Skip to content

feat(agents): manage agent templates from an Agent templates tab - #12481

Merged
chenmingwei23 merged 1 commit into
mainfrom
feat/agent-templates-page
Sep 22, 2026
Merged

chenmingwei23 merged 1 commit into
mainfrom
feat/agent-templates-page

Conversation

@CrysisDeu

@CrysisDeu CrysisDeu commented Sep 21, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

The definitions a chat or a crewmate runs — the agent templates under
~/.kiro/agents/ — have no management surface of their own. The crew editor's
Template pane edits one crew's private copy of a template (blueprint
semantics), so a template no crew is bound to, which is every freshly installed
package agent, has nowhere to be seen. The dashboard can change only a template's
model and skills; prompt, tools and description say "edit the file". Creating a
template means publishing some crew's private copy; deleting one means removing
the file by hand and hoping nothing still points at it.

Why it matters

#12224 made templates first-class execution choices in chat without enrolling
them as crewmates. That only pays off if a user can also see, edit, create and
retire templates in the app: otherwise "try this template" leads to a file the
UI cannot show. And a delete done on disk while a crewmate, a schedule or a
private copy still names the template leaves sessions that silently start on
the default agent — the exact failure a guard in the product prevents.

What changed (motivation → approach → change)

Goal: manage the shared templates themselves, beside the pane that manages
private copies, without changing what either a crewmate or a private copy is.

Approach: one read-only roster endpoint that returns editability and references
with each row, so the tab can say up front what it can edit and what a delete
would hit; the existing detail PATCH widened to the definition keys, gated by a
server-side read-only rule; and the template delete rebuilt from the handler
this repo retired with the old standalone page (lock discipline and 409 guard
kept, guard widened). MCP servers stay read-only on purpose — a server is a
capability grant with its own admission path.

  • src/kiro_crew/dashboard/handlers/agent_templates.py (+ routes/agents.py):
    GET /api/agents/templates — every global discovery row (package-controlled
    strings through _roster_mask, as the sibling rosters do) plus read_only
    (null, or package / runtime / markdown / private_copy; a row with
    no spec file beneath the agents directory — an edition catalog row with an
    empty or foreign filename — is runtime, never offered an action that
    could only 404) and used_by
    (crews, the default agent, schedules by what they dispatch — a multi-entry
    agent_sequence over a dormant agent_id, else the captured
    execution_context.template_id — where the dispatcher reads the template a
    job runs, and the ONLY place a schedule created from a template chat names
    it — falling back to agent_id for a legacy record, script jobs over
    neither; that rule now lives once, in src/kiro_crew/cron.py::dispatched_agents_from_disk(loadable_only=…),
    which the guard calls with True (only records the scheduler could build
    pin a template) and kirocrew doctor's job_agent_names_from_disk wraps
    best-effort with False, so the two readers cannot drift; the same module
    exposes its store lock as cron_store_lock (one implementation, _file_lock
    delegates) for the delete guard to hold across walk + rename — chat
    folders by default_agent, webhook tokens by agent, and private copies by
    forked_from; an open chat that picked the template is deliberately not a
    holder — the module docstring records why); POST /api/agents/templates — create blank or as a
    lineage-free copy of an installed template, refusing a name an installed spec
    or a crew binding resolves (409 ambiguous_template_name when two files
    declare the name); DELETE /api/agents/detail/{name} — refused as
    template_read_only for the four kinds above and as template_referenced
    (with the list, masked like used_by) while anything names it; a row whose
    filename is not a plain basename resolving to a file directly under the
    agents directory is a 404, never an unlink; a name more than one file reaches
    (by declared name or stem) is 409 ambiguous_template_name, never the first
    hit — the matches come from the raw spec-file walk the PATCH uses, not the
    deduplicated roster, so foo.json declaring bar beside bar.json, or two
    files both declaring bar (one roster row), refuse instead of unlinking
    whichever the scan listed first. Under the
    spec lock the FILE decides: it is re-read and classified from its own contents
    (the way discovery and the PATCH classify a file), the requested name must be
    one it answers to (declared name or stem, else 404), the ambiguity check is
    re-run so the name still reaches exactly this file (a claimant landing after
    the probe is 409, not an unlink), it must be deletable from
    this tab (else 409 template_read_only with the file's reason), and its own
    declared name joins the reference guard's aliases — an edition row saying
    rogue over victim.json, or calling a package file plain, unlinks nothing. The reference check and the unlink
    are one off-loop critical section under the locks whose writers this handler
    can serialize with — the folder store held across it (a new snapshot-handing
    state.hold_folders, the sibling of mutate_folders' own held write), the
    config.json advisory lock and the config.local.json overlay lock nested
    inside it (cross-process, writing nothing; the effective merged config and
    the fork sidecar are read under them), and the spec lock around the unlink
    — the shape _unlink_copy_unless_referenced established; both mutations
    (create's write, delete's section) run through drained_to_thread, so a
    request cancelled mid-write drains the worker before the locks unwind
    instead of committing a file nobody audits. Webhook tokens are read inside
    that section and their writers now commit under the same spec lock
    (handlers/hooks.py::_commit_pinned_token re-verifies the agent inside
    agents_spec_lock before a mint or a re-pin), so a token cannot land pointing
    at a template being removed. Schedules are pinned with the scheduler's own
    store lock: cron.cron_store_lock (the one implementation, which
    CronService._file_lock now delegates to) is held from the dispatch-reference
    walk through the rename, innermost in the delete's lock order, so a schedule
    cannot be saved against the template between "nothing dispatches this" and
    the file going; a holder that keeps it past the bounded wait is 503 schedule_store_busy, nothing unlinked. A writer that bypasses the lock (a
    hand edit) is still caught: once the file is gone the store is re-read and a
    schedule that landed anyway is named in a WARNING and a SEL row (outcome
    dangling_reference). A crons.json that is present but unreadable fails the
    delete closed (503 schedule_store_unreadable, nothing unlinked) — before the
    unlink; after it, a store that fails to read is a WARNING, never a refusal
    about a file that is gone. The unlink is a rename to a one-deep tombstone
    (<name>.json.bak.<epoch>, the janitor's own backup shape, invisible to
    discovery; the rename comes first and the older grave goes only after it
    succeeded, so a refused rename keeps both the live file and the previous
    recovery copy; the sweep matches only a genuine grave, <name>.bak.<digits>,
    never a spec — foo.json.bak.5 is a legal template name), so the tab's one irreversible action is undoable by hand and
    the SEL row names where the file went. The create's
    locked check re-scans by declared name too, so a package file landing after the
    probe refuses instead of producing an ambiguous name. The default holder is
    agent.default_agent (a template name); the top-level default_agent is a
    crew alias and that crew is already counted as a crew. Create applies the
    same two-layer binding check, so a name bound only in the overlay is refused,
    and a duplicate re-resolves its source inside the spec lock — the name must
    reach exactly the file the probe chose, so a second claimant or a replacement
    landing after the probe is a 409, not a copy of a definition never validated —
    and re-reads it there (the fork/publish shape). Both mutations keep the
    dispatch snapshot current as well as the roster cache: a create publishes the
    new name (publish_materialized_agents) before scheduling the off-loop
    rescan, so "Chat with this template" resolves it on the next turn instead of
    a slot being stored against the default agent; a delete awaits
    refresh_materialized_agents off-loop before answering, so the deleted name
    is not bindable in a deferred-refresh window. A successful create or delete
    emits an operation-labelled SEL line beside the middleware's request-level
    record.
  • src/kiro_crew/dashboard/handlers/agents.py: the detail PATCH accepts
    description, prompt, tools, allowedTools (shape-checked, 400 invalid_definition, each tool entry capped in length) and refuses them on a
    read-only spec (409 template_read_only) and when two files claim the name
    (409 ambiguous_template_name, neither touched — for every PATCH key, and
    re-checked under the write lock). The read-only rule classifies the
    FILE the PATCH targets (name, declared name, fork sidecar), not the
    deduplicated roster's answer for its declared name, so a plain atlas.json
    beside a package SomePkg-atlas.json cannot let the package file through;
    model / skills keep their existing reach for the crew pane's private copies.
  • website/src/pages/overview/AgentTemplatesTab.tsx, wired as the Agent
    templates
    tab in website/src/pages/CapabilitiesPage.tsx: list grouped by
    origin via lib/templateSource.ts (Mine / Crewmate overrides / From packages
    / Built-in — a crewmate's private copy is a "crewmate override" in the UI,
    including the shared lib.templateSource badge the crew pane shows for the
    same file; Duplicate yields "a template of your own"), detail editor with one dirty draft
    (description, model, prompt, tools with auto-approval marks) saved through
    api.agentPatch with ONLY the keys the draft changed (a non-empty model
    is read by the server as an explicit pin that flips model_managed off, so a
    prompt-only save must not resend a managed spec's concrete model) — a
    background refetch never reseeds a dirty draft — skills
    through AgentSkillsEditor (saved on their own, under its own heading; each
    saved list is written into the detail cache before the refetch, so a second
    toggle in that window cannot PATCH the first away),
    resources and MCP servers read-only (string-valued fields only, since a spec
    is hand-editable; a rejected detail read shows its error rather than
    Loading). The header holds two controls: Chat with this template
    (createSlot with agent_kind: 'template') and an overflow menu with Enroll
    as crewmate (its row says what enrolling starts), Duplicate and Delete
    (reference dialog on refusal); a read-only template's banner carries the
    reason once with Duplicate to edit beside it (a private copy gets Open
    crewmate
    instead, and its reference row links to the crew's pane); a tool
    pill is name + worded state tag (the toggle, styled with the shared button
    recipe; the caption says a click switches it and that removing or toggling
    changes nothing until saved, and on a read-only template — inert tags — says
    "read-only here" instead of instructing a click) + divider-separated remove; the Add tool input
    shows an example placeholder; the enroll row says the crewmate appears under
    Agents; the save bar's buttons never wrap;
    resources are plain rows, not chips; the save bar says how many crewmates a
    save affects and carries both halves of the save-vs-restart model (new chats
    use a save at once; running chats after Apply & Restart), and a refused save is reported
    in that bar beside Save (inline ErrorNotice) rather than at the top of the
    scrolling pane; the glossary says in visible text that Apply & Restart only
    relaunches running sessions; a private copy's banner ends "open the crewmate
    to edit it", matching its button; the model badge's title spells out "auto
    (backend default)"; list badges read "Runs N crewmates" / "N crewmate
    overrides" (the group heading's word), both counts in the same muted pill as the model, and the
    usage line says "Runs N crewmates" in the badge's words, not "Runs as"; read-only reasons say outcomes, not mechanism; a delete leaves
    "Deleted “name”." in the bar's slot over the next row; the tab's own links (usage-line holders,
    Open crewmate, the refused-delete dialog) ask the shell's leave gate before
    navigating and beforeunload is armed while a draft is dirty; the glossary is
    term-led (Template / Chat / Crewmate, one line each) with the save-vs-restart
    sentence saying chats and history stay; the delete confirm states the blast
    radius, and is asked only when the row shows no holder — with holders in hand Delete opens the reference list directly, no confirm and no request; read-only banners lead with a bold two-word reason and a read-only
    prompt is a visibly locked block; the save bar wraps at 320px; Chat with this
    template stays enabled while dirty and asks first; the header's
    Apply & Restart (components/RestartButton.tsx) now carries its own tooltip
    saying what it is for and what a relaunch does to work in progress, and asks
    the same in a confirm at the click (chats and history stay; a reply in
    progress stops); the MCP servers heading is glossed "external tool
    connections"; the glossary's Chat line says "a one-off session that runs a
    template — nothing is created or enrolled"; the chat
    button, the tool-pill remove and the model badge carry titles saying what they
    do. The tab is lazy-loaded so it does not ride in the dashboard shell's App
    chunk, and the lazy chunk sits inside an ErrorBoundary in
    CapabilitiesPage.tsx so a stale chunk after a deploy degrades to the tab, not
    the dashboard.
    website/src/api/client.ts adds agentTemplates, agentTemplateCreate,
    agentTemplateDelete.
  • website/src/i18n/locales/*.json + pluralKeys.json: the tab's copy in all
    twelve catalogs, plural forms per locale.
  • src/kiro_crew/dashboard/folder_repository.py + dashboard/state.py:
    hold / hold_folders, the store lock held across an awaitable read-only
    section (the delete guard's use above). .github/black-baseline.txt: the
    touched test/test_folder_store_writer.py graduated off the black baseline,
    as the ratchet requires.
  • Docs: docs/system-specs/modules/crew-mode.md (new section, components and
    test rows), docs/system-specs/modules/learn-cron-dashboard.md (endpoint
    contract; the "Template deletion" passage now describes the revived route),
    docs/feature-map/README.md (Agent Templates row; the Capabilities rail now
    hosts nine tabs), src/kiro_crew/docs/agents.md.

Backwards compatibility

Compatible: every existing request keeps its behaviour. The detail PATCH's
model and skills paths are unchanged (the crew pane still edits private
copies with them); only requests carrying the new definition keys meet the
read-only rule. DELETE /api/agents/detail/{name} was absent on main, so no
caller depended on its absence. No config key, spec field or route was renamed
or removed; package, runtime and markdown specs are never written.

Tests

  • test/test_agent_templates_endpoint.py (70): the roster marks editability per
    origin (a row with no spec file beneath the agents directory — empty, foreign
    or absent filename — is read-only for the runtime's reason), lists crew / schedule (as dispatched: sequence over dormant
    agent_id, script jobs skipped) / chat-folder / webhook / private-copy
    references, and masks package-controlled strings (dropping a row whose name
    would be masked); a cron record the scheduler cannot load holds nothing; the
    detail PATCH refuses an ambiguous name for every key (model included) with neither file touched, and re-checks ambiguity under the write lock; a row whose filename is absolute, traversing or nested is
    never unlinked (404, file intact); a delete refusal masks its reference
    labels;
    create writes a minimal runnable spec, copies a package template without
    lineage, refuses an invalid, reserved, taken or crew-bound name and a
    malformed body (a name bound only in config.local.json included), copies
    the source as it is inside the spec lock, refuses a source made ambiguous or
    replaced after the probe (409, nothing written), answers 409 (not 500) for an
    ambiguous source name, publishes a created name to the dispatch snapshot and
    awaits the rescan after a delete (a refusal touches neither), and a successful create/delete emit their SEL events; delete refuses a
    referenced template listing the references (including a folder pin, a
    webhook pin and a binding that lives only in the overlay), checks and unlinks
    inside one folder-store hold so a pin that lands while the guard runs is
    seen, holds the cron store lock from the reference walk through the rename
    (probed on both sides) and answers 503 schedule_store_busy with the file
    intact when another holder keeps it, names a schedule written past the lock
    (warning + SEL row), fails closed on an unreadable cron store (503, file intact), runs create and delete through the drained seam, refuses a create whose name a file declared after the probe, does not count a template that merely shares the default crew's alias, refuses the default agent by stem, refuses package and runtime specs,
    retires an unreferenced template to a one-deep tombstone (and prunes its
    bookkeeping; the rename precedes the prune, so a refused rename keeps the
    older grave, same-second graves stay distinct, and a live template whose
    name looks like a grave is spared), never turns a post-delete store read failure into a failed
    response, 404s an unknown name, refuses a name two files reach (a crossover, and a same-name twin the roster collapses) with neither unlinked, refuses a row whose file does not answer to the requested name, a second claimant landing after the probe (ambiguity re-checked under the lock) and a row that calls a package file plain (file re-read under the lock);
    the detail PATCH writes the definition keys on an owned template (an MCP ref
    survives the governance sanitizer, a builtin auto-approval is withheld by
    it), refuses them on a package template, classifies the targeted file rather
    than its declared name, and validates their shape including the per-tool
    length cap.
  • test/test_agent_spec_hardened_reads.py: the call-site label ratchet lists
    the new _read_agent_spec caller.
  • test/test_folder_store_writer.py: hold_folders keeps the store lock
    across a section that hops off the loop, hands it a snapshot, and a mutation
    queued meanwhile commits only after the section returns.
  • test/test_webhooks_api.py (+1): a mint whose agent vanishes between the
    loop-side pre-check and the locked commit mints nothing, and a re-pin in the
    same window leaves the stored pin alone, while a label-only PATCH never takes
    the spec lock.
  • test/test_doctor_deprecated_agents.py (+2): the shared cron walk carries job
    ids and gates on loadability only for the guard; doctor still reads a partial
    record, and reports the captured execution's template over a stale or empty agent_id (a malformed context falls back to agent_id).
  • website/src/components/RestartButton.cov80.test.tsx (+1, 7): Apply & Restart
    asks first, naming what stays and what stops; declined does nothing, and the
    existing confirmed-path cases answer the ask yes.
  • website/src/test/AgentTemplatesTab.test.tsx (25): grouping by origin; the
    two-control action row and its overflow menu (enroll hint / already-enrolled
    state, Delete on an owned row vs Duplicate to edit and a disabled prompt on a
    package row, the read-only reason shown once); the definition save through
    agentPatch with the auto-approval toggle; a dirty draft survives a
    background refetch of the detail; a rejected detail read renders its error
    rather than Loading; only string-valued MCP fields render from a hand-edited
    spec; a dirty draft is not carried into a newly created template (New template
    asks first; the create drops the draft before selecting); a save is confirmed
    in the bar's slot and the Add tool label stays visible while typing; deleting
    a delete names the deleted template over the next row, and deleting
    a dirty template on a narrow viewport returns to the list; every in-app link
    the tab owns routes through the shell's leave gate with its target, and
    beforeunload is armed only while the draft is dirty; the overrides group
    is glossed and its row described as an override; resources render as plain rows, not chips; a read-only template with no
    stored prompt says so instead of showing an empty editor (no character count beside it); one Skills heading;
    the dirty-draft confirm on a row switch; the server's read-only refusal
    reported inside the save bar beside Save, cleared by Discard (which asks
    first; declined keeps the draft); a saved skill list lands in the detail
    cache before the refetch; Chat with this template
    dispatches createSlot in the template namespace; a referenced delete opens
    the reference dialog, naming a chat folder among the rows and linking a
    private-copy row to its crew; a private copy's banner offers Open crewmate;
    blank create vs a
    duplicate that sends from, with the created row selected after the roster
    refetch; a name the server would refuse is refused before sending; the usage
    line names chat-folder and webhook holders and each drills in to where the
    holder lives. The ApiError stub carries the
    raw response text like the real one, which is what caught the reference list
    being read off the wrong shape.
  • Catalog parity and every locale style suite pass with the new keys;
    delete_confirm is pinned in QUOTED_OPERAND_CONFIRM_KEYS.

Manual verification

Against an isolated gateway (KIROCREW_HOME under scratch, KIRO_HOME pointed
at a scratch agents directory holding two package templates and three owned
ones, port 6799, this worktree's dist), with a crewmate pr-bot and a chat
folder Reviews both pinned to pr-reviewer: the roster lists pr-reviewer
as used by the crewmate and the folder, each holder a link to where it is held; editing its prompt raises the save bar
("Unsaved changes · affects 1 crewmate", "Press Save template to keep these
edits: new chats use them at once; chats already running pick them up after
Apply & Restart"); the overflow menu shows Enroll with its explanation, Duplicate and
Delete; the Skills section carries one heading; confirming Delete opens "Can't delete pr-reviewer yet" with a Crewmate
row and a Chat folder row, each with Open →, and the file stays; the package
template atlas shows "Read-only · path" in the header, the reason once in the
banner with Duplicate to edit beside it, and a disabled editor; Duplicate
to edit
opens New template in duplicate mode with atlas in "Copy of". The
reference dialog probe is what exposed the ApiError.body shape bug fixed in
this round.

Screenshots / video

Owned template — two-control header, usage line linking to the crewmate and the chat folder, editable definition:

Agent templates tab with an owned template open

Delete refused while referenced — one row per holder (crewmate, chat folder), each with a jump:

Referenced delete refused with the reference list

Package template — "Read-only · path" header, the reason once in the banner with Duplicate to edit, disabled editor:

Package template shown read-only

Save bar, Tools editor (editable and read-only), overflow menu, New template in duplicate mode

Save bar after a prompt edit — affected crewmates, the save hint, Save template and Discard:

Save bar after a prompt edit

After Save — the bar is gone and "Template saved." sits in its slot:

Template saved confirmation in the save bar's slot

Tools editor on an owned template — ✓ marks auto-approval, Add tool open; Skills, Resources and MCP servers below:

Tools editor, editable

The same sections on a package template, read-only:

Tools and lower sections, read-only

Overflow menu — Enroll as crewmate with what it starts, Duplicate, Delete:

Overflow menu open

New template opened from the header — the blank / duplicate choice:

New template dialog

Duplicate dialog, seeded from the selected template and titled by its entry point:

Duplicate dialog

The other read-only variants — runtime (Built-in group), markdown-defined, and a crewmate's private copy (Private copies group) — and the empty state:

Runtime template, read-only

Markdown-defined template, read-only

Private copy, read-only

Empty state

Related Issues

N/A — follows #12224 (templates as chat execution choices); no tracked issue.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|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

@CrysisDeu
CrysisDeu requested review from a team and Zedmor September 21, 2026 07:42
@CrysisDeu

Copy link
Copy Markdown
Contributor Author

Intent: Give the shared agent templates under ~/.kiro/agents/ a management surface — see every installed template by origin and what runs it, edit the ones the user owns (description, model, prompt, tools, skills), create and delete safely, and start a chat or enroll a crewmate from one — as the counterpart of the crew pane that edits private copies.
Not a goal: Editing MCP servers or resources from this tab, changing what a crewmate or a private copy is, writing package/runtime/markdown specs, or managing project-scoped templates.

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Sep 21, 2026
@github-actions

github-actions Bot commented Sep 21, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Fable 5) — 🟡 CONCERNS

Design-level review of 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

The delete guard's holder list is completeness-by-convention: any future store that persists a template name silently reopens the dangling-reference hole the guard exists to close.

Watch

The guard enumerates six holder kinds by hand (template_references: crews, default, schedules, folders, webhooks, private copies) — "it counts every CONFIGURATION that names a template". Each kind is tested, but nothing pins the enumeration itself, so a future feature that persists a template name (an app manifest, a crew-mode slot binding, a workflow) ships without joining the guard, and its deletes recreate the exact "sessions that silently start on the default agent" failure the PR names. The author already recognized this drift class for schedules ("that rule now lives once… so the two readers cannot drift") but the holder list has no equivalent, in a codebase whose pattern is pinned unions (test_sandbox_governance_mask.py).
Clears when: a pinning test or registry ties template-name-persisting stores to template_references (even a scan asserting known name-bearing config/store keys are all counted), or the crew-mode spec records the invariant as a rule new stores must satisfy.

[DESIGN-REVIEWED] 7b46278

@github-actions

github-actions Bot commented Sep 21, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5) — 🟡 CONCERNS

Premise-level review of 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0 — 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: CONCERNS

Apply & Restart's new confirm dialog is an added step every Capabilities-tab user pays, justified only by a hypothesis about hesitant readers — no named failure.

Not justified as shipped

  • Item 7 — inherited: the confirm on RestartButton.tsx rests on "a reader who reads 'Restart' as breaking something never presses it" — a guess, not a report; the button serves all nine tabs (1 mount site counted, CapabilitiesPage.tsx), so every skills/prompts restart now costs an extra click forever, while the tooltip and save-bar text this PR also adds already carry the same reassurance.
    Subtraction: delete the window.confirm call in RestartButton.tsx (keep the new title tooltip and save-bar wording).
    Clears when: a report of a user losing an in-flight reply to an unwarned restart, or observed avoidance of the button, is linked.

What this change ships

Inventory (10 items) — 8 justified

Intent: let a user see, edit, create, and retire the shared agent templates from the dashboard instead of hand-editing files under ~/.kiro/agents/ — an ADDITION.

  1. New "Agent templates" tab listing every installed template by origin, with editability and what-still-runs-it — justified
  2. Owned template's description, prompt, tools and auto-approve marks now editable in-app (detail PATCH widened, read-only rule server-side) — justified
  3. New template creation, blank or as a duplicate of an installed one, without publishing a crew's private copy — justified
  4. In-app template delete returned, refused while referenced, file retired to a recoverable tombstone — justified (the retired handler's own spec text invited exactly this return)
  5. Delete guard and roster now count schedules, chat folders, webhooks and private copies as holders, shown before a delete — justified
  6. Webhook token mint/re-pin now commits under the agents spec lock and can refuse at commit time — justified
  7. Apply & Restart now asks a confirm before restarting, on every Capabilities tab — inherited: no named person was failing; hypothesis-only provenance
  8. Schedules created from a template chat now pin their template for both the delete guard and kirocrew doctor, via one shared walk — justified
  9. The templates tab's code loads lazily on first visit, behind an error boundary — justified
  10. test/test_folder_store_writer.py reformatted and dropped from the black baseline — rides along, harm-free

More items exist (13 locale catalogs, spec/docs updates in the same commit — both mandated by documented invariants); capped at the 10 a person would notice.

[FIRST-PRINCIPLES-REVIEWED] 7b46278

@github-actions

github-actions Bot commented Sep 21, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Fable 5) — 🟡 CONCERNS

UX-level review of 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

Users still won't dare press Apply & Restart — its reassurance lives behind the click, and nothing shows which running chats it would touch.

Watch

  • Apply & Restart: the blind reader, after reading the glossary explainer, said "I would not dare click it… I can't see which chats are running or what 'restart' does to them." The window.confirm this PR adds ("Your chats and their history stay; a reply in progress stops") answers exactly that fear, but only post-click — a reader who won't press it never sees it, so saved edits never reach running chats. Frequent (every edit while a chat runs) × value lost silently × every time. Smallest fix: surface the running-chat count and the what-stays line pre-click (in the save bar's hint or beside the button), not only in the confirm and hover title.
  • Padlocked rows under "Mine" (e.g. release-notes-md) carry no visible reason; the reader said "in the left list I cannot tell which reason applies until I open the entry." Hover-title only. Low impact, but recurring for anyone with markdown templates.

Suggestions

  • In list rows whose group doesn't explain the lock (the "Mine" group), pair the Lock icon with its short reason word (e.g. "Markdown") instead of a title-only hint.
  • The reader had to "slow down" to keep model: auto apart from the tools' auto-approve ✓; consider the editor's existing clear-label wording "auto (backend default)" for the list badge too.

[UX-REVIEWED] 7b46278

@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 Sep 21, 2026
@github-actions

github-actions Bot commented Sep 21, 2026 •

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 7b46278

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0: <one-sentence reason>

@github-actions

github-actions Bot commented Sep 21, 2026 •

Copy link
Copy Markdown
Contributor

Opus 4.8 Review — ✅ no blocking findings

Reviewed 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0 — this comment is updated in place on each push.

Review details

I've reviewed the full backend logic (cron.py, agent_templates.py, agents.py PATCH widening, hooks.py token-commit serialization, folder_repository.py/state.py lock helpers, routes) and the frontend (AgentTemplatesTab.tsx, client.ts, RestartButton.tsx). The candidate list was empty, and my own extension pass turned up nothing groundable to the (a)/(b)/(c) bar at 80+: the concurrency-sensitive delete path re-validates every reference and re-classifies the file under the same cross-process locks its writers take before the tombstone rename, and the pre-lock ambiguity/name-taken checks are all backstopped by locked re-checks, so the plausible smells resolve to correct outcomes rather than observable failures.

No findings.

[OPUS-REVIEWED] 7b46278

Verdict parsed from the review's SHA-scoped output markers for commit 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 7b46278948fb04819cbfbbbb9abdf444ecb9c0f0: <one-sentence reason>

@CrysisDeu
CrysisDeu force-pushed the feat/agent-templates-page branch from 35c3e11 to 7a87e87 Compare September 21, 2026 08:45
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Sep 21, 2026
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=adef54b47f5d — tool strings remain effectively unbounded / deduped lookup bypasses package read-only protection

Fixed in 7a87e87. Both findings share this span (agent_templates.py, validation) and are one rationale: the PATCH's validation trusted the wrong evidence. validate_definition_patch now caps every tools / allowedTools entry at MAX_TEMPLATE_TOOL_CHARS (256) on top of the count cap, so a list's size is bounded by count × length. read_only_reason_for_path no longer looks the declared name up in the deduplicated roster: it classifies the FILE the PATCH is about to write — runtime by OWNED_KIRO_AGENT_FILES, markdown by suffix, then _global_agent_info(path, spec) (the same package/builtin rule discovery applies to every row) plus the fork sidecar for private_to. atlas.json beside SomePkg-atlas.json therefore answers None for the plain file and package for the package file, whichever the name-first lookup lands on.
Pinned by test_patch_definition_classifies_the_targeted_file_not_its_name (both files present, both answers asserted) and the {"tools": ["x" * 257]} case of test_patch_definition_shape_is_validated.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=3b42ca861125 — action row exceeds two-button limit / error notices omit the required handoff decision

Fixed in 7a87e87. Same span (AgentTemplatesTab.tsx), one rationale: the pane's chrome did not follow the two frontend rules. The header row now holds exactly two controls — Chat with this template and one overflow DropdownMenu (Enroll as crewmate with its explanation, Duplicate / Duplicate to edit, Delete). The read-only remedy moved into the read-only banner beside its reason (a separated region), so no row anywhere exceeds two. Both ErrorNotice sites that sit beside an unsaved draft (the pane notice over the definition draft; the create dialog's notice over the typed name/description) now carry the {/* No hand-off: … */} decision naming that draft; the two load/detail failure notices already pass askAgent.
Pinned by the roster test asserting the row's button count is 2 and the menu's item list per editability.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=2f6082500a24 — feature map incorrectly includes skills in the shared draft

Fixed in 7a87e87. The Agent Templates row now says the draft is description, model, prompt and tools, and that skills save on their own through the shared skills editor (a computed view over resources, not part of the draft). The owning spec (crew-mode.md) says the same and adds the consequence the Opus lane found: because the skills editor invalidates the prefix the detail query shares, the editor reseeds from a refetch only while the draft is clean.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=9f4f9fc067a2 — reseed effect overwrites a dirty draft after the skills editor's invalidation

Fixed in 7a87e87. The reseed effect now returns early while the draft is dirty (if (dirty || !detailQuery.data || detailQuery.isFetching) return, dirty in its dependency list). A row switch clears the draft before changing selectedName, so a clean editor is the only one reseeded; after a save the baseline is adopted first, so the post-save refetch reseeds from the server's sanitized truth as before.
Pinned by keeps a dirty draft through a background refetch of the detail: types into the prompt, swaps the detail mock, invalidates ['agent-templates'] on the real QueryClient, waits for the second read, and asserts the typed prompt and the dirty bar are still there.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=0432628e34e4 — _AmbiguousTemplateName escapes api_agent_template_create as a 500

Fixed in 7a87e87. Both _load_template_specs calls in the create handler are inside one try that answers 409 ambiguous_template_name with the same message the fork/publish handlers use, so two files declaring one name is reported as the user's to untangle rather than as a server fault.
Pinned by test_create_from_an_ambiguous_name_is_a_409_not_a_500 (atlas.json + SomePkg-atlas.json, from: atlas; asserts 409, the code, and that nothing was written).

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=05ea5f7259c9 — the reference guard left chat-folder default_agent pins uncounted

Fixed in 7a87e87. A chat folder's default_agent is a fifth holder and is now counted: template_references takes the folder pins and emits {kind: "folder", id, label} rows, so they appear in used_by on the roster and in the 409 references the delete guard returns (the tab renders them as "Chat folder" rows with a jump). The pins are snapshotted from the dashboard's folder store on the event loop (state.read_folders) before the roster / guard runs in the executor, so the guard reads committed folder state rather than the file. The endpoint contract, the owning spec and the shipped user doc now name the folder among the holders.
Pinned by test_roster_counts_a_chat_folder_pin_as_a_reference (a pinned folder counts, an inheriting one with default_agent: "" does not) and test_delete_refuses_while_a_chat_folder_pins_the_template; the frontend test names a chat folder among the references renders the row.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=083f9c50ce23 — "Save template" vs the page header's "Apply & Restart"

Fixed in 7a87e87. The unsaved-changes bar now carries a second line, "Save writes the file; no restart needed." (save_hint, all twelve catalogs), so the distinction is stated where both controls are on screen: saving is a file write that new sessions pick up at once, and the header's restart is for sessions already running. The bar's first line also changed to "Unsaved changes · affects N crewmates" so it reads as one sentence about the save. Screenshot 02 shows the bar.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=7f252eeaebb3 — "Enroll as crewmate" had no explanation

Fixed in 7a87e87. Enroll moved into the header's overflow menu as a two-line item: the label, and under it "Adds a crewmate with its own memory that runs this template. Nothing starts until you chat with it." (enroll_hint); when a crewmate of that name already exists the item is disabled and the line says "Already enrolled as a crewmate." Screenshot 06 shows the open menu.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=46b6979c93a2 — read-only reason repeated in the header subtitle and the banner

Fixed in 7a87e87. The subtitle is now "Read-only · ~/.kiro/agents/" (read_only_short); the banner alone carries the sentence, with Duplicate to edit beside it as the remedy. Screenshot 03 shows the pair.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=117d24546ca4 — save bar's bare "· 0 crewmates"

Fixed in 7a87e87. The bar uses a new plural key affects_crewmates ("affects {{count}} crewmate(s)", per-locale plural forms registered in pluralKeys.json) instead of the bare count badge string.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=0b7702f1bab6 — tools_hint used internal vocabulary

Fixed in 7a87e87. Now "click a tool to toggle auto-approval ✓; {{productName}}’s permission rules may still ask first" (the product name is the interpolation variable the catalog contract requires), translated in every catalog.

@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 Sep 21, 2026
@CrysisDeu
CrysisDeu force-pushed the feat/agent-templates-page branch from 7a87e87 to 5a255dd Compare September 21, 2026 09:26
@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Sep 21, 2026
@github-actions

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — ⚠️ review incomplete

GPT 5.6 did not produce a complete verdict for 11c72b654354719631475767a2c62207e1637756; inspect the workflow logs and re-run it.

This comment is updated in place on each push.

See the GPT 5.6 Review job logs; this commit has no completed GPT verdict.

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 11c72b654354719631475767a2c62207e1637756: <one-sentence reason>

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=adef54b47f5d — Template mutations leave the dispatch snapshot stale

Correct — only the roster cache was cleared, while "Chat with this template" and every template-bound turn resolve through _materialized_kiro_agent's in-memory snapshot. Create now calls publish_materialized_agents([name]) (loop-safe set union) and then schedule_materialized_agents_refresh(), so a slot created before the rescan lands is not normalized — and stored — onto the default agent; delete AWAITS refresh_materialized_agents on the discovery executor before answering, since a removal has no publish shortcut. Tests: test_create_publishes_the_name_to_the_dispatch_snapshot (publish then schedule, in that order) and test_delete_refreshes_the_dispatch_snapshot_before_answering (a 409 refusal touches nothing; a 200 refreshes). Spec updated. In 366db25.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=c16ef20aad20 — One fact, three phrasings: "Overridden by N crewmates" vs "Crewmate overrides"

The count badge now says "{{count}} crewmate override(s)" — the group heading's and the banner's word — in all 12 locales (zh-CN "{{count}} 个队友覆盖版", ja "クルーメイトの上書き {{count}} 件", de "{{count}} Crew-Kameraden-Überschreibung(en)", ru with few/many). Test asserts the badge text on the source row. Shot 01 re-taken. Spec updated. In 366db25.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=a0bf9c7f4263 — Apply & Restart — reader would not dare click it

Sixth round on this button. Every remedy raised so far is in: both halves in the save bar, the glossary sentence, the tooltip, and a confirm at the click whose first sentence says what stays. What the reader now asks for — seeing which chats are affected before pressing — is information the header button does not have on this tab, and adding it means wiring live session state into a control shared by every Capabilities tab. That is a dashboard-wide behaviour decision I have handed to the PR owner (fcab0e7a2a7d, 1fe7dd8b3acf); copy iteration on this tab stopped two rounds ago.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=596070d31878 — Say "Relaunch the agent behind N open chats…" in the confirm

Concrete and worth doing, but it belongs with the decision above: the count of open chats lives in the chat store, and RestartButton is a store-free shared component — wiring it in is the same session-aware change as "show restart only when sessions run", which the owner is deciding. Deferred to that follow-up rather than widened into this PR.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=567bd240b294 — Item 9 — the private_copy → "Crewmate override" relabel is unjustified

Same item as 106380e2010d two rounds ago; the linked record is unchanged: #12481 (comment). The UX lane's blind reader on this PR met "Private copy" (crew pane) and "Crewmate override" (this tab) for one object and flagged the pair as confusing; the relabel is what closed that, and the tab's group heading, banner, usage line and dialog row all say "override", with "Open crewmate" landing in the pane where the badge lives. Reverting re-opens an observed confusion to save a 13-catalog one-string diff.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=fb166ce05be4 — Item 8 — the Apply & Restart confirm rides along (tag only)

Acknowledged as the lane itself tags it: the harm (a reply in progress stops with no warning) is named, so the step is justified by its own text; the finding asks for nothing further. Kept, with the rationale on record at 54131862db82 (the UX reader's two refusals to press the button).

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=15471147d303 — A ratchet test for the delete's lock order

Not in this PR: the order (hold_folders → config → config.local → agents_spec_lock → token store) is spelled out in the handler's docstrings and crew-mode.md, and the two writers that share a pair of these locks (hooks._commit_pinned_token, this delete) are both exercised by tests that would deadlock or fail if either inverted the pair. A generic acquisition-order ratchet needs instrumentation on every lock helper (update_config_locked, agents_spec_lock, the folder store hold, the token store), which is a repo-wide testing seam worth its own PR; I would rather add it once for all writers than a bespoke probe for this one.

@github-actions

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — 🔴 changes requested (blocking)

GPT 5.6 found at least one blocking issue that must be resolved before merging 366db2592a64f7f5a594dc31d47daa8cc70dc73b. Adjudication produced no [GPT-ADJUDICATED] marker for 366db25: the adjudication stage FAILED TO RUN (its raw output is in the Adjudication details block -- typically an API or setup error), so no proportionality review took place and GPT's blocking verdict stands unreviewed.

This comment is updated in place on each push.

BLOCKING -- src/kiro_crew/dashboard/handlers/agent_templates.py:831 -- Schedule creation can race template deletion
tombstone[0] = _tombstone(spec_path)
Concurrent cron creation after _cron_agent_ids() -> job persists before _tombstone() -> both operations succeed, but the job cannot execute its selected template.
Anchor: residual/crash-data-loss-corruption
Fix: Serialize cron writes with the reference check and rename.
[BLOCK-MERGE] 366db25
[GPT-REVIEWED] 366db25
False positive or not applicable? A repository writer can comment:
/ai-review override gpt 366db2592a64f7f5a594dc31d47daa8cc70dc73b: <one-sentence reason>

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=adef54b47f5d — Schedule creation can race template deletion

Closed with the scheduler's own lock rather than a new one: the cron store's cross-process advisory lock (.crons.lock) is now exposed as cron.cron_store_lock(store_dir) — the ONE implementation, which CronService._file_lock (and so every store mutator) delegates to after its loop-safety guard — and the delete holds it from the dispatch-reference walk through the tombstone rename, innermost in its lock order (folder hold → config → overlay → spec lock → store lock; the mutators take only the store lock, so no inversion). A schedule therefore cannot be saved against the template between "nothing dispatches this" and the file going; a holder that keeps the lock past the bounded spin is 503 schedule_store_busy with nothing unlinked. The post-rename re-read stays as a WARNING + SEL row for a writer that bypasses the lock (a hand edit). Tests: test_delete_holds_the_cron_store_lock_from_the_check_through_the_rename probes the lock from a second fd at both the walk and the rename, and test_delete_refuses_while_the_cron_store_lock_is_held_elsewhere pins the 503 with the file intact; the 308 cron locking/regression tests still pass with _file_lock delegating. Specs crew-mode.md and learn-cron-dashboard.md updated in the same commit. In eec87f8.

The definitions a chat or a crewmate runs had no management surface of
their own. The crew editor's Template pane edits one crew's PRIVATE copy
of a template, so a template no crew is bound to -- every freshly
installed package agent -- had nowhere to be seen, and the dashboard could
change only a template's model and skills. Creating a template meant
publishing some crew's private copy; deleting one meant removing the
file by hand.

Add an Agent templates tab beside Crews under Agent Capabilities. It
lists every installed global template by origin (Mine / Private copies /
From packages / Built-in) with what still runs it, edits an owned
template's description, model, prompt, tools and auto-approval marks as
one draft, edits skills through the existing skills editor, and shows
resources and MCP servers read-only. A package, runtime, markdown or
private-copy template is read-only with Duplicate to edit. New template
starts blank or copies an installed one; Delete is refused with the
reference list while a crewmate, the default agent, a schedule or a
private copy still names the template. Chat with this template opens a
chat in the template namespace; Enroll as crewmate binds a member to it.

Backend: `handlers/agent_templates.py` adds `GET/POST /api/agents/
templates` and revives `DELETE /api/agents/detail/{name}` from the
handler retired with the old standalone page, keeping its lock
discipline and widening the reference guard to schedules and private
copies. The detail PATCH accepts the definition keys and refuses them on
a read-only spec (`409 template_read_only`).
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=35999f6106ec — Handler docstring says the scheduler's writers hold none of these locks

Stale prose, not stale code: the delete already holds cron.cron_store_lock across the walk and the rename. The handler docstring now lists the store lock as the innermost lock in the order and says the post-rename re-read is for a writer that bypasses it (a hand edit); the inline comment at the re-read says the same. crew-mode.md already carried the correct text. No behaviour change. In 7b46278.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • fixed span=10ca1a0b6a35 — Empty state says "ships agents" while the page says template

empty_subtitle now says "Create one here, or install a package that ships templates." in all 12 locales (zh-CN "…带有模板的包", ja "テンプレートを含むパッケージ", de "…ein Paket mit Vorlagen"). One word for one thing on the page. In 7b46278.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=8bd3398c90fe — Apply & Restart — reader would not dare click it

Seventh round on this button; every remedy raised has landed (save-bar halves, glossary line, tooltip, click-time confirm). The remaining ask — renaming or session-gating a header control shared by every Capabilities tab — is a dashboard-wide decision I have handed to the PR owner (fcab0e7a2a7d, 1fe7dd8b3acf, a0bf9c7f4263). Copy iteration on this tab is closed.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=6f7309cebe8d — folders_count links to bare /chat

The chat page has no per-folder route today (/chat opens the folder rail; a folder is selected there), so there is nothing deeper to link to without adding a route param to the chat shell, which is outside this PR. The holder's kind label ("Chat folder") and the folder's own name are both in the usage line and the refused-delete dialog; a folder deep-link is a small chat-shell follow-up, noted for the owner.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=fb5424cf342b — Item 6 — _commit_pinned_token serializes every mint on the spec lock

Same item as 4bdf2c955e3a (round 21), answer unchanged: the refused-call outcome lands on a THIRD PARTY's first call, not on the operator who minted the token, and nothing on the operator's side says why the integration they just wired is dead — that asymmetry with schedules (whose failure lands on the operator's own Schedule page) is the named case. The lock turns it into a 409 at the mint, in the operator's click. This race was itself a GPT [BLOCK-MERGE] on an earlier head; removing the lock re-opens a security-fenced finding to save one lock acquisition on an operation that runs a handful of times per install.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=9e1d55cc65f0 — Item 7 — the RestartButton confirm rests on a hypothesis

Same item as 54131862db82 and 0a202c70783a; the named report is linked again: #12481 (comment) quotes the UX lane's blind reader declining to press Apply & Restart twice on shots 01 and 14 ("'Restart' sounds like it interrupts something already happening and I don't know what I'd break"). That is the observed hesitation the clearing condition asks for. Kept.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=25713d8d75ab — Ratchet the six holder kinds in template_references

Same suggestion as 82a45618e2cf (round 23), answer unchanged: the six holders live in five stores with no shared schema, so a registry is a second hand-written list, and a pin test can only assert the list the author already wrote — it cannot see a future config field that carries a template name. The behaviour tests name every holder kind (crew, default, schedule-as-dispatched, folder, webhook, private copy), so dropping one goes red. A cross-store 'names a template' registry is a repo-wide seam worth its own PR.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=c315dcaee3ae -- Holder enumeration in template_references is hand-maintained

The six holder kinds are recorded in docs/system-specs/modules/crew-mode.md as the guard's contract, and the module docstring names the same list, so a new name-persisting store has a written rule to satisfy. A registry or scan test that ties every config/store key carrying a template name to the guard is repository-level test infrastructure across five stores with no shared schema; a heuristic pin would be brittle. Tracked as a follow-up PR, not folded into this one (sixth round of this item).

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=f69e7de8fb76 -- Item 7: the confirm on RestartButton is justified by observed reader avoidance

The confirm was not a guess: the UX lane's blind reader on this PR twice reported not daring to press Apply & Restart because the effect on running chats was unstated (UX Review on heads 63a080b and a494525, cited in the earlier disposition on eec1f4e). That is the observed avoidance this item's own clearing condition asks for. The confirm is one line plus one i18n key, tested in RestartButton.cov80.test.tsx, and independently revertable.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=a31eb089ddd0 -- Apply & Restart: reader cannot see which chats are running

Showing which chats are running, or disabling the button when none are, requires the shared page-header RestartButton (used by all nine Capabilities tabs) to read live session state. That is a dashboard-wide behaviour change outside this tab's scope; the owner has been asked to decide it separately (eighth round of this item). Within scope, the confirm text now states the effect on running chats and the glossary says when a restart is needed.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=2000d8b85d19 -- Padlocked rows under Mine carry no visible reason

The reason is shown as soon as the row is opened, as a bold two-word lead in the read-only banner (Markdown file / Built in / From a package / Crewmate override), and the row lock carries a title. Adding a reason word to every list row changes the row layout for a low-impact case the lane itself rates as low; deferred to the follow-up copy round rather than another push on a review-ready head.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=bf0fe70fb75e -- Pair the Lock icon with its reason word in Mine rows

Same item as 2000d8b85d19: the reason appears as the bold lead of the banner on open, and the lock has a title. A per-row reason word is deferred to the follow-up copy round; no push on a review-ready head for a suggestion-level item.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • rebutted span=5e18cd75c60b -- List badge model: auto vs tools auto-approve

The tools tag was already renamed to auto-approve in an earlier round to separate it from the model value, and the list badge carries a title explaining that auto means the backend default. Spelling out auto (backend default) on every list row widens the badge and repeats the editor label; deferred to the follow-up copy round.

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.

2 participants