc -bclean, eslint clean,i18n:checkall green underI18N_BASE_REF=origin/main(en-XA regenerated, 46 changed lines matching exactly its 36 + 10 English source lines),lint:i18nandi18n:renderat baseline, phantom-classes clean, docs-lint passing, and the black / isort / flake8 / mypy (1680 files) / brand / comment-history / focus-cue / harness-parity / changelog-history / feature-map / loop-bound-locks / testpaths / vendor-manifest gates all pass diff-scoped.Frontend: the full scoped set, 395 files and 9931 tests, all pass. That matters here: an earlier run of only the obviously-related files missed three that pin the renamed strings —
previewSurfaces.test.tsx(the toggle's accessible name),ChatSidebar.instanceSessionsMerge.test.tsx(fivechecking remote instancesregexes) andtest_remote_instances_doc.py(the doc page's pinned label). All three were genuinely red and are fixed.Backend scoped: 95 failures reduce to 1 caused by this change — the doc test above, now green. The remaining 94 were measured against a pristine worktree at the merge-base: 93 fail identically there (host-environmental —
faiss, no Node ≥ 20,ssh -G,env --default-signal, process reaping, Windows path rules), and 1 (test_perf_sampler) is a load flake that passes 3/3 in isolation on both trees.The two per-locale style gates were the last to go green.
bnregister (§5, no আপনি forms) andfrspacing (§1, U+202F before double punctuation) hold a branch responsible for every value it touches, and the restored translator text predates both guides — so 22 values were brought into line. Generated artifacts are idempotent: regenerating twice is byte-identical.Rebasing onto a moving main surfaced a live conflict worth noting: #12603 landed a Fargate view whose test asserted the old string, so it was red against the restored catalog until fixed here, along with four new harness assertions it added.
Every value change was applied by script with an explicit per-locale map and a machine check that each term introduced is attested in translator-authored text in the same catalog — that check caught eight Korean values corrupted by hand-written escapes (
원감for원격,여기for열기).Why this change exists
This is a maintainer decision, recorded here: the repo owner asked for the "Remote Instances" vocabulary to go back to "Crew". It is a relabel, not a defect fix, and I am not claiming a user report — there is none.
What it is grounded in is a concrete incoherence rather than symmetry with the past. Shipped
mainlabels the feature "Remote Instances", while in-flight work onfeat/remote-crew-railpresents the same objects as "crews" in the rail header, hover and switcher. Those two cannot both ship: whichever lands second makes the product say both words for one thing. This PR settles it in the direction the owner chose, which also happens to be the direction the rail already assumes.The counter-argument deserves stating, because the First Principles lane is right that it exists: #6950 retired "Remote Crew" precisely so the label would match the code and config it sits on (
/api/instances,instances.json,InstancesPanel), and that reason has not evaporated — this change reinstates a displayed/stored split. The judgment is that a displayed name should read for the person using it and a stored name for the person maintaining it, and that "Remote Crew" beside "Kiro Crew", "Crew Member" and "Issue Radar crew" is the coherent product vocabulary even though it diverges from the route names.remote-instances.md's glossary now carries both names so the split is documented rather than hidden. If you disagree with that premise, this PR is the wrong shape and the three pre-existing fixes (launch_done_unsigned,deploy_state_gone, theruterm) can be lifted out and shipped alone — say so and I will split them.The UX lane raises the sharpest version of the objection: after this change "crew" names two different things — the machine you connect to, and the group of agents ("Crew Members"). That overload is real on
maintoday. It is also being resolved from the other side: #11869 (chore/agents-vocabulary, open) renames Crew Members → Agents across the dashboard, which leaves "crew" meaning only the remote machine. The two PRs are complementary rather than competing, and whichever lands second ends the ambiguity. If #11869 is abandoned, the overload becomes permanent and UX's suggested mitigation — a one-line gloss where the two senses collide — is worth taking. I have deliberately not invented that gloss here: it would be new product copy in 12 locales for a naming problem whose better fix may be the other PR.Reviewer note
Anyone who learned "Remote Instances" from the shipped docs sees the term change under them. That is the deliberate call this PR asks you to confirm.
No linked issue: this reverses a shipped rename at the maintainer's request rather than closing a tracked defect.