[bkudiess]

What Problem This Solves

Users need a complete automation inventory and reliable everyday controls without reproducing the entire Gateway web editor.

User Impact

User impact: browse all accessible agents and paused or advanced jobs, run or pause jobs, edit ordinary schedules, inspect recent results, and open chat or web management for deeper work. Owner delivery and unsupported timing writes/copies remain web-managed; basic edits omit unchanged unsupported fields and preserve hidden saved settings.

Removal is not atomic: the client checks the loaded revision with cron.get, then sends the released Gateway's id-only cron.remove. A concurrent edit between those calls is not detected. Reported conflicts are surfaced without an unguarded removal retry.

Why This Change Was Made

The native page prioritizes discovery, status, timezone-aware presets, delivery visibility and daily actions. Permission checks, confirmations, draft lifetime and stale-response fencing remain explicit instead of silently converting advanced jobs. Landed Usage and Dashboard owners are preserved.

A reproduced queued-chat ordering defect is also repaired: delayed acceptance no longer reopens a completed turn or places its prompt after the response. Unrelated/newer runs and queued follow-ups remain isolated. The combined fixture inventory uses a measured 1500-second aggregate allowance within the unchanged 30-minute outer step; individual deadlines, assertions and cleanup gates are unchanged.

The final reader repair relinquishes bottom anchoring for an actual keyboard-directed destination outside the follow range. Matching native completion and existing tail-navigation owners retain follow/restoration responsibilities; no extra scrolling or forced layout is introduced.

Evidence

Current-head CI passed on attempt 1, including the required CI Gate. Head 35715c5900e0368f3d43d13bd0dd20e367f70a67 was tested as merge b6288e913dec3aa111c36b205961b9852f73569d (main a6ff6a16 plus this head), whose tree 9582ba29a84a747f2c6b369e60c2a7c3d8420005 exactly matches the local candidate.

The requested simplification removed 373 net lines of temporary diagnostics across three test files, with no production, test-case, assertion, action, wait-budget or CI-selection changes. The reduced candidate passed all 122 Debug fixture cases, both Debug/Release reader widths, two Release queued-chat cases, native UI 1243 cases with nine policy skips, and full unfiltered hosted Tray 6276/6276.

Previous intermittent failures are preserved. The reader counterfactual demonstrated that changing anchoring released the already-pending requests; it did not identify the source of the nine-DIP measurement oscillation. The original reader assertions passed through the production repair in Debug and Release. Temporary reader, reconnect and notification diagnostics are now retired without removing their regression tests; their removal is not a reconnect or notification root-cause fix. Fixture evidence does not establish live Gateway timing or installer/Store upgrade behavior.

Review scope and resolved findings

Independent Opus 5.5 and GPT-6 Astra code-review contexts covered implementation/proof-plan scope at 03f45691 against 045d13ac, 128 paths per reviewer. The rubber-duck role was unavailable. Ten retained findings were repaired in cab3b592 and subsequent lifetime, rendering and proof corrections. Supplemental coverage followed narrative exposure, so this was not wholly blind; reviewers did not execute runtime proof.

Separate bounded P0-P3 reviews covered integration conflicts, CI capacity, queued-turn repair, diagnostic safety, reader-anchor ownership, editor geometry and the Systems/Usage proof-readiness corrections. Accepted agent-delta correlation and Release-build prerequisite findings were repaired and rechecked. The simplification mapped 206 authored repair hunks to 193 retained and 13 removed; its independent review found no significant issues, and AST checks preserved cases, assertions, actions and waits. This is not a fresh whole-head dual-model review; subsequent reviewer model identities are unverified.

The older durable review was reconciled against source and executed regressions: Main normalizes to systemEvent/text; accepted invocation/run IDs govern completion, with conservative ID-less handling; unreachable full-history UI was removed while recent full results remain; hidden-field preservation and strict protocol tests protect existing jobs. Released portable-profile preservation is not installer/Store or live Gateway cron-storage migration proof.

Change Type

• Bug fix
• Feature
• Refactor
• Docs or instructions
• Tests or validation
• Security hardening
• Chore or infrastructure

Scope

• Tray or WinUI UX
• Windows node capability
• Local MCP or winnode
• Gateway, connection, or pairing
• Setup or onboarding
• Permissions, privacy, or security
• Tests, CI, or docs

Required proof pools

• windows-winui-interactive: native editing, navigation, focus, accessibility, responsive layout and visible result/hero rendering. Declaration is not complete-pool acceptance.

Validation

Current local source: head 35715c5900e0368f3d43d13bd0dd20e367f70a67, tree 9582ba29a84a747f2c6b369e60c2a7c3d8420005. Validation ran on the frozen source manifest whose exact tree became this commit. Its parent integrates pinned main a6ff6a16. $ownedResults denotes each invocation's separate task-owned output directory, with its private path redacted.

Recorded command	Result
.\build.ps1 -NoTrustRepository	All five targets passed
dotnet build .\tests\OpenClaw.Tray.UITests\OpenClaw.Tray.UITests.csproj -c Debug -r win-arm64 -p:DevBuild=false --disable-build-servers -nodeReuse:false	Passed, zero warnings/errors; compilation only
dotnet build .\tests\OpenClaw.Tray.UITests\OpenClaw.Tray.UITests.csproj -c Release -r win-arm64 -p:DevBuild=false --disable-build-servers -nodeReuse:false -p:RequireSignedCleanupHost=false	Passed, zero warnings/errors; unsigned non-Dev compilation only
dotnet test .\tests\OpenClaw.Shared.Tests\OpenClaw.Shared.Tests.csproj --disable-build-servers --logger "trx;LogFileName=shared.trx" --results-directory $ownedResults	5760 passed, 28 ordinary opt-in skips, zero failed
dotnet test .\tests\OpenClaw.Tray.Tests\OpenClaw.Tray.Tests.csproj --disable-build-servers --logger "trx;LogFileName=tray.trx" --results-directory $ownedResults --filter "FullyQualifiedName!=OpenClaw.Tray.Tests.SettingsManagerIsolationTests.OpenClawTrayDataDirRedirectsSettingsAwayFromRealAppData"	6275 passed, zero failed/skipped; partial local suite
dotnet test .\tests\OpenClaw.Tray.Tests\OpenClaw.Tray.Tests.csproj --disable-build-servers --filter "FullyQualifiedName~ChatTimelinePresentationTests" --logger "trx;LogFileName=chat-focused.trx" --results-directory $ownedResults	103/103 passed, zero skips
git diff --check	Passed

The one locally excluded Tray case reads OS-roaming settings outside isolation overrides. No repository/CI filter was changed. The exact case passed in both current hosted Tray variants; this does not retroactively make the local filtered run unfiltered.

Current-head hosted evidence	Result
Native UI and production resources	Standard UI 1243 passed, nine policy skips, zero failed; production-resource cases 31/31; reader 480/1100 and four editor geometry guards passed
Gateway fixture and Release proof	Debug fixture 122/122, profile 55/55, released-profile 1/1; Release queued-chat 2/2 and reader 2/2; no selected skips
Unfiltered Tray	6276/6276 plus Dev 31/31; exact locally excluded privacy case passed in both

Earlier evidence stays attributed: the 621/621 provider-owner and 12/12 ordering/isolation runs belong to 575bd3d8; unchanged managed cases re-executed in current full Tray closeout. Documentation, skill, skill-validator and workflow-contract scripts passed at 3dec70a9; those inputs are unchanged by this three-test-file reduction. Actual-source reader characterization reproduced the missing handoff on the prior owner and passed on the repair; this does not substitute for native event/layout execution.

Real Behavior Proof

• Environment tested: local isolated ARM64 managed/build validation; hosted Windows x64 Debug and targeted unsigned non-Dev Release fixtures.
• PR head or commit tested: head 35715c59; hosted merge b6288e91, identical tree 9582ba29, as recorded above.
• Exact steps or command run: recorded local commands above and hosted fixture invocations below.
• Evidence after fix: weekly editor journeys passed in Light/Dark; both reader widths passed in Debug and Release without the counterfactual; queued-chat ordering passed both Release variants. Current Usage captures show the complete expanded header, cost/share and all facts through Errors, unobstructed in both themes.
• Observed result: all required CI passed on the reduced candidate. The retained Release assembly identifies Release / OpenClawTray, and its SHA-256 906225BD5D23C040F4388F6EC105D13B95E1CB0573E41A5D58F9223E14A936C0 matches both running-app receipts.
• Screenshot or artifact links verified? (Yes/No/N/A) Yes: current job artifacts were downloaded, raw TRX inspected and the Light/Dark expanded-detail images visually checked.
• Not verified or blocked: local native input remains held after foreground refusal; full ARM64 Release/High Contrast/installed acceptance, live Gateway/provider timing, installer/Store upgrade and Gateway cron-storage migration are not established by the synthetic x64 fixtures.

.\scripts\test-gateway-fixture.ps1 -AppPath .\src\OpenClaw.Tray.WinUI\bin\Debug\net10.0-windows10.0.22621.0\win-x64\OpenClaw.Tray.WinUI.exe -ArtifactsDirectory .\TestResults\GatewayFixture -TestTimeoutSeconds 1500 -Screenshots -NativeInputProof -NoBuild
.\scripts\Invoke-CiTest.ps1 -Project tests\OpenClaw.Tray.UITests -Configuration Release -ResultsDirectory TestResults\GatewayFixtureRelease -TrxFileName GatewayFixtureRelease.trx -Runtime win-x64 -Filter "FullyQualifiedName~GatewayFixtureUiTests.CompletedRunLateSnapshotLeavesNativeComposerReadyAndQueueDraining" -ExpectedTestCount 2 -RequireAllTestsPassed -HangTimeoutSeconds 300 -CollectNativeFailureMetadata -CaptureHostExit
.\scripts\Invoke-CiTest.ps1 -Project tests\OpenClaw.Tray.UITests -Configuration Release -ResultsDirectory TestResults\ReaderRelease -TrxFileName ReaderRelease.trx -Runtime win-x64 -Filter "FullyQualifiedName~ReactorChatLayoutProofTests.ConversationFeedback_ThinkingToolReviewsAndHistoryRetainComposerWithoutSearch" -ExpectedTestCount 2 -RequireAllTestsPassed -HangTimeoutSeconds 120 -CollectNativeFailureMetadata -CaptureHostExit

CI explicitly selects the Release app, enables isolated fixtures/screenshots, and retains its DLL for configuration, identity and runtime-hash verification. Unsigned fixture builds do not change production signing defaults.

Security Impact

• New permissions or capabilities? (Yes/No) No.
• Secrets or tokens handling changed? (Yes/No) No.
• New or changed network calls? (Yes/No) Yes.
• Command or tool execution surface changed? (Yes/No) No.
• Data access scope changed? (Yes/No) Yes.
• If any answer is Yes, explain the risk and mitigation: native discovery reads all jobs accessible to the connected operator. Existing authenticated Gateway authorization governs reads and mutations. Explicit filters, capability checks and generation fencing reject stale results and unsupported writes. Web handoff retains existing credential policy; run-result rendering stays inert.

Compatibility and Migration

• Backward compatible? (Yes/No) Yes, with explicit web-management fallback for unsupported edits or Gateway revisions.
• Config or environment changes? (Yes/No) Yes: ordinary hosted fixture allowance is 1500 seconds; runner default remains 900 and accepted range remains 1 through 1800.
• Migration needed? (Yes/No) No.
• If yes, list the exact upgrade steps: N/A. Cron definitions remain Gateway-owned. Basic update patches preserve inherited fields rather than migrating or replacing them; the non-atomic removal caveat above remains.

Review Conversations

• I replied to or resolved every bot review conversation addressed by this PR.
• I left unresolved only conversations that still need maintainer judgment.

The final review-thread check found zero inline threads. No formal approval or bot reply is implied; the old-head durable findings are reconciled above and final required CI is green.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

ClawSweeper review complete

ClawSweeper finished reviewing this revision. The review result is being finalized.

View the workflow run.

[clawsweeper]

Codex review: blocked before merge.

What this changes

This PR replaces the Cron page with agent-scoped native Automations management and upgrades Reactor.

Example: Browse the ops agent’s daily automations.

• Before: The legacy Cron page requests an unscoped inventory.
• After: Automations requests agentId=ops and offers an explicit All agents view.

Review scores

Measure	Result	What it means
Overall readiness	🧂 unranked krab (1/6)	Useful authorized direction with substantial supporting tests, but current runtime evidence is insufficient and source review identifies actionable defects.
Proof confidence	🧂 unranked krab (1/6)	Real behavior proof is not required for maintainer- or bot-authored pull requests.
Patch quality	🦐 gold shrimp (3/6)	3 actionable review findings remain.

Product

Kind: Feature · Worth it: Yes · Fix scope: Complete
User problem: Operators need native agent-scoped scheduling without losing advanced saved settings or sending requests released Gateways reject.
Reason: The collaborator-authored direction provides a coherent daily workflow and explicitly leaves advanced authoring in the web app; no additional product approval is required.

Merge readiness

⛔ Blocked before merge - 5 items remain

Keep this maintainer-authored draft open: it delivers distinct native scheduling work, but two behavior defects and unreachable production UI remain.

Priority: P2
Reviewed head: f7ae188fe3ae6805a1e66e41075778f4ee6b7b20

Before merge

• Convert the payload when selecting the Main session (P2) - Create a new automation with the default assistant task, enter a name and prompt, then select Main. This handler changes only SessionTarget; Normalize leaves AgentTurn unchanged and validation accepts it. Save therefore sends sessionTarget="main" with payload.kind="agentTurn", which the released Gateway rejects in assertSupportedJobSpec. The old editor generated systemEvent/text for Main. Restore that mapping in the existing form normalization and cover the serialized request.
• Match run completion to the accepted invocation (P2) - If run A completes while the cached job still reports an earlier LastRunAtMs, starting run B captures that stale baseline. B's watcher then accepts A's history entry because its timestamp exceeds the baseline, clearing B's running indicator before B finishes and enabling actions prematurely. The acknowledgement's RunId is currently used only for a notice. Retain it with the run owner and require matching completion, including in list reconciliation; timestamp-only completion must not settle an ID-less acknowledgement. The same distinction is documented in fix: keep older Run Now attempts from clearing newer runs #2082.
• Remove the unreachable full-history component (P2) - The narrowed page never instantiates AutomationRuns: its only production use is the static RenderRun helper from AutomationDetail, which renders three recent results and sends All runs to the web. The component still adds its own search, filters, paging, score projection, and store view machinery for an unreachable surface. Keep the shared recent-result renderer and remove the unreferenced component and exclusively dependent state rather than carrying the superseded UI alongside the daily workflow.
• Add data-model compatibility proof - The review found that existing stored data may not work after upgrade. Show that existing data still loads and works with this change.
• Explain or restore the original intent of Session-target payload selection - 61fcd39: The earlier Cron editor selected systemEvent/text for Main and agentTurn/message otherwise, matching the Gateway contract.

Findings

• [P2] Convert the payload when selecting the Main session — src/OpenClaw.Tray.WinUI/Automations/AutomationEditor.cs:550-552
• [P2] Match run completion to the accepted invocation — src/OpenClaw.Shared/Cron/AutomationsStore.cs:901-903
• [P2] Remove the unreachable full-history component — src/OpenClaw.Tray.WinUI/Automations/AutomationRuns.cs:22-29

Provenance

• Session-target payload selection changes intended behavior without addressing why it exists (61fcd39: The earlier Cron editor selected systemEvent/text for Main and agentTurn/message otherwise, matching the Gateway contract.)

Tests

• Missing end-to-end proof: Current-head native scheduling, Main-session creation, successive-run completion, and stable-release upgrade preservation remain unverified; historical screenshots and unrun regressions do not establish them.

Agent review details

How this fits together

The WinUI page sends authenticated scheduling requests through the shared Automations store and Gateway client, then renders jobs, recent results, and mutation feedback.

flowchart LR
  User[Operator] --> Page[Native Automations page]
  Page --> Store[Automations store]
  Store --> Client[Operator Gateway client]
  Client --> Gateway[Gateway scheduler]
  Gateway --> Overview[Overview and Inbox facts]
  Overview --> Page

Loading

Technical review

Best possible solution:

Land the daily-workflow replacement with valid session payloads, invocation-bound run tracking, and only reachable UI.

Do we have a high-confidence way to reproduce the issue?

Source establishes the Main-session failure against the released Gateway validator and the stale-history completion path; neither was executed during this read-only review.

Is this the best way to solve the issue?

The shared store and retained daily editor are appropriate, but session normalization and run identity need correction, and the superseded full-history component should be removed.

Full review comments:

• [P2] Convert the payload when selecting the Main session — src/OpenClaw.Tray.WinUI/Automations/AutomationEditor.cs:550-552
  Create a new automation with the default assistant task, enter a name and prompt, then select Main. This handler changes only SessionTarget; Normalize leaves AgentTurn unchanged and validation accepts it. Save therefore sends sessionTarget="main" with payload.kind="agentTurn", which the released Gateway rejects in assertSupportedJobSpec. The old editor generated systemEvent/text for Main. Restore that mapping in the existing form normalization and cover the serialized request.
  Confidence: 0.99
• [P2] Match run completion to the accepted invocation — src/OpenClaw.Shared/Cron/AutomationsStore.cs:901-903
  If run A completes while the cached job still reports an earlier LastRunAtMs, starting run B captures that stale baseline. B's watcher then accepts A's history entry because its timestamp exceeds the baseline, clearing B's running indicator before B finishes and enabling actions prematurely. The acknowledgement's RunId is currently used only for a notice. Retain it with the run owner and require matching completion, including in list reconciliation; timestamp-only completion must not settle an ID-less acknowledgement. The same distinction is documented in fix: keep older Run Now attempts from clearing newer runs #2082.
  Confidence: 0.96
• [P2] Remove the unreachable full-history component — src/OpenClaw.Tray.WinUI/Automations/AutomationRuns.cs:22-29
  The narrowed page never instantiates AutomationRuns: its only production use is the static RenderRun helper from AutomationDetail, which renders three recent results and sends All runs to the web. The component still adds its own search, filters, paging, score projection, and store view machinery for an unreachable surface. Keep the shared recent-result renderer and remove the unreferenced component and exclusively dependent state rather than carrying the superseded UI alongside the daily workflow.
  Confidence: 0.98

Overall correctness: patch is incorrect
Overall confidence: 0.95

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning medium; reviewed against 1c223f14f6b6.

Provenance checked

• Full run-response presentation keeps the original intent (638386e: Improve Cron run history UX #814 made complete sanitized responses available without leaving Cron history.)
• Legacy snapshot ownership changes intended behavior with a stated reason (ece8758: The previous AppState extraction centralized observable snapshots; this PR explicitly transfers scheduling ownership to the typed store and removes the old forwarding path.)
• Reactor dependency upgrade changes intended behavior with a stated reason (a4dd026: The preview.16 pin adopted the native session-switch workaround; this PR moves to official preview.17 packages for subsequent control fixes while retaining chat regression requirements.)

Testing

Proof path: none.

Security

None.

Evidence

Acceptance criteria:

• ./build.ps1
• dotnet test ./tests/OpenClaw.Shared.Tests/OpenClaw.Shared.Tests.csproj --no-restore
• dotnet test ./tests/OpenClaw.Tray.Tests/OpenClaw.Tray.Tests.csproj --no-restore
• dotnet test ./tests/OpenClaw.Shared.Tests/OpenClaw.Shared.Tests.csproj --no-restore --filter 'FullyQualifiedNameCron|FullyQualifiedNameAutomations'

What I checked:

• Pinned change and main behavior: The introduced diff replaces the legacy Cron implementation; current main still contains the old page. The feature is not already implemented on main. (src/OpenClaw.Tray.WinUI/Pages/CronPage.xaml.cs:65, f7ae188fe3ae)
• Previous findings reconciled: History requests omit view; removal performs a revision preflight followed by id-only removal; native validation blocks unsupported timing and Owner changes; expanded successful results use safe Markdown. The maintainer-authored description explicitly records the non-atomic removal tradeoff. (src/OpenClaw.Shared/Cron/CronGatewayRequests.cs:56, f7ae188fe3ae)
• Main-session regression: The new session selector changes SessionTarget without converting AgentTurn to SystemEvent. Main previously selected both the payload kind and text field from the session target. (src/OpenClaw.Tray.WinUI/Automations/AutomationEditor.cs:552, f7ae188fe3ae)
• Released Gateway contract: The PR executes this repository’s cron protocol. At stable v2026.11.2, assertSupportedJobSpec rejects agentTurn with sessionTarget=main; the closed request schemas also confirm the earlier view, removal, and delivery.target compatibility repairs. (src/cron/service/jobs-validation.ts:60, b556fdf457b0)
• Run completion identity: Polling accepts any history timestamp newer than the cached job baseline, although the run acknowledgement supplies a RunId. The related repair describes the same timestamp-versus-invocation distinction: fix: keep older Run Now attempts from clearing newer runs #2082. (src/OpenClaw.Shared/Cron/AutomationsStore.cs:902, f7ae188fe3ae)
• Proof and current checks: The four prepared screenshots were inspected and show the superseded split-pane UI. The full PR description explicitly leaves current native and stable-Gateway upgrade proof unverified. Exact-head fast-validation has passed; other required suites remain in progress. No target code, builds, or tests were run during this review. (f7ae188fe3ae)

Likely related people:

• Régis Brid: Raw commit 638386e adds src/OpenClaw.Shared/OpenClawGatewayClient.cs:785 relative to its recorded parents. This identifies author metadata, not feature responsibility or a PR merger. (role: source-line author; confidence: high; commits: 638386ed73ae; files: src/OpenClaw.Shared/OpenClawGatewayClient.cs)

Preflight source

Source unknown. Source comparison at the observed PR snapshot only. Completion is self-reported; body, policy, tests, execution, and report bytes are not verified.

Review metrics

Metric	Value	Why it matters
Source and test delta	src +11,901/-4,579; tests +9,174/-813	The replacement includes substantial localization and native coverage; the unreachable full-history component is avoidable production growth.

Labels

Label changes:

No label changes.

Label justifications:

• P2: Native scheduling is a useful improvement, with bounded editor and running-state defects.
• merge-risk: 🚨 compatibility: The replacement changes released-Gateway request shapes and persisted-job updates; selecting Main currently constructs a rejected job definition.
• rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🦐 gold shrimp.
• status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action.

Rating scale

6/6 🦀 challenger crab · 5/6 🦞 diamond lobster · 4/6 🐚 platinum hermit · 3/6 🦐 gold shrimp · 2/6 🦪 silver shellfish · 1/6 🧂 unranked krab. Overall follows the weaker of proof and patch quality; ✨ marks media proof (a screenshot, video, or linked artifact) that directly shows the changed behavior.

Workflow

ClawSweeper edits this one comment on every review. Comment @clawsweeper re-review for a fresh review only. Repair and merge execution lanes are retired.

History

Review history (1 earlier review cycle)

• reviewed 2026-10-27T19:13:43.567Z sha 97bbae6 :: blocked before merge. :: [P1] Keep history requests compatible with the released Gateway | [P1] Preserve removal on Gateways without revisioned deletion | [P2] Check Gateway support before offering new scheduling fields | [P2] Retain Markdown rendering for successful full responses

Reviewed November 9, 2026, 6:13 PM ET / 23:13 UTC (Revision 2).

[bkudiess]

Native Automations proof

Windows 11 ARM64, isolated synthetic Gateway. No production jobs or provider calls. The app source is unchanged since 776f8df3; the later commits change only the proof tests. Current-head source correspondence and executable/library SHA-256 hashes were recorded locally.

Debug proof: Light and Dark populated states, independent Light/Dark disconnected Axe scans, and the actual final message in 240-message chat histories at both window widths: 5 passed, 0 failed, 0 skipped.

Details in Light

Details in Dark

Other main states, editor and confirmation

List

Run history and transcript

All runs

Editor in Light

Editor in Dark

Pause confirmation

Empty agent

Disconnected recovery

[bkudiess]

Real Windows High Contrast (Night sky), production-shaped Release ARM64, fixture Gateway only. The test recorded Windows high contrast: True and passed all main states plus populated Axe checks (1 passed, 0 failed, 0 skipped). Both stable disconnected Axe cases also passed in the preceding High Contrast run. The original None contrast setting was restored and verified. An earlier duplicate Light case completed its UI checks but failed on temporary-profile cleanup; that failed attempt remains recorded locally.

[bkudiess]

Automations parity checklist (web → native)

Reference: openclaw/openclaw@5089aa71 ui/src/pages/cron/, ui/src/lib/cron/.
Status: ☐ planned · ☑ done · ⏸ deferred (with reason).

#	Web item	Native equivalent	Data source	Status
1	Job list scoped to the selected agent	Left list pane; agent:{id}:cron passes agentId	cron.list { agentId }	☑
2	Search	Search box (debounced)	cron.list { query }	☑
3	All / Active / Paused	SelectorBar	cron.list { enabled, includeDisabled }	☑
4	Sort (next run, recently updated, name; asc/desc)	Sort DropDownButton with radio items	cron.list { sortBy, sortDir }	☑
5	Advanced filters (schedule kind, last status, condition)	Filter flyout	cron.list { scheduleKind, lastRunStatus, trigger }	☑
6	Paging / load more, shown of total	"Load more" + count	cron.list { limit, offset } → hasMore,nextOffset,total	☑
7	Scheduler status, next wake, totals, failing	Status strip at top	cron.status, scoped cron.list	☑
8	Job row: state, name, agent, schedule, next/last run, trigger	List card	CronJob	☑
9	Job details	Detail pane, Overview tab	cron.get	☑
10	Pause / resume (disable confirms)	Native action buttons + confirmation for pause	cron.update {enabled, expectedConfigRevision}	☑
11	Run now / run if due, running state, feedback	Run split button, running badge	cron.run {id, mode}	☑
12	Edit	Editor pane	cron.update	☑
13	Clone	More menu → prefilled editor	cron.add	☑
14	Remove (confirm)	More menu → ContentDialog	cron.remove	☑
15	Schedule: once, interval, cron, event; on-exit/stream read-only	Schedule editor	schedule object	☑
16	Time zone	Cron time zone box	schedule.tz	☑
17	Stagger window / exact timing	Exact toggle + amount/unit	schedule.staggerMs	☑
18	Active hours	Toggle + start/end + tz	activeHours	☑
19	Readable schedule summary	Live summary under editor + list	local	☑
20	Assistant task prompt / timeline message	Action section	payload	☑
21	Model, thinking, timeout	Editable combos + number	payload.model/thinking/timeoutSeconds	☑
22	Light context, include reasoning, skip empty scratch	Advanced toggles	payload flags	☑
23	Read-only command/script payloads	Read-only block	payload	☑
24	Session target, wake mode, idle only, delete after run	Advanced	job fields	☑
25	Condition trigger script	Advanced	trigger	☑
26	Delivery: mode none/announce/webhook	Delivery section	delivery.mode	☑
27	Target conversation (suggestions)	Editable combo of conversations	conversations.list	☑
28	Channel, account, owner DM, direct policy, best effort	Delivery section	delivery.*	☑
29	Webhook URL + validation	Delivery section	delivery.to	☑
30	Failure alerts inherit/disabled/custom	Failure alerts section	failureAlert	☑
31	Event-triggered jobs from MCP servers	Event schedule: server, event, args JSON	config.get (mcp.servers), mcp.events.list	☑ (live mcp-events.status subscription diagnostics ⏸ deferred: needs a long-lived event subscription the page host does not keep; definitions and configured servers are shown)
32	Per-job scratch	Scratch tab (load, save, remove, size, conflict)	cron.scratch.get/set	☑
33	Run history per job + all runs (Activity)	Runs tab + Run history view	cron.runs	☑
34	Run filters: search, status, delivery, sort, load more	Filter bar	cron.runs	☑
35	Run status, delivery outcome, suppression, error, model, tokens	Run card	CronRunLogEntry	☑
36	Run transcript	Transcript panel with paging	cron.history	☑
37	Condition activity (checks, last checked, last fired)	Runs tab header	job state	☑
38	Auto-disabled status	Status pill + InfoBar	state.autoDisabled	☑
39	Suggested schedules / starter automations	Empty state cards	local	☑
40	Non-admin limited actions	Read-only banner, mutations hidden	operator scopes	☑
41	Revision conflict handling	Reload latest + InfoBar	CRON_JOB_CHANGED	☑
42	Empty / loading / error / no-match states	Native states	—	☑
43	Disconnected + Open Connection	InfoBar + empty state	connection	☑
44	One-time job finished notification (Windows)	InfoBar + app notification	run-now watcher + cron.get not-found	☑

Screenshots: Light/Dark main-state gallery and actual Windows High Contrast.

This is the implementation inventory, not a claim that every control has separate live-provider proof. Runtime proof uses the isolated fixture Gateway.

[roboclaw-bot]

PR cleanup review

Reviewed head: d11898a373bece41ced81b6f476abb8f860adf20. Main comparison: e59374c064f43d68f5c72d070d4c89ef2d20ec3b.

Thanks for the work on this. I would keep #1893 (feat(automations): native Automations page with web parity) open as a draft, rather than close it as superseded: pinned main still has the legacy Cron page, and this branch has a distinct native daily-workflow implementation.

At d11898a, the latest CI run has completed, not remained pending: the production-resource lane passed 31/31, while the UI and Mock Gateway jobs failed on reader scrolling, a Usage title-bar peer and composer readiness. Those failures need attribution before rerunning; this read-only pass did not establish baseline flakes.

The next useful pass is focused lifecycle and proof work: bind the page store to its originating connection, cover omitted-owner cloning and multiple expanded rows across reconnect, and preserve visible results when persistent error toasts fill the queue. Please also keep capture observational and restore the native timezone-transition matrix. These are source-backed concerns, not an executed Windows reproduction.

Please refresh the body to the actual head and current daily-workflow scope, reconcile the older review against the repairs already present, and attach current-head native/compatibility evidence before moving out of draft. Historical screenshots and the old parity checklist are useful provenance, but not proof of this UI.

Action in this cleanup pass: leaving this PR open. No branch changes or runtime-proof waiver. The findings above come from source, discussion and recorded-CI review; Windows validation was not rerun in this pass.