Repository navigation
fix(members): one roster reading, and no New team door for now - #17835
Conversation
300e746 to
609664c
Compare
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS The RFC ( [DESIGN-REVIEWED] 53d44d9 Description read[DESCRIPTION-READ] 70b94ff93b0ff749a8f9f36766e07cec126e2b8a3b9c1b1accae248c7b508c1f That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
UX Review (Opus 5.5) — 🟡 CONCERNSUX-level review of UX-Verdict: CONCERNS Rows that appear only when you search the chip look identical to listed crewmates, so the reader would not dare click them. Watch
[UX-REVIEWED] 53d44d9 Description read[DESCRIPTION-READ] faaa32650f405abcfff8b231b5454533cf2a6be044db08969f60da891a49365c That sha256 names everything this verdict read: the PR description, plus the 9 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ human override accepted@CrysisDeu overrode this lane for |
Opus 5.5 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: |
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: |
609664c to
59b1819
Compare
59b1819 to
4b18f54
Compare
4b18f54 to
bb76105
Compare
The Crewmates column lists rows through `rosterPopulation` / `listedByDefault` -- created, chatted-with, starred rows and the default crew -- while the header chip's `CrewmateSwitcher` was handed the raw roster and listed everything, so a user with two crewmates read "2 crewmates" beside a chip that said 10 and offered every sync-generated template. The switcher now runs the same two functions over the same roster: the count, the face stack and the needs-you cue read the listed population, and a typed search falls back to the full roster exactly as the column's search does. The chip's agent-template search term moves into the shared `matchesSearch`, so one definition of "what the search reaches" serves both placements.
b24e086 to
2911745
Compare
Human judgment recorded@CrysisDeu marked the first-principles AI finding as false positive, not applicable, or explicitly accepted for
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: First Principles Review. Re-run the lane's latest workflow run for |
Teams are not ready to be offered as something to create, so the roster header's "+" no longer carries New team: its only door is New crewmate. Everything a team already has keeps working -- the roster groups by team, a group header opens its team view, and that view's Edit still opens the dialog in edit mode. The dialog's create mode, its labels and `api.teams.create` all stay, so the door comes back by reverting this commit. The two page cases that drove the removed item become one that pins its absence, and create-mode coverage moves to TeamDialog's own test, which can still reach it.
2911745 to
53d44d9
Compare
Human judgment recorded@CrysisDeu marked the first-principles AI finding as false positive, not applicable, or explicitly accepted for
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: First Principles Review. Re-run the lane's latest workflow run for |
Problem / Motivation
Goal: The Crewmates page gives one answer to "who is on my crew?", and stops offering a team you cannot finish making.
Two things on that page disagreed with themselves. The header's crewmate chip counted every row the gateway returns, while the roster column counted only the crewmates you have actually chatted with, created or starred — so one user read
2 crewmatesin the column beside a chip saying10, and another read6beside a chip offering 12. The chip is also the only roster affordance on the desktop DM view, so the extra rows it offered were reachable nowhere else.Separately, the
+menu still offered New team, a door into a concept that is not ready to be offered.Why it matters
The count is the page's one-line answer to who you work with. Two numbers on one screen make both untrustworthy, and the chip let a user switch into a sync-generated template they never created and the column refuses to list. The team door is the part of an unfinished feature a first-time user is most likely to walk into.
Not a goal
api.teams.createstay; hiding the door is one revertible commit./api/members.What changed (motivation → approach → change)
The chip and the column were asking the same question through different code: the column ran the roster's hide rule, the chip ran a bespoke filter over the raw list. Now there is one reading.
CrewmateSwitchercalls the column's ownrosterPopulationandrosterShows, so its count, its faces and its needs-you dot read the listed crewmates, and typing in its search reaches a hidden row exactly as the column's search does. The chip's one extra search term — matching a row by the agent template it runs — moved into the shared matcher, so "what the search reaches" has one definition and the column gained that term too.+menuThe team door is simply not drawn; the comment where it was says what returns it. The two page tests that drove it become one that pins its absence, and create-mode coverage moves to the dialog's own test, which can still reach it. The docs that named the door — the feature map, the crew-mode spec, the
learn-cron-dashboard.mdspec'sTeamDialogline and the shippedcrew-members.mduser doc — now describe what the page actually offers, and the feature map also records that the folded roster runs the column's hide rule. A new capture harness underwebsite/scriptsproduces the screenshots below and asserts both readings, and the existing create-page harness now expects one door.Backwards compatibility
Compatible: no API, config key or stored shape changes; a crewmate hidden from the chip was already hidden from the column, and
/api/membersreturns exactly what it did.This diff takes away one UI path (the roster
+menu's New team item), so every reader of team creation:Reader:
website/src/pages/members/TeamDialog.tsx:save — crew page — create mode is untouched and still posts{name, members}; covered byTeamDialog.test.tsx(gate on a name, roster-ordered body, failure notice, draft guard)Reader:
src/kiro_crew/dashboard/handlers/teams.py:api_teams_create — crew page — route, payload and validation unchanged; no dashboard door calls it while the item is hiddenReader:
src/kiro_crew/crew_teams.py:create_team — release — the operator's ownkirocrew agent create/deletepath into the team document is unchangedReader:
website/src/pages/members/TeamView.tsx:team-edit — crew page — the team view's Edit still opens the same dialog in edit mode; covered by the existing MembersPage team casesReader:
website/scripts/capture-crewmates-page-create.mjs:assertHeldMenu — release — asserted the item stayed live under a create hold; now asserts the menu offers one doorReader:
test/gui_user/scenarios/crewmate-team-view.yaml— nightly GUI run — its only purpose was to create a team through the+menu, which no longer exists; the scenario is retired (file deleted, dropped fromSHIPPEDand the members group intest_scenarios_and_report.py, catalog count 41 → 40 scenarios, 9 → 8 nightly), andcrewmate-create-first.yamlnow describes the menu as one rowReader:
src/kiro_crew/docs/crew-members.md:"Creating and editing" / "Teams" — release — the shipped user doc said the+menu had two rows and that New team creates one; it now names the one row and says team creation is not offered for now, with Edit team as the way to change an existing teamTests
CrewmateSwitcher.test.tsx— an unchatted app row is neither listed nor counted while its name is still reachable by typing; the open crewmate stays listed even when the hide rule would drop it.rosterFilter.test.ts— the shared search matches a row by its agent template.TeamDialog.test.tsx(new) — create mode: the checklist and the one-team rule, Create gated on a name, the roster-ordered body it posts, a failed create said inside the dialog, and the unsaved-draft guard against Escape.test/gui_user/test_scenarios_and_report.py— the retired nightly team scenario is gone from the shipped set and the catalog counts.MembersPage.test.tsx— the+menu offers exactly one item and never opens the dialog in create mode; the draft guard now runs through the reachable Edit door.website/src/pages/memberspass.Manual verification
Both readings were captured against a real build of this branch with a fixture roster (six chatted crewmates across two teams, beside six package templates nobody has chatted with): before, the column said
6 crewmatesand the chip said12; after, both say6, and typingconductorin the chip still reaches five hidden templates.One foreign red worth naming: on this machine
src/test/ArtifactDetailPage.test.tsxfails six editor cases inside the lazyPierreEditorimport, and it fails identically on a cleanmainworktree with none of these changes, so it is local dependency state, not this diff.Screenshots / video
The roster column and the header chip in one frame, before and after.
The hidden rows are still reachable by typing, and the
+menu now has one door.The same pair on a two-crewmate roster
Related Issues
Closes #17833
Closes #17834
Pattern harvest
Rule candidate: review-prompt
Pattern: a folded or collapsed view of a list filters the list itself instead of calling the full view's own filter, so the two disagree about how many items there are.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)