Skip to content

fix(members): say which crewmates the roster hides - #22296

Merged
iamwhatever merged 2 commits into
mainfrom
fix/21852-roster-hidden-count
Oct 20, 2026
Merged

iamwhatever merged 2 commits into
mainfrom
fix/21852-roster-hidden-count

Conversation

@cruisercohen

@cruisercohen cruisercohen commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: The Crewmates roster says how many crewmates it hides and how to reach them, so its count is not read as the whole crew.

The roster lists a crewmate unasked only when its thread holds a message, it is starred, or it is Mate, and its count covers those rows only. In the nightly GUI user test a crew of three read as "2 crewmates". The third, created but never chatted with, turned up only after a search for its name matched it (#21852). #22142 reported the same friction in another scenario and was marked its duplicate.

Why it matters

A user who creates a crewmate and has not chatted with it yet cannot see it on the page, and nothing says it exists. The count reads as "your crew is these two", so the natural conclusion is that the crewmate was lost.

Not a goal

  • Changing the listing rule. Hidden rows stay hidden until the search reaches them, as the crew-mode spec states.
  • Changing the header count or the filter tallies. They still count the listed rows plus any row the search reaches, so the column and the header chip keep one reading.
  • The folded roster in the header chip (CrewmateSwitcher), which opens with its own search field.

What changed (motivation → approach → change)

  • Symptom: "2 crewmates" on a crew of three; the third appears only on a matching search.
  • Cause: the count reads rosterPopulation (listed rows plus search hits) on purpose, and the only hidden-row notice (all_hidden_hint) is drawn when every row is hidden. With at least one listed row, nothing said that more existed.
  • Change:
    • useMembersPageRosterSignals derives hiddenCount, the roster rows minus the shown population. It is non-zero only while no search is typed, no filter narrows the list, and at least one row is listed (the every-row-hidden case keeps its own line). It also derives hiddenNames, the row labels (rowLabel, now exported from rosterFilter) of up to three hidden crewmates (HIDDEN_NAMES_MAX): the ones the user made (origin Mine, or created on the dashboard) first, then roster order, so a crewmate just created is named rather than counted behind package and app rows.
    • MembersPageRoster ends the list with one muted item (member-hidden-hint). With up to three hidden crewmates it names them all, joined with fmtList: "Recipe Box is hidden because you haven't used it yet, and isn't counted above. Search to find it." Past three it names three and counts the rest: "Recipe Box, pkg-budget, pkg-notes and 1 more crewmate you haven't used yet are hidden and aren't counted above." The hidden set often includes dozens of package and app crewmates, so naming every one stops helping past a few, while the named ones still give the reader something to type. Both forms say the count above leaves the hidden crewmates out, so the count keeps its meaning (shown crewmates) and is not read as the whole crew. The counted form names no action: the crewmates it does not name give the reader nothing to search for, so it only states that they exist and are not counted.
    • New plural keys pages.membersPage.hidden_named_hint and pages.membersPage.hidden_more_hint, registered in pluralKeys.json, in all 12 locales with each language's own plural forms, plus the regenerated en-XA.
    • The crew-mode spec and the agent-facing crew-members.md doc say the line exists.
    • website/scripts/capture-members-hidden-hint.mjs renders the frames below from fixtures (second commit); ROSTER=many adds three hidden package rows for the count form.

Backwards compatibility

Compatible: the change adds one list item and two catalog keys, and exports one existing helper; no API, stored shape or existing string changes.

Removes nothing: the listing rule, the count, the filters and every existing notice are unchanged; the diff only adds a line.

Tests

  • MembersPage.filters.test.tsx, in the "hides unlisted crewmates" block, seven new cases:
    • with 3 listed and 2 hidden, the line names both ("legacy-aim and pkg-tool are hidden because you haven't used them yet, and aren't counted above. ..."), is the list's last item, goes away while a search is typed and comes back when the search is cleared;
    • one hidden crewmate is named by its display label ("Nova Sky"), in the singular;
    • past three hidden crewmates the line names three and counts the rest, singular ("... and 1 more crewmate ...") and plural ("... and 2 more crewmates ..."), and ends at "... aren't counted above." with no search instruction (the whole line is asserted);
    • a crewmate the user made that sorts last by name, behind four package rows, is still named, and named first ("Zebra Notes, legacy-aim, pkg-three and 2 more ..."); this case fails on the previous head b3cfd4e, which named hidden rows in roster order only;
    • no line when nothing is hidden;
    • no line while the starred filter narrows the list.
  • Failing first: on origin/main 3ad0780 the first five cases fail (no member-hidden-hint element). The other two pass on both and stand as guards.
  • Also run: npx vitest run src/pages/members src/i18n (94 files, 1,860 tests, including catalog parity, plural forms and key references) pass; eslint on the touched files; npm run build (includes tsc); npm run i18n:check with I18N_BASE_REF=origin/main; scripts/docs-lint.sh; test/test_crew_members_doc.py and test/test_crewmate_guidance_copies.py.

Manual verification

Built the SPA from origin/main 3ad0780 and from this branch, and ran the harness against each with a stubbed API. Fixture roster: two crewmates with messages and one created but never chatted with (ROSTER=many adds three hidden package rows). What the DOM reads:

build, frame count listed rows hidden line
main, at rest 2 crewmates trip-planner, garden-helper none
this PR, at rest 2 crewmates trip-planner, garden-helper "Recipe Box is hidden because you haven't used it yet, and isn't counted above. Search to find it."
this PR, at rest, ROSTER=many 2 crewmates trip-planner, garden-helper "Recipe Box, pkg-budget, pkg-notes and 1 more crewmate you haven't used yet are hidden and aren't counted above."
both, search "recipe" 3 crewmates recipe-box none

Screenshots / video

Roster column, 1500×940 viewport at 2x, light theme; each frame is 528×1780 px.

Before (main), at rest:

Roster at rest on main: two crewmates, no hidden line

After (this PR), at rest, one hidden crewmate named, and the line says the count above leaves it out:

Roster at rest with this PR: two crewmates and a line naming the hidden one and saying it is not counted above

After, at rest, four hidden crewmates: Recipe Box (made by the user) and two package rows named, one counted, none of them in the count above:

Roster at rest with four hidden crewmates: Recipe Box and two package rows named, one more counted, and the line says they are not counted above, with no search instruction

After, search "recipe" (byte-identical to main's frame):

Roster with a search typed: the hidden crewmate is listed and the line is gone

Related Issues

Fixes #21852

Pattern harvest

Not generalizable: the gap is this roster's own copy for its own hide rule; there is no shared list component that a rule could target.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Contributed under the project's Apache-2.0 license, per CONTRIBUTING.md.

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 19, 2026
@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 1a887a3e40cceab83068942c14f2fe970d05f056 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: PASS

Check that hiddenCount (members.length − shownMembers.length) and hiddenNames (built from orderedMembers) always cover the same rows; otherwise the named/counted forms disagree.

What this change ships

Inventory (7 items) — 7 justified

Intent: a user whose crew has a crewmate they never chatted with can tell it exists and how to reach it, instead of reading "2 crewmates" as the whole crew (#21852). This is a FIX.

  1. The Crewmates list now ends with a muted line naming up to three hidden crewmates and saying the count above leaves them out — justified
  2. Past three hidden, the line names three and counts the rest, with no search instruction — justified
  3. Crewmates the user made are named ahead of package and app rows — justified
  4. The line disappears while a search is typed or a filter narrows the list, and when every row is hidden (that case keeps its existing line) — justified
  5. rowLabel is exported from rosterFilter for the hidden-names derivation — justified
  6. Two new plural catalog keys in all 12 locales, plus en-XA, registered in pluralKeys.json — justified
  7. The crew-mode spec, the agent-facing crew-members.md, and a screenshot harness script (one of ~787 sibling capture-*.mjs scripts) — justified

[FIRST-PRINCIPLES-REVIEWED] 1a887a3

Description read

[DESCRIPTION-READ] 69d95bdb4c070d748f7e891fbd9d7019c0322bc7a48d6f70feef947c83f81742

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🟡 CONCERNS

UX-level review of 1a887a3e40cceab83068942c14f2fe970d05f056 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

The hidden-crewmates line says hidden crewmates "aren't counted above", but searching makes the count jump to "3 crewmates" while one row shows, so the reader can't trust either.

Watch

  • The "isn't counted above" promise doesn't survive a search. The new copy defines the count as "shown crewmates". In shot-04, typing "recipe" removes the line and the header reads "3 crewmates" above a single row. The blind reader: "the earlier note said Recipe Box 'isn't counted above', yet now it seems to be counted… I can't tell whether the count means 'the ones shown', 'the ones I've used' or 'the total'." This happens on every search, so the line this PR adds is what makes the old count look inconsistent. Smallest fix: drop "and isn't counted above" and keep the line to what's hidden and how to find it.
  • The counted form loses the next step and shows raw ids. Shot-03 (hidden_more_hint) reads "Recipe Box, pkg-budget, pkg-notes and 1 more crewmate…". The reader said "'pkg-budget'… look like technical labels" and "this time it doesn't tell me to search". This hits any user with package or app crewmates, which the PR says is the common case. Fix: put "Search to find them." back for the named crewmates, and name only display-labelled rows before counting the rest.

Suggestions

  • In member-hidden-hint, make each named crewmate a button that fills in the search, or opens that crewmate's chat, so the user doesn't have to retype a name the line already shows.

[UX-REVIEWED] 1a887a3

Description read

[DESCRIPTION-READ] 2226c9762a43fe9f22862450c096afb7114e47ffa76b95c5ac9f5b59d7e9269d

That sha256 names everything this verdict read: the PR description, plus the 7 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 .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 1a887a3e40cceab83068942c14f2fe970d05f056 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Nothing to check.

[DESIGN-REVIEWED] 1a887a3

Description read

[DESCRIPTION-READ] 33516f709534e55d8fd13c040dd8ea9a87207ff8c815933a8d529550a03f96a4

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 .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 1a887a3e40cceab83068942c14f2fe970d05f056 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 1a887a3

Verdict parsed from the review's SHA-scoped output markers for commit 1a887a3e40cceab83068942c14f2fe970d05f056.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 1a887a3e40cceab83068942c14f2fe970d05f056: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 1a887a3e40cceab83068942c14f2fe970d05f056 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 1a887a3

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 1a887a3e40cceab83068942c14f2fe970d05f056: <one-sentence reason>

@cruisercohen
cruisercohen force-pushed the fix/21852-roster-hidden-count branch 2 times, most recently from e528775 to d03567b Compare October 19, 2026 22:09
@cruisercohen cruisercohen changed the title fix(members): say how many crewmates the roster hides fix(members): say which crewmates the roster hides Oct 19, 2026
@cruisercohen

Copy link
Copy Markdown
Contributor Author

Screenshots for head d03567b (fixture render, roster column at rest): one hidden crewmate named, and four hidden crewmates counted. The same frames are in the description.

Roster at rest with this PR: two crewmates and a line naming the hidden one

Roster at rest with four hidden crewmates: the line gives their count

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running and removed readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention labels Oct 19, 2026
@cruisercohen

Copy link
Copy Markdown
Contributor Author

Updated screenshot for the next head (fixture render, ROSTER=many, roster column at rest): past three hidden crewmates the line names the first three in roster order and counts the rest.

Roster at rest with four hidden crewmates: the line names three and counts one more

@cruisercohen
cruisercohen force-pushed the fix/21852-roster-hidden-count branch from d03567b to b3cfd4e Compare October 19, 2026 22:43
@cruisercohen

Copy link
Copy Markdown
Contributor Author

Updated screenshot for the next head (fixture render, ROSTER=many, roster column at rest): the crewmates the user made are named first, so Recipe Box is named ahead of the package rows and one package row is counted.

Roster at rest with four hidden crewmates: Recipe Box and two package rows named, one more counted

@cruisercohen

cruisercohen commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor Author

Review findings on b3cfd4e, by finding. The fixes are on cb2c828.

Design Review, wrong rows named: fixed in cb2c828. Hidden crewmates the user made (origin Mine, or created on the dashboard) are named before package and app rows, then roster order. A new case covers a user-made crewmate that sorts last behind four package rows; it fails on b3cfd4e. The ROSTER=many frame in the description is re-captured on the new code: "Recipe Box, pkg-budget, pkg-notes and 1 more crewmate ...".

First Principles, HIDDEN_NAMES_MAX export: removed in cb2c828.

UX Review:

  1. Unnamed remainder: not changed here. With the user's own crewmates named first, the remainder is package and app rows. A control that lists them would switch off the listing rule, which is a roster mode of its own rather than part of this fix. The search still reaches each of them.
  2. Names as plain text: not changed. The line is a status note at the end of the list, under the search field it points to. Making each name a control that fills the search adds an interaction the roster has nowhere else, so it is left out of this fix.
  3. pkg- wrapping: not changed. The names are interpolated into one translated sentence, so a span per name needs a rich-text path through the catalog, and a no-wrap name could overflow the narrow column. The break after the hyphen is the browser default.
  4. "haven't used it yet": kept, so the line reads the same as the existing every-row-hidden line (all_hidden_hint, "Crewmates you haven't used yet are hidden"), which describes the same rule.

@cruisercohen
cruisercohen force-pushed the fix/21852-roster-hidden-count branch from affaafd to cb2c828 Compare October 19, 2026 23:19
@cruisercohen

Copy link
Copy Markdown
Contributor Author

Review notes on cb2c828, by finding.

First Principles, stale ROSTER=many row in the manual-verification table: fixed in the description. The row now reads "Recipe Box, pkg-budget, pkg-notes and 1 more crewmate ...", which is what the harness printed for this tree.

Design Review, members vs orderedMembers: they hold the same rows. orderedMembers (useMembersPageRosterView) is built from the same rosterRead.members that hiddenCount reads. It is either the sorted list of those rows or a kept order that is used only while it names exactly the same rows. shownMembers is rosterPopulation over that same list. So the hidden names and the count come from one set of rows. The case "counts every hidden crewmate past the three it names" checks both together.

UX Review:

  1. Header count vs the line: not changed. The count covers listed rows on purpose (feat: hide unused crewmate rows on the Crewmates page #15363, fix(members): one roster reading, and no New team door for now #17835), and this PR leaves it alone. The line is what says more exist, and it names them.
  2. "and 1 more" with no name: not changed. With the user's own crewmates named first, the unnamed rest are package and app rows. A control that lists them would switch off the listing rule, which is a roster mode of its own rather than part of this fix.
  3. A word for package rows: not changed. The line shows each row's label as the roster does, and the roster shows package rows by the same label.

@cruisercohen
cruisercohen marked this pull request as ready for review October 20, 2026 00:11
@cruisercohen
cruisercohen requested review from a team and KiroCrewCaptain as code owners October 20, 2026 00:11
@cruisercohen
cruisercohen enabled auto-merge (squash) October 20, 2026 00:11
@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 20, 2026
@iamwhatever

Copy link
Copy Markdown
Contributor

HOLD (AI review concern not fully answered) at cb2c828.

The code looks correct. hiddenCount and hiddenNames come from the same rows as the listed set, and the line only renders with no search, no filter and at least one listed row (useMembersPageRoster.ts L259-L278). CI is green, and the Design, First Principles, Opus and GPT items are answered.

What blocks approval is the UX Review CONCERNS on this head:

  1. The count and the line disagree. The lane suggested keeping the count and changing the line's copy to say the count covers only shown crewmates. Your reply explains why the count stays, which the lane agreed with, but does not say whether the copy should change. The PR's goal is that the count is not read as the whole crew, and the blind reader still read it that way.
  2. For "and N more", the lane also offered a copy-only option ("search by name to find the rest"). The reply rebuts only the list-control option.

What clears it: adopt the copy changes, or reply to each one with why the current wording is better.

— crew-pr-reviewer (r113)

@cruisercohen
cruisercohen force-pushed the fix/21852-roster-hidden-count branch from cb2c828 to 426e1be Compare October 20, 2026 02:29
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 20, 2026
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 20, 2026
The Crewmates roster lists only crewmates whose thread holds a message, starred ones and Mate, and its count covers the listed rows only. A crew of three read as "2 crewmates" and the third turned up only when a search matched it. The list now ends with one muted line naming the hidden crewmates by their row label (the first three, then a count of the rest) and saying the search finds them. The rule and the count are unchanged.
@cruisercohen
cruisercohen force-pushed the fix/21852-roster-hidden-count branch from 426e1be to 1a887a3 Compare October 20, 2026 03:58
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: passed Eligible automated validation passed for the current revision labels Oct 20, 2026
@cruisercohen

Copy link
Copy Markdown
Contributor Author

Reply to the HOLD at cb2c828 and the UX rounds since, on 1a887a3.

1. The count and the line disagree: adopted. The count stays: listed rows plus any row a search reaches, as before. Both forms of the line now say the hidden crewmates aren't counted above, in all 12 locales and en-XA:

  • "Recipe Box is hidden because you haven't used it yet, and isn't counted above. Search to find it."
  • "Recipe Box, pkg-budget, pkg-notes and 1 more crewmate you haven't used yet are hidden and aren't counted above."

2. "and N more": the copy option "search by name to find the rest" went in on 426e1be. The UX round on that head found it sends the reader to a search they cannot run for the crewmates the line does not name. So the counted form now asks for no action and only states that the rest exist and aren't counted. The named form keeps "Search to find it/them", because every name it gives can be searched. A control that lists the hidden rows would turn off the listing rule, which is a roster mode of its own and outside #21852.

UX round on 1a887a3:

  • "Isn't counted above" vs the count during a search: with a search typed, the line is gone and the header counts what the search reached. That is the header's existing behavior, the same as on main (the description's table, search "recipe": "3 crewmates" on both builds), and this PR leaves the count alone. While the line is shown, no search is typed and the statement holds. Dropping it would bring back the first finding: a count read as the whole crew.
  • The next step and raw ids in the counted form: no search step, for the reason in 2. pkg- names are not changed. The line shows each row's label as the roster shows it, and which crewmates are named stays as it is (yours first, then roster order). Renaming crewmate ids or changing which are named is outside ux(members): Nova Sky wasn't in the list at first (it said 2 crewmates); it only appeared after I sear… #21852.
  • Names as buttons: left out, as before. It adds an interaction the roster has nowhere else.

Tests: the three counted-form cases in MembersPage.filters.test.tsx now assert the whole line, so a search instruction appended to it fails them. npx vitest run src/pages/members src/i18n passes (94 files, 1,860 tests), as do npm run i18n:check and the build. The ROSTER=many still in the description was re-captured from fixtures on this head; the one-hidden still is byte-identical to the previous capture.

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 20, 2026

@iamwhatever iamwhatever left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

APPROVE at 1a887a3.

The r113 HOLD is cleared. Your reply on this head answers each UX Review item: the line now says the hidden crewmates are not counted above in both forms, the counted form drops the search step for crewmates it cannot name, and the remaining suggestions are declined with reasons that hold. The line is drawn only with no search typed and no filter narrowing, so the count during a search is main's existing behavior, not something this line contradicts.

The First Principles check also holds: hiddenCount and hiddenNames are derived from the same rows, because orderedMembers is a reordering of members (useMembersPageRoster.ts L250-L278). The tests assert whole lines, CI is green, and no sensitive files are touched.

— crew-pr-reviewer (r115)

@iamwhatever
iamwhatever merged commit 3a18b2a into main Oct 20, 2026
152 of 161 checks passed
@iamwhatever
iamwhatever deleted the fix/21852-roster-hidden-count branch October 20, 2026 14:57
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

ux(members): Nova Sky wasn't in the list at first (it said 2 crewmates); it only appeared after I sear…

2 participants