Skip to content

feat: welcome users with Mate, a first crewmate they get to name - #18317

Merged
buluoray merged 1 commit into
feat/guide-platformfrom
feat/first-crewmate-welcome
Oct 9, 2026
Merged

buluoray merged 1 commit into
feat/guide-platformfrom
feat/first-crewmate-welcome

Conversation

@buluoray

@buluoray buluoray commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Give a new user one ready crewmate, Mate, that opens the conversation itself, explains what crewmates are and lets the user give it a name of their own, and have every crewmate the user creates greet them with its goal, so nobody starts in front of an empty chat.

Stacked on #18301 (crewmate guides and change cards); review that one first. The diff here is only the commit on top of it. With it, any crewmate can find controls and prepare changes, but an install still starts with no crewmate that talks first: the user has to know what to ask, and what a crewmate is, before anything happens. A crewmate the user creates on the dashboard already greets with its goal through the create flow's seeded first turn on main; this PR keeps that and adds none of its own.

Why it matters

The first minutes decide whether someone keeps using Kiro Crew. A crewmate that is already there, speaks first, and explains in plain words what it and other crewmates can do turns the guide and card tools from the platform PR into something a newcomer actually meets. Naming it is the first thing the user does with a crewmate, and seeing the name change at once shows that crewmates are theirs to shape.

Not a goal

  • A special or protected assistant. Mate is an ordinary crewmate: it can be renamed, rebound or deleted like any other, and a deleted one is never re-created.
  • Releasing the Crewmates page. It stays behind its Feature Preview; Mate is created in the background either way, but nothing about it is shown until the preview is on and the page is opened.
  • Changing where a new user lands when the app opens; that stays as on main.
  • Greeting on every open. Mate's welcome is owed once and never repeats.
  • A second greeting for crewmates the user creates: main's create-flow greeting stays the only one.
  • Rerouting error hand-offs or finished guides to Mate: they behave as on feat: let any crewmate guide users to the control on their screen #18301.
  • Turning the default member into Mate: it stays the shared fallback crew, its template and conversations untouched.

What changed (motivation → approach → change)

Approach: reuse the ordinary crewmate create path, the member event log the page already folds, and the platform guide tools; add a once-only pre-create, a once-only hidden first turn for Mate, and one narrow self-rename tool.

  • Mate (agent_materialization/first_crewmate.py): every install, fresh or upgraded, creates one ordinary crewmate once (key mate, shown as Mate while it has no display name), bound to the ordinary crewmate template with its own private memory store. A marker records that it ran and whether it wrote the row, so deleting Mate never brings it back, and a crewmate the user already keeps under the key mate is left alone, memory binding included.
  • Mate's first-turn welcome (dashboard/mate_welcome.py, website/src/pages/members/useFirstGreeting.ts): the first time Mate's own thread opens empty, the gateway starts one hidden kickoff turn. Only the first-crewmate step owes it; the claim is an O_CREAT|O_EXCL marker in the member's directory, so two tabs, a reload or a retry never greet twice. The welcome explains in everyday words what a crewmate is and invites the user to give Mate a name. A user who already has other crewmates hears Mate introduce itself as one more teammate, not as their first.
  • Live rename (kirocrew-guide tool rename_self, POST /api/guide/agent/rename, members.rename_member_display): when the user answers with a name, Mate renames itself. The member renamed is the caller's own pinned thread, never an argument; only a turn sent in the dashboard may call it, like every agent-half guide route; the name is cleaned and checked against MEMBER_NAME_MAX_CHARS, and the default crew is refused. The rename is a member/config event, so the open chat header and the roster show the new name at once, with no reload and no confirm card.
  • Crewmates page (MembersPage.tsx, rosterFilter.ts): with the preview on, a bare visit opens Mate while no message has been exchanged with it, ahead of a remembered crewmate, as an early rule of its own; after that the page lands as before. While that Mate is still owed its first visit, Meet CrewMates does not open on its own, neither at the end of the tour nor on a Crewmates page visit, so it never covers Mate's welcome; once Mate has history, or was deleted, the flow opens as on feat: let any crewmate guide users to the control on their screen #18301 (useMeetCrewmatesGate.ts, reading pendingMate from lib/assistantMember.ts). Mate is listed from the start; every other row keeps feat: let any crewmate guide users to the control on their screen #18301's listing rule, including the rule for rows from older gateways.
  • Mate's chat (crewmateBubbles.ts, useCarryHeldAnswer.ts, GuideMateNote.tsx): folds the narration written before a tool call, and a guide note offered from Mate's thread names Mate while it has no display name.
  • Docs: crew mode, memory, MCP architecture, learn-cron-dashboard, feature map. Two new strings in all 12 locales.

Moving parts the bullets above do not name:

Area What it does
agent.py, agent_files.py, agent_materialization/__init__.py Call the once-only Mate create at startup; the mate key and its default display name.
src/kiro_crew/members.py, memory_stores.py rename_member_display and the welcome-owed marker; at startup, give the Mate row the first-crewmate step wrote, if it still points at Global memory, its own private store.
mcp_guide.py, validation.py, config/defaults.json The rename_self tool, its argument schema and its exact-name grant.
src/kiro_crew/docs Regenerated ui-index.generated.json; agent-spec-fields.md names the first crewmate.
store/chatSlice.ts, store/chat/runState.ts, website/src/api Mark a gateway-started welcome turn as running in the tab; clients for the welcome and rename calls.
app-sdk/messageRenderers.tsx, pages/chat/*, components/AgentSelector.tsx, pages/KiroCrewAgentsPage.tsx, pages/members/CrewNotesTab.tsx Mate's label where a crewmate is named; the chat identity carries whether it is Mate.
website/src/cards, guide/findByName.ts, uiLocations/areas/members.ts Card text names the agent generically; comments.
hooks/useMeetCrewmatesGate.ts hasNoCrewmates counts Mate as a crewmate; both automatic triggers wait for the roster while the Crewmates preview is on and stay shut for the session if Mate is still owed its first visit.
website/scripts The preview-gateway walkthrough scripts used for the screenshots.
website/src/i18n Mate's default name and the greeting-failed line in all 12 locales.

Backwards compatibility

Compatible: existing installs keep their members and conversations. An upgraded install gains Mate once, as an added member created through the existing create path, with a welcome worded for a user who already has crewmates; nothing existing is renamed, rebound or reserved, and a row the user already keeps under the key mate is left as it is.

Reader: website/src/hooks/useMeetCrewmatesGate.ts:useMeetCrewmatesGate -- first run and crew page -- with the Crewmates preview on, neither the tour-end nor the page-entry auto-open fires while Mate exists with no chat history, and one held that way stays held for the session; with Mate chatted with, deleted, the roster unreadable, or the preview off, both fire exactly as on #18301 (useMeetCrewmatesGate.test "the tour ending does not open the flow while Mate exists and has never been chatted with", "the tour ending still opens the flow once Mate has history", "the tour ending still opens the flow when Mate was deleted", "a Crewmates page entry does not open the flow while Mate has never been chatted with")

Tests

  • Backend: test_first_crewmate_ordinary.py (created once, ordinary, deletable, never re-created), test_welcome_owed.py (only the first-crewmate step owes a welcome, on an upgrade with other crewmates too; a dashboard or CLI create owes none), test_mate_welcome.py (kickoff turn, refusals, the first-crewmate wording alone and the one-more-teammate wording beside other crewmates), test_mate_rename_self.py (renames only the caller, refuses a channel or non-dashboard turn, a non-crewmate thread and the default crew, refuses a turn the user did not send, cleans the name; the welcome asks for Mate's name in plain words and states principles rather than prohibitions), test_assistant_agent.py (a user's own mate row keeps its Global binding at the next start; a created Mate left on Global is moved), updated test_change_cards.py, guide-route and composition-contract tests.
  • Frontend: useFirstGreeting.test.tsx, ChatPane.mateChat.test.tsx, KiroCrewAgentsPage.mateDisplayName.test.tsx, useCarryHeldAnswer.test.ts, assistantMember.test.ts, rosterFilter.test.ts (Mate listed, feat: let any crewmate guide users to the control on their screen #18301's rule otherwise), Members page cases (never-chatted Mate beats a remembered crewmate, a chatted one does not, preview off lands and greets nothing, only Mate's thread asks for a welcome, Meet CrewMates not announced over Mate's landing and announced once Mate has history), useMeetCrewmatesGate.test.tsx (tour-end and page-entry both held while Mate is owed its first visit, tour-end still held after Mate's history arrives in the same session; both fire once Mate has history or is deleted; nothing held with the preview off).
  • Each new guard was broken once to confirm a named test turns red.

Manual verification

Run on a preview gateway built from an earlier head of this branch, fresh data home, Crewmates preview on, real model turns: Mate opened on the first Crewmates visit and greeted first; "Call you Pebble" renamed it in the header, composer, switcher and roster with no reload; with the preview off there was no Crewmates rail entry and no welcome. Not re-run on this head: the welcome wording for a user who already has crewmates, and Meet CrewMates staying shut over Mate's landing (both covered by the tests above). Gates on this head: black, isort, flake8, mypy, tsc, npm run build, the i18n checks with I18N_BASE_REF=origin/main, and the related pytest and vitest sets. Not exercised: Windows and macOS hosts.

Screenshots / video

All frames are real model turns on a preview gateway built from earlier heads of this branch. Since then the welcome wording changed (plain words, a variant for a user who already has crewmates), so greeting text differs from a run on this head.

  • Mate's first welcome on the first Crewmates visit
  • After "Call you Pebble": header and composer renamed live
  • Phone width
  • Roster lists Pebble; the never-used default row is not listed
  • A created crewmate, Scribe, greets with its goal (taken with a kickoff this head removed; on this head the create flow's seeded first turn from main does it)
  • Preview off: /members typed directly shows no welcome and no Crewmates rail entry
  • Preview off in Settings

Related Issues

Part of #18315

Stacked on #18301; together they replace #15326.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • 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

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

1-first-open-mate-welcome-v2

2a-pebble-header-live-v2

5-phone-mate-welcome-v2

2c-pebble-roster-column

3b-scribe-short-welcome

4a-preview-off-members-typed-url

4b-preview-off-settings

@buluoray
buluoray requested review from a team and KiroCrewCaptain as code owners October 9, 2026 03:25
@buluoray
buluoray requested review from bolichen97 and removed request for a team October 9, 2026 03:25
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 9, 2026
@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — 🟡 CONCERNS

Design-level review of 46db2368bef1e875a7050ba8ca29139207c17ce9 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

The Mate row appears outside the Crewmates preview. Upgraded users with the preview off will see an unexplained new crew, which contradicts the description.

Watch

  • Mate is visible even with the preview off. The description says "nothing about it is shown until the preview is on and the page is opened." But create_first_crewmate_once() runs in rebuild_agent_config on every install, including upgrades, and writes an ordinary agents.mate row into config.json. AgentSelector and KiroCrewAgentsPage read that same roster with no preview gate. This PR even gives Mate a localized label in both (crewDisplayName → assistantWelcome.default_name). So an upgraded user who never turned on the preview finds a new "Mate" crew on the Crews page and in every agent picker (chat, job form, channel page). They never asked for it, and its welcome never fires for them. Deleting it fixes it, since it is never re-created, so the harm is reversible. Still, it puts a new thing in front of every upgraded user, and the description says it doesn't.
    Clears when: either only create Mate (or only list it) while the Crewmates preview is on, or change the description and Backwards compatibility to say that upgraded users see Mate on the Crews page and in agent pickers whether or not the preview is on, with a screenshot of that state.

Suggestions

  • is_assistant_member(name, entry) throws away entry and matches on the key alone. A user's own crewmate kept under mate therefore gets Mate's label and fallback in the UI as well. Gate the UI label on first_crewmate_was_created, which the backend already uses for memory, so the two halves agree.

[DESIGN-REVIEWED] 46db236

Description read

[DESCRIPTION-READ] e470f26b132d98dd8fe82d1d04ffd6635fbe8110a2175eaaa2df3a1838ffbd1d

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 9, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🔴 BLOCK

UX-level review of 46db2368bef1e875a7050ba8ca29139207c17ce9 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: BLOCK

cannot evaluate: missing the new "first greeting could not be started" notice and its Retry button, which no screenshot shows.

Evidence gaps

  • member-greeting-error-row (MembersPage.tsx) appears in none of shot-01 to shot-07. That row is ErrorNotice with "The first greeting could not be started. You can still write your message below." plus a Retry button. Mate's welcome is the PR's main path, and this row is what a new user sees when the welcome fails. To close the gap, attach a still of Mate's empty thread with the greeting request failing, with the notice and Retry visible.

Watch

  • Mate's welcome asks for the wrong name. In shot-04 the welcome invites "give me a name", then ends "What should I call you, and what's the first thing…". That last question asks for the user's own name, and it is the line a person answers first. Then rename_self never fires, or renames Mate to the user's name. This hits every new user once, and the first impression is what the PR is for. Fix: in mate_welcome.py's kickoff prompt, end on "What would you like to call me?"
  • Mate appears with the Crewmates preview off. In shot-06 the preview is off (shot-07), yet the chat page shows a "Mate · Idle" pill and "Message Mate…". The PR says "nothing about it is shown until the preview is on." The blind reader read it as "Mate again, even though I renamed it Pebble". This affects every user without the preview, and it persists. Fix: hide Mate from the chat-page agent picker while the preview is off. If showing it there is intended, add a stated reason.

Suggestions

  • After a rename, the welcome still opens "Hi. I'm Mate." under the "Pebble" header (shot-02: "feels a bit odd to see both names at once"). Have Mate's first reply after the rename confirm it plainly, for example "From now on I'm Pebble."

[UX-REVIEWED] 46db236

Description read

[DESCRIPTION-READ] 2cd80b3d7a1bf18c71d0f44f030b6e4569d8ac69f4430ba602b167e244896a50

That sha256 names everything this verdict read: the PR description, plus the 10 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 9, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — 🔴 BLOCK

Premise-level review of 46db2368bef1e875a7050ba8ca29139207c17ce9 — 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: BLOCK

product-shape change without accepted RFC: every install now gets a crewmate it never asked for, plus a model-written welcome that rfc-crewmate-greeting §6 rejected.

Not justified as shipped

  • Item 1 — product-shape: "every install, fresh or upgraded, creates one ordinary crewmate once"; no RFC on the base covers this.
  • Item 2 — product-shape: rfc-crewmate-greeting (accepted) §6 rejected this: "A model-written welcome turn. Rejected: it spends a model call…"
  • Item 4 — product-shape: this overrides rfc-crewmates-launch §02's landing rule, "the remembered crewmate first, activity second".
  • Item 5 — product-shape: this holds back §08's decided Meet CrewMates first run, whose step 2 is "Name your first crewmate".
  • Item 6 — product-shape: §02's decided "No crewmates yet" empty state can no longer be reached.

What this change ships

Inventory (10 items) — 3 justified

Intent: give a newcomer a crewmate that talks first and gets named by them. This is an ADDITION.

  1. Fresh and upgraded installs gain a crewmate named Mate — product-shape
  2. Mate opens with a hidden, model-written welcome turn — product-shape
  3. Mate renames itself through the new rename_self tool or route — justified
  4. A bare Crewmates visit opens a never-chatted Mate ahead of the remembered crewmate — product-shape
  5. Meet CrewMates no longer opens on its own while Mate is unvisited — product-shape
  6. The empty state on a fresh install is gone — product-shape
  7. Mate's upgraded row is moved off Global memory onto a private store — rides along
  8. Narration written before a tool call is folded in crewmate chats — rides along
  9. Guide notes and card text name Mate, or the agent generically — justified
  10. Mate's default label shows in the selector, the agents page and notes — justified

Blockers

  • product-shape change without accepted RFC: auto-created Mate, its model-written first turn, and the changed landing and onboarding
    Items 1, 2, 4, 5 and 6 change what every install gets without anyone choosing it. I grepped the base for first crewmate|\bMate\b|pre-?creat under docs/request-for-change and docs/decisions. The only match on this shape is the §08 flow, which this PR holds back. The accepted RFCs also contradict items 2, 4 and 6.
    Clears when: an RFC covering a pre-created, self-greeting first crewmate and the changed landing and onboarding is merged on the base as non-draft, followed by a push or rebase; or a maintainer posts /ai-review override first-principles 46db2368bef1e875a7050ba8ca29139207c17ce9: <decision>.

[FIRST-PRINCIPLES-REVIEWED] 46db236

Description read

[DESCRIPTION-READ] 0ba610ef04ec59f87bec046707ff008b04d37fbb4f1960edc965c695d073ed10

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 9, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — 🔴 changes requested (blocking)

GPT 6.1 found at least one blocking issue that must be resolved before merging 46db2368bef1e875a7050ba8ca29139207c17ce9. 1 of 1 blocking finding(s) are security-class and were withheld from adjudication, so the blocking verdict stands.

This comment is updated in place on each push.

BLOCKING -- website/scripts/verify-crewmate-assistant.mjs:92 -- Walkthrough aborts waiting for a nonexistent welcome card
const welcome = page.getByTestId('assistant-welcome') / await welcome.waitFor()
Running the script with the current SPA build -> /members -> no production component renders assistant-welcome -> uncaught Playwright TimeoutError before the walkthrough’s checks or screenshots.
Anchor: residual/crash-data-loss-corruption
Fix: Use the mate fixture and thread, stub its greeting endpoint, and replace the static-card waits and assertions with the actual transcript greeting.
[BLOCK-MERGE] 46db236
[GPT-REVIEWED] 46db236

Adjudication (Opus 5.5) — is blocking on each finding proportionate?

F1 is real: no production component renders assistant-welcome, so website/scripts/verify-crewmate-assistant.mjs:92 throws a Playwright TimeoutError every time the script runs. It does not cause the crash, data loss or corruption that its residual/crash-data-loss-corruption anchor claims.

  • Where the script runs: it is a manual walkthrough. You run it by hand and give it an output directory (:10). Nothing in website/package.json, .github or docs calls it, and it runs only against stubbed routes (:31, :79-88).

  • Where assistant-welcome appears: only in tests, and those check that it is absent (ChatPane.mateChat.test.tsx:104, MembersPage.test.tsx:219).

  • What happens when it fails: a developer running the script sees the error immediately, and no user or gateway data is touched. The harm is LOW, not unbounded.

  • What it costs to fix: it is a test-script-only fix, but the shipped app does not rely on this script, so a reviewer can reasonably clear it with an override.

    [ADJUDICATION] 46db236 total=0 uphold=0 downgrade=0
    [GPT-ADJUDICATED] 46db236

    [ADJUDICATION-FENCED] 46db236 fenced=1 flagged=1
    FLAG F1 website/scripts/verify-crewmate-assistant.mjs:92 -- The defect is a loud, deterministic TimeoutError in a manual dev-only walkthrough against stubbed APIs, not wired into CI or package scripts (no references found), with no reachable path to crash, data loss or corruption in the shipped product, so the unbounded-harm anchor does not hold and a writer may reasonably override.
    [GPT-ADJUDICATED-FENCED] 46db236

🏷️ Fenced finding(s) machine-flagged as likely edge case

The security fence keeps these findings blocking regardless of adjudication; the only clearance path is a human override. The annotate-only pass judged the condition combination each one requires extreme, and pre-drafted the override rationale below. A repository writer must independently verify a rationale before posting it — it is machine-authored, and a wrong override on a security-class finding ships exactly the class the fence exists to stop.

  • F1 website/scripts/verify-crewmate-assistant.mjs:92 — The defect is a loud, deterministic TimeoutError in a manual dev-only walkthrough against stubbed APIs, not wired into CI or package scripts (no references found), with no reachable path to crash, data loss or corruption in the shipped product, so the unbounded-harm anchor does not hold and a writer may reasonably override.

    /ai-review override gpt 46db2368bef1e875a7050ba8ca29139207c17ce9: The defect is a loud, deterministic TimeoutError in a manual dev-only walkthrough against stubbed APIs, not wired into CI or package scripts (no references found), with no reachable path to crash, data loss or corruption in the shipped product, so the unbounded-harm anchor does not hold and a writer may reasonably override.
    

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 46db2368bef1e875a7050ba8ca29139207c17ce9: <one-sentence reason>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 9, 2026
@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 46db2368bef1e875a7050ba8ca29139207c17ce9 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 46db236

Verdict parsed from the review's SHA-scoped output markers for commit 46db2368bef1e875a7050ba8ca29139207c17ce9.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 46db2368bef1e875a7050ba8ca29139207c17ce9: <one-sentence reason>

buluoray added a commit that referenced this pull request Oct 9, 2026
Record the product owner's 2026-10-08 decision, taken after the team
review, so the First Principles lane can read it from the base branch
when #18301 and #18317 change default mounts and first-visit landing.

Part 1: one read-or-offer MCP server, kirocrew-guide, mounted for the
kirocrew template, its crewmates, worker and dashboard sessions. Guides
highlight the real control and change cards apply only on the user's
press, from a dashboard turn the user sent; trust-root areas are
answered in words only.

Part 2: Kiro Crew creates one ordinary crewmate, Mate, once. It opens
first on the Crewmates page while never chatted with, writes its own
welcome, and renames itself through rename_self. default stays the
shared fallback crew.

The document amends sections of rfc-crewmates-launch,
rfc-crewmate-greeting and rfc-crewmates-chatted-list rather than
editing those accepted documents. The index gains its row and audit
note.
@buluoray
buluoray force-pushed the feat/guide-platform branch from 3637e85 to 639dc8c Compare October 9, 2026 04:24
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 9, 2026
Every install, fresh or upgraded, creates one ordinary crewmate once,
shown as Mate while it has no display name. It is bound to the ordinary
crewmate template, gets its own private memory store, and can be
renamed, rebound or deleted like any other; a deleted one is never
re-created, and a crewmate the user already keeps under the key mate is
left as it is.

The first time Mate's own thread opens empty, it speaks first with one
hidden kickoff turn. Only the first-crewmate step owes that welcome,
recorded in the member's directory, and the claim is an O_CREAT|O_EXCL
marker so two tabs, a reload or a retry never greet twice. The welcome
explains in everyday words what a crewmate is and invites the user to
give Mate a name; a user who already has crewmates hears Mate introduce
itself as one more teammate. A crewmate the user creates on the
dashboard keeps greeting through the create flow's seeded first turn.

When the user answers with a name, Mate calls the new kirocrew-guide
tool rename_self. It renames only the calling crewmate, derived from
the caller's own pinned thread, only from a turn sent in the dashboard,
and the new label reaches the open chat and the roster through the
member/config projection with no reload or confirm card.

With the Crewmates preview on, a bare visit to the Crewmates page opens
Mate while no message has been exchanged with it, ahead of a remembered
crewmate, and that visit does not also announce Meet CrewMates; after
that the page lands and announces as before. With the preview off
nothing about Mate is shown: no landing and no welcome.
@buluoray
buluoray force-pushed the feat/first-crewmate-welcome branch from c3ea98a to 46db236 Compare October 9, 2026 05:26
@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=266306e752a8 Mate opens ahead of the remembered crewmate; "you are their first crewmate" -- fixed (wording); landing kept by owner decision

The never-chatted Mate landing first is the owner's decision, behind the Crewmates preview, once per page mount, and gone as soon as Mate's thread has history.
The "first crewmate" sentence is now conditional: a user with other crewmates hears "one more teammate" (46db236); test_mate_welcome.py covers both variants.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=48878a145c18 A private V2 memory store is provisioned for it -- rebutted (owner decision); narrowed

A private store is what every created crewmate gets; Mate is created through that same path, so it is not new on-disk state of a new kind.
Narrowed in 46db236: the start-of-process repair moves only the row the first-crewmate step wrote (first_crewmate_was_created), never a user's own mate row.
Test: test_assistant_agent.py::test_an_existing_key_is_left_alone[base] (reddens when the gate is removed).

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=60e225441937 Upgrade path contradicts the stated scope -- fixed (description and wording; upgrade creation kept by owner decision)

The owner decided Mate is created once for every install, fresh or upgraded; the description now says so instead of "a fresh install".
The welcome kickoff now reads the roster: with other crewmates it uses FIRST_WELCOME_WITH_CREW ("one more teammate on their crew"), never "first crewmate" (46db236).
Tests: test_welcome_owed.py::test_an_upgrade_with_other_crewmates_still_gets_mate_once, test_mate_welcome.py::test_mate_beside_other_crewmates_is_welcomed_as_one_more, test_mate_rename_self.py::test_mates_welcome_fits_a_user_who_already_has_crewmates.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=774d38c894f4 A mate row is written into their config.json -- rebutted (owner decision)

Adding Mate to every install once is the owner's product decision; the row is ordinary, deletable and never re-created (one-time marker).
A row the user already keeps under mate is left untouched, and the marker now records that the step skipped it (46db236). Stated in the description's Backwards compatibility.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=780253bca169 A WELCOME_FIRST record is written -- fixed

The record now exists only for the Mate row the first-crewmate step wrote, and the welcome it triggers fits a user who already has crewmates (46db236).
The separate short-welcome kind is gone: dashboard and CLI creates record nothing (test_welcome_owed.py).

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=58ab3ac8329f create_first_crewmate_once has no fresh_install check -- rebutted (owner decision); wording fixed

Mate for upgrades too is the owner's decision; the comment and spec now say every install, fresh or upgraded.
The welcome is roster-aware, so a long-time user hears "one more teammate", not "first crewmate" (46db236); test_mate_welcome.py::test_mate_beside_other_crewmates_is_welcomed_as_one_more.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=5ccf98a58a41 assistant_needs_private_memory moves a user's own mate row -- fixed

assistant_needs_private_memory now also requires first_crewmate_was_created(): the one-time marker records created only when the step wrote the row (46db236).
A user-created mate row on Global keeps its V1 binding at every start. Tests: test_assistant_agent.py::test_an_existing_key_is_left_alone (reddens with the gate removed) and ::test_a_created_first_crewmate_left_on_global_is_moved_at_next_start.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=1421fb00e4d8 Item 1 -- upgrades get Mate too -- rebutted (owner decision); description fixed

Creating Mate once for every install, upgrades included, is the owner's decision; the description no longer says "a fresh install" only.
The welcome for an upgraded user with crewmates no longer calls Mate their first crewmate (46db236).
Test: test_welcome_owed.py::test_an_upgrade_with_other_crewmates_still_gets_mate_once.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=424f1495e53e Item 5 -- error hand-offs and finished guides reroute to Mate -- fixed (subtracted)

Deleted website/src/lib/mateHandoff.ts and its test, the Mate targeting in errorReport.ts, AskAgentButton, ErrorNotice, ErrorBoundary, notificationSheet, SkillsTab, SecurityPanel, code-review-sage and App.tsx, the "Back to {name}" guide exit and its strings (46db236).
Those files are now byte-identical to #18301; their tests were restored to #18301's.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=5d626c6442f5 Item 7 -- older-gateway row rule changed -- fixed

listedByDefault restores #18301's rule for a row without last_chat_ts, the default crew included (rfc-crewmates-chatted-list section 5); only the Mate line is added (46db236).
rosterFilter.test.ts and MembersPage.filters.test.tsx are back to #18301's expectations for those rows.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=e10e075a9f87 Item 3 -- duplicate of greeting_seed -- fixed

Verified: main's MembersPage seeds greeting_seed/greeting_seed_with_job after a dashboard create. B's short CREWMATE_WELCOME path is deleted (46db236): mark_welcome_owed is gone from crew_records.py and cli_commands.py, and the page asks /greet only on Mate's thread.
One greeting mechanism per create path: dashboard create -> greeting_seed; first-crewmate step -> Mate's kickoff.
Tests: test_welcome_owed.py (dashboard and CLI creates owe nothing), MembersPage.test "only Mate's thread asks for a first welcome".

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=5f825e6b5f50 New autouse fixtures use the shared patch stack -- fixed

_quiet_sel and _events in test_mate_rename_self.py, and _owner_caller in test_first_crewmate_ordinary.py and test_welcome_owed.py, now request _floor_monkeypatch (46db236).
Same change clears the three semgrep.kirocrew.test-autouse-shared-monkeypatch hits.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=1104f1e0d0d7 Two names stay on screen after a rename -- rebutted

The "Hi. I'm Mate." bubble is history: it is what Mate said before the user renamed it, and rewriting past turns would misreport the transcript.
Every live surface carries the current name at once (header, roster, switcher, composer) through the member/config projection; screenshot 2a-pebble-header-live-v2 shows it.
The rename reply itself is the model's own words by owner decision; the kickoff asks it to use the new name from then on.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=32a110fb5078 The welcome asks for the user's name instead of Mate's -- fixed (earlier round)

The kickoff already says "the name in question is yours, since you are the one being named" and the welcome ends on what the user would like to call Mate.
Evidence: the v2 screenshot 1-first-open-mate-welcome-v2 in the description shows the closing question asking for Mate's name; test_mate_rename_self.py::test_mates_first_welcome_asks_for_a_name_for_mate pins the wording.
Fixed product text is not adopted: the owner decided the model writes the welcome.

@buluoray

buluoray commented Oct 9, 2026

Copy link
Copy Markdown
Contributor Author
  • span=7aa32b88fb1a The welcome uses words a newcomer does not know -- fixed

In 46db236 the kickoff drops the session explanation, asks for everyday words with no developer terms, and explains a crewmate as a teammate who remembers what you work on together.
Its offer stays the three plain items (set up Kiro Crew, sort out what is not working, show around); no PR/commit/CR examples.
Test: test_mate_rename_self.py::test_mates_first_welcome_asks_for_a_name_for_mate asserts "no developer terms" and no "session".

@github-actions github-actions Bot removed the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 9, 2026
@buluoray
buluoray merged commit 357d004 into feat/guide-platform Oct 9, 2026
22 of 27 checks passed
@buluoray
buluoray deleted the feat/first-crewmate-welcome branch October 9, 2026 05:28
@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Oct 9, 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.

1 participant