Repository navigation
refactor: rename Ultra Mode to conductor mode - #36924
Conversation
Sensitive Change ReviewSensitive files changed:
How to clear this check:
Only the author's comment counts this way. A reviewer can instead |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks. Every item is the rename. Issue #36869 and the maintainer-signed decision entry are the source for it. The values an install has already saved keep their old spelling on purpose. What this change shipsInventory (10 items) — 10 justifiedIntent: people see the crew feature under its final name, conductor mode, and nothing an existing install saved stops working. This is a FIX: a relabel.
[FIRST-PRINCIPLES-REVIEWED] 3503e0f Description read[DESCRIPTION-READ] 1a3df101b2ef0a2d22b378fbc190ef6044cd8be49063fb049b8fdb0a9caebc7c That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — ✅ human override acceptedReviewed Human judgment by @bolichen97 overrides the Opus 5.5 finding for [OPUS-OVERRIDE] 3503e0f Verdict recorded from an authorized human decision for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 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: |
3c4b8e6 to
54a2f62
Compare
|
The title and the Goal: line now use the final product name, conductor mode, at a maintainer's request. The goal is otherwise unchanged. |
|
/ai-review override design 54a2f62 not-applicable: agent: owner-authorized (maintainer ruling): diff is a mechanical rename over the lane size limit; reviewed locally in slices by GPT and Opus, no open findings; evidence in the PR body |
|
/ai-review override ux 54a2f62 not-applicable: agent: owner-authorized (maintainer ruling): diff is a mechanical rename over the lane size limit; reviewed locally in slices by GPT and Opus, no open findings; evidence in the PR body |
Human judgment recorded@bolichen97 recorded a not-applicable release (class
This decision applies only to this commit. A new push requires a new judgment. |
Human judgment recorded@bolichen97 recorded a not-applicable release (class
This decision applies only to this commit. A new push requires a new judgment. |
Design Review (Opus 5.5) — ✅ human override accepted@bolichen97 overrode this lane for |
UX Review (Opus 5.5) — ✅ human override accepted@bolichen97 overrode this lane for |
|
/ai-review override gpt 54a2f62 not-applicable: agent: owner-authorized (maintainer ruling): the blocking finding is the by-design recorded-decisions gate on the new superseding decision entry; the diff is a mechanical rename reviewed locally in slices by GPT and Opus, no open findings; evidence in the PR body |
Human judgment recorded@bolichen97 recorded a not-applicable release (class
This decision applies only to this commit. A new push requires a new judgment. |
54a2f62 to
29534d3
Compare
The crew feature's final product name is conductor mode. Every user-facing place that said Ultra Mode says conductor mode: UI strings in all locales (kept untranslated), settings labels and their generated search entries, gateway notices, agent prompts and skills, shipped docs, specs and the feature map. Code names that carried the old name (components, CSS classes and tokens, test ids, i18n keys, tests, capture harnesses) follow it, with a ConductorMode / conductor_mode / conductor-mode stem so they never collide with the existing conductor agent-role names. Saved state keeps working: ids that never said ultra stay (crew_mode, dashboard.new_chat_crew_mode, mc-preview-crew-mode), the ultra_mode_only refusal code and the ultra root_kind trace value keep their spelling, and a transcript that stored the old agent-switch notice still renders as that notice. A new decision entry records the rename and maps the earlier entries to the new name; those entries stay as written.
29534d3 to
3503e0f
Compare
|
/ai-review override design 3503e0f not-applicable: agent: owner-authorized (maintainer ruling): diff is a mechanical rename over the lane size limit; reviewed locally in slices by GPT and Opus, no open findings; evidence in the PR body |
|
/ai-review override ux 3503e0f not-applicable: agent: owner-authorized (maintainer ruling): diff is a mechanical rename over the lane size limit; reviewed locally in slices by GPT and Opus, no open findings; evidence in the PR body |
Human judgment recorded@bolichen97 recorded a not-applicable release (class
This decision applies only to this commit. A new push requires a new judgment. |
|
Human judgment recorded, but the following reviewer lane(s) could not be re-run automatically: Design Review. Re-run the lane's latest workflow run for |
Human judgment recorded@bolichen97 recorded a not-applicable release (class
This decision applies only to this commit. A new push requires a new judgment. |
|
Human judgment recorded, but the following reviewer lane(s) could not be re-run automatically: UX Review. Re-run the lane's latest workflow run for |
|
/ai-review override fable 3503e0f not-applicable: agent: owner-authorized (maintainer ruling): the blocking finding is the by-design recorded-decisions gate on the new superseding decision entry; the diff is a mechanical rename reviewed locally in slices by GPT and Opus, no open findings; evidence in the PR body |
Human judgment recorded@bolichen97 recorded a not-applicable release (class
This decision applies only to this commit. A new push requires a new judgment. |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Reviewed head 3503e0f. The rename is complete. No removed Python symbol is still referenced in src, test or website. The i18n keys map one to one from ultra_* to conductor_mode_* in the catalogs compared, with no placeholder drift. Every renamed CSS class, token and test id has a definition. Test case counts are the same before and after, with no assertion dropped. A merge with current main is clean and leaves no old name outside the documented allowlist.
Done well: the values running installs already hold keep their spelling, and each has a comment saying why. These are the ultra_mode_only refusal code, the ultra trace kind, the span names and telemetry.traces_ultra_mode. The gatewayNotice regex also accepts both wordings, with a test for each.
Optional:
- test/test_session_control_crew.py:653 compares against
crew.CONDUCTOR_MODE_ONLY, so it would still pass if the wire value changed. Asserting the literal"ultra_mode_only"would make the backwards-compatibility claim hold. - The decision entry supersedes two recorded decisions. A short PR comment from the deciding maintainer that states the ruling, cited from the entry, would give it a clear record.
- Open PRs #36830, #23512, #31518 and #37044 add new "Ultra Mode" text. #23512 also asserts the old aria-label. These authors will need to rebase onto the new name after this lands.
The crew feature's final product name is conductor mode. Every user-facing place that said Ultra Mode says conductor mode: UI strings in all locales (kept untranslated), settings labels and their generated search entries, gateway notices, agent prompts and skills, shipped docs, specs and the feature map. Code names that carried the old name (components, CSS classes and tokens, test ids, i18n keys, tests, capture harnesses) follow it, with a ConductorMode / conductor_mode / conductor-mode stem so they never collide with the existing conductor agent-role names. Saved state keeps working: ids that never said ultra stay (crew_mode, dashboard.new_chat_crew_mode, mc-preview-crew-mode), the ultra_mode_only refusal code and the ultra root_kind trace value keep their spelling, and a transcript that stored the old agent-switch notice still renders as that notice. A new decision entry records the rename and maps the earlier entries to the new name; those entries stay as written. Hand-resolved backport; original author: Bolin Chen. (cherry picked from commit 176820c)
…ename The rename changes the text of the two conductor agent specs and of _CONDUCTOR_SYSTEM_PROMPT. Main re-pinned the same pins in #36924, but release/0.13.0 builds those specs from its own prompts, so its digests differ from main's. Only kirocrew-conductor.json and kirocrew-ledger-conductor.json change; every other pinned spec keeps its bytes.
The crew feature's final product name is conductor mode. Every user-facing place that said Ultra Mode says conductor mode: UI strings in all locales (kept untranslated), settings labels and their generated search entries, gateway notices, agent prompts and skills, shipped docs, specs and the feature map. Code names that carried the old name (components, CSS classes and tokens, test ids, i18n keys, tests, capture harnesses) follow it, with a ConductorMode / conductor_mode / conductor-mode stem so they never collide with the existing conductor agent-role names. Saved state keeps working: ids that never said ultra stay (crew_mode, dashboard.new_chat_crew_mode, mc-preview-crew-mode), the ultra_mode_only refusal code and the ultra root_kind trace value keep their spelling, and a transcript that stored the old agent-switch notice still renders as that notice. A new decision entry records the rename and maps the earlier entries to the new name; those entries stay as written. Hand-resolved backport; original author: Bolin Chen. (cherry picked from commit 176820c)
…ename The rename changes the text of the two conductor agent specs and of _CONDUCTOR_SYSTEM_PROMPT. Main re-pinned the same pins in #36924, but release/0.13.0 builds those specs from its own prompts, so its digests differ from main's. Only kirocrew-conductor.json and kirocrew-ledger-conductor.json change; every other pinned spec keeps its bytes.
Problem / Motivation
Goal: Every place a person reads the crew feature's name says conductor mode instead of Ultra Mode, and nothing an existing install saved stops working.
A maintainer ruling settled the feature's final name as conductor mode. Today the composer chip, the Sessions row tag, the Board panel, settings labels and search entries, gateway notices, agent prompts, skills, shipped docs and specs all still say Ultra Mode.
Why it matters
The old name is on every surface of a feature that is being launched. Each day it stays, more people, docs and saved prompts learn a name that is about to go away, and the code keeps naming one feature two ways.
Not a goal
crew_mode,crew_root,dashboard.new_chat_crew_mode,agent.crew_mode_agent, themc-preview-crew-modeflag, thecrewview id.ultrareply-verbosity level, or the glassultrathin/ultrathickthicknesses. They are other features.CHANGELOG.md.What changed (motivation → approach → change)
People now see conductor mode wherever they saw Ultra Mode. It is written the way the dashboard writes "bots": "Conductor mode" at the start of a label or sentence, "conductor mode" inside one. Non-English catalogs keep the English name, written "Conductor mode" as they write "Bot". One mechanical rule set does the rename, so the code, the docs and the UI cannot drift apart.
ConductorModeCrewPanel,ConductorModeEffect,conductorModeEffect.ts,conductor_mode_*i18n keys,conductor-mode-*CSS classes, tokens and test ids, test names, capture harnesses,conductor_mode_shape()in tracingdocs/decisions/2026-11-13-the-crew-mode-feature-is-named-conductor-mode.mdand its README rowThe same swap reaches comments, strings and names in
src/kiro_crew/agent_materialization,src/kiro_crew/agent_parts,src/kiro_crew/builtin_skills,src/kiro_crew/config,src/kiro_crew/context_assembly,src/kiro_crew/crew_log,src/kiro_crew/crew_mode.py,src/kiro_crew/crew_pause.py,src/kiro_crew/dashboard_agentic.py,src/kiro_crew/docs,src/kiro_crew/mcp_dashboard_caller_scope.py,src/kiro_crew/mcp_dashboard_session_runs.py,src/kiro_crew/mcp_tools,src/kiro_crew/security_posture_sinks_gateway.py,src/kiro_crew/slack,src/kiro_crew/tracing, andwebsite/src/api,website/src/hooks,website/src/shell,website/src/store,website/src/stylesandwebsite/src/uiLocations;config-baseline.jsonis regenerated."Conductor" already names the agent role (
kirocrew-conductor,conductor_close.py, theconductor-needs-you/conductor-status/conductor-view-*test ids). So every renamed code name carries aConductorMode/conductor_mode/conductor-modestem and never collides with them. Prose such as "the Ultra Mode conductor" is rewritten ("the conductor in conductor mode") rather than swapped.Two labels keep Title Case because their sibling labels are Title Case: "Start New Chats in Conductor Mode" (settings) and "Traces: Conductor Mode Shape" (config).
The new decision entry supersedes, in part, the 2026-10-29 entry that named the surfaces Ultra Mode, and the 2026-10-30 line that kept the word "conductor" out of user text, for the feature's own name only. The other entries written under the old name stay byte-identical, and the new entry says they apply to conductor mode.
After the change,
git grep -i ultrafinds only:ultraverbosity level and itsultra-briefchange-card fixture, glassultrathin/ultrathick, effort and quality test values,jev-ultrafast, a font name, plain English and Portuguese words, and false matches such asdefaultRaf;src/kiro_crew/_vendor/(2 files);CHANGELOG.md;Allowlist (kept on purpose):
ultra_mode_only: the remote-dispatch refusal code. A conductor on another machine matches on it. The Python constant is nowCONDUCTOR_MODE_ONLY, and the test that pins the code keeps its name.ultra: the value of thekirocrew.crew.root_kindtrace attribute, the trace kind behindtrace.ultra, and the span namesultra round N/ultra outcome .... Exported traces carry them and saved trace queries filter on them.telemetry.traces_ultra_mode: a saved config key, read as-is.Ultra Modeinwebsite/src/pages/chat/gatewayNotice.tsand one case in its test, so a transcript that stored the old agent-switch notice still renders as that notice.Backwards compatibility
Compatible: no config key, localStorage key, saved setting, route or event a running install stores is renamed. The wire values, the saved config key and the stored notice wording above keep reading.
Reader: website/src/pages/chat/gatewayNotice.ts:agent_switch_blocked_crew_mode -- chat -- the regex matches both the "conductor mode" and the stored "Ultra Mode" wording; gatewayNotice.test.tsx covers both
Reader: src/kiro_crew/dashboard/handlers/remote_dispatch_routes.py:CONDUCTOR_MODE_ONLY -- crew page -- value stays
ultra_mode_only; test_a_plain_owner_chat_is_refused_ultra_mode_onlyReader: src/kiro_crew/tracing/crew.py:ROOT_KIND_CONDUCTOR_MODE -- chat -- value stays
ultra; test/tracing/test_tracing_runs.pyReader: src/kiro_crew/tracing/provider.py:conductor_mode_shape -- chat -- still reads
telemetry.traces_ultra_mode; test/tracing/test_tracing_work.pyReader: website/src/components/commandPalette/settingsRegistry.gen.ts:developer.conductor-mode -- chat -- setting ids are slugs of the English label and are not stored; an old
?highlight=developer.ultra-modelink still opens the Feature Previews page, without the highlightReader: website/src/i18n/locales/.json:components.crewMode.conductor_mode_ -- chat -- keys ship inside the bundle; every caller is renamed in the same commit (
npm run i18n:check)Tests
Renamed tests keep their assertions; text assertions follow the new casing.
gatewayNotice.test.tsxadds one case: a stored "Ultra Mode" agent-switch notice still resolves to the notice text.Proof: not provable -- other: refactor (rename) PR with no behavior fix; the one behavior pin is a frontend test
Determinism (testing-conventions D12): the 9 renamed backend tests passed 20 of 20 runs at
-n0, their files passed 10 of 10 at-n 4(288 tests), and they passed in a shuffled order.gatewayNotice.test.tsxpassed 20 of 20 runs, and the 7 changed frontend test files passed shuffled (335 tests).Local runs on the reviewed patch: black (26.10.1), isort, flake8, mypy (4627 files), comment-history, brand name, file line cap, docs-lint, decision-ledger history, feature map,
npm run i18n:check(diff-scoped),tsc, semgrep (0 findings), the flake-pattern ratchet, 197 related frontend test files and 274 related backend test files plustest/tracing. Two backend tests fail the same way on a cleanorigin/maincheckout:test_find_ui_auto.py::test_the_generated_auto_tier_answers_labels_only_it_carries[continue with github...]andtest/tracing/test_tracing_otlp.py::test_a_header_echoed_back_by_a_failure_reaches_no_log_record[library_error].Local review in slices
The diff is 1.1 MB, over the 1 MB limit at which the AI review lanes fail closed. It was reviewed locally on this head in five slices, each by the model-pinned GPT and Opus lanes with the CI contract prompts:
src/kiro_crew/**,docs/**,AGENTS.md,config-baseline.json(397 KB)maindid) leftSee what ... looks likeregex that could no longer match, so the check was vacuous; heading-comment casing and ruler width): fixedgit grepfor leftovers, every new name against its definition and imports, collisions with existing conductor names, interim-name check?highlight=links and stored cards): rebutted, neither id has a card write path (resolve_settingrefuses both withno_write_path), and an old link still opens the right pageThe four slice diffs sum to the whole diff (190 files, 1,111,772 bytes). After the fixes, one GPT and one Opus verifier checked the fix delta on this head: both PASS, no findings.
Manual verification
I built the SPA before and after the change and loaded it gateway-free (dark theme, English): a conductor chat with the Board open at 1400 px and 390 px, and a new chat with the chip pressed at 1400, 390 and 320 px. The chip is 141 px wide (it was 109 px), the Sessions row tag fits beside the time, the footer mark fits, and the Board tab still reads "Board". The welcome title stays on one line at 390 px and wraps cleanly onto two lines at 320 px.
Screenshots / video
Board open, and the welcome at 390 px and 320 px
Related Issues
Closes #36869
Waiting for a maintainer (human-only):
recorded-decisions-are-not-regressions: the new decision entry supersedes part of two recorded decisions, so this gate blocks by design (docs/decisions/README.md rule 3) until a maintainer restates the ruling on this PR.src/kiro_crew/security_posture_sinks_gateway.pychanges (one description string), so it needs the reasoned approval of a maintainer.Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)