Human-gated AI development workflows

Worker dispatch ready

KAN-17Review gate

Implementer automatisk PR merge gate efter godkendt workflow review

Repository

danubii/openclaw-orchestrator

ai/KAN-17-implementer-automatisk-pr-merge-gate-efter-godke

Created

Apr 26, 09:36 AM UTC

Updated Apr 26, 10:04 AM UTC

Owner

dev-user

Review Approval / Review approval

Mode

Feature

PlanWaitForApproval

Run timeline

9/14 steps complete

Ticket Intake

Apr 26, 09:36 AM UTCDuration 2s

Completed

Context Build

Apr 26, 09:38 AM UTCDuration 2m 00s

Completed

Codex Plan

Apr 26, 09:38 AM UTCDuration 2m 00s

Completed

Plan Approval

Apr 26, 09:39 AM UTCDuration 55s

Completed

Codex Implementation

Apr 26, 09:55 AM UTCDuration 15m 59s

Completed

Test Run

Apr 26, 09:58 AM UTCDuration 2m 19s

Completed

Cross Repository Integration Test

Apr 26, 09:36 AM UTCDuration 0s

Completed

Smoke Test

Apr 26, 10:01 AM UTCDuration 2m 49s

Completed

Claude Review

Apr 26, 10:04 AM UTCDuration 3m 34s

Completed

Review Approval

Apr 26, 10:04 AM UTCElapsed 5h 48m 06s

Pending

Git Push Branch

Queued

Pending

Test Deploy

Queued

Pending

Manual Qa

Queued

Pending

Pull Request

Queued

Pending

Artifact viewer

claude-agent-log.txt · v3

Loading artifact

Lifecycle

Active workflow

Gate control

Review Approval

PlanningImplementation

Decision comment

6 blocker/high review finding(s) must be returned or resolved.

5 medium review finding(s) will be overridden by approval.

Approval history

PlanApproved

dev-user / Apr 26, 09:39 AM UTC

Repositories

1 target

danubii/openclaw-orchestrator

ai/KAN-17-implementer-automatisk-pr-merge-gate-efter-godke

Completed

Target: main

Review findings

13 total / 11 priority

Manual override API/UX not implemented

Blocker

src/OpenClaw.Orchestrator.Infrastructure/Services/WorkerJobProcessor.cs

The AC requires a complete override mechanism: an API endpoint for submitting overrides, enforcement that only overrideable checks can be overridden, mandatory comment, and audit log entry. The implementation only stores override fields in the database and a toggle in settings. No endpoint, no check-policy engine, no audit write.

Implement POST /api/workflows/{id}/merge-gate/override with required comment body, audit entry write to MergeGateRun, and enforcement that MergeGateStatus must be Blocked and AllowManualOverride must be true for the target check.

No dedicated merge gate unit/integration tests

Blocker

tests/OpenClaw.Orchestrator.Application.Tests/WorkflowOrchestratorTests.cs:1901-1904

The Jira AC requires tests for: successful merge, failed GitHub checks, timeout, PR conflict, head SHA changed, idempotent PR update, and delete-branch setting. The diff contains zero new test methods targeting these scenarios. Only test-fake stubs were extended to satisfy the updated interface contracts.

Add WorkerJobProcessor test scenarios using FakeWorkspaceContainerRunner and configurable FakeGitHubClient that returns controlled responses (failed checks, SHA mismatch, conflict). Cover all seven AC-listed scenarios before merge.

GitHub checks polling blocks worker thread up to 15 minutes

High

src/OpenClaw.Orchestrator.Infrastructure/Services/WorkerJobProcessor.cs

WaitForGitHubChecksAsync loops with Task.Delay(10s) for up to GitHubChecksTimeoutSeconds (default 900s) inside a worker job. This occupies a worker slot continuously, starving other jobs under the default ConcurrentWorkers cap.

Extract polling to a separate re-enqueueing WorkerJobType (e.g. PollGitHubChecks) that runs once, checks state, and re-enqueues itself with AvailableAt delay if still pending. This releases the worker slot between polls.

MergeGate workflow step always created regardless of AutoMergeEnabled

High

src/OpenClaw.Orchestrator.Infrastructure/Services/WorkflowOrchestrator.cs:306-312

WorkflowOrchestrator.CreateWorkflowAsync unconditionally adds a pending MergeGate step to every workflow. When auto merge is disabled, this step remains Pending forever, creating a misleading UI and incorrect workflow state for all manually-managed PRs.

Either conditionally add the step only when AutoMergeEnabled is true at workflow creation time, or mark it Skipped automatically in the GitPushBranch completion path when auto merge is off.

QualityGateTimeoutSeconds stored but never applied

High

src/OpenClaw.Orchestrator.Infrastructure/Services/WorkerJobProcessor.cs

MergeGateSettingsDto.QualityGateTimeoutSeconds is persisted and exposed in the settings UI, but RunFinalQualityGateAsync uses hard-coded TimeSpan.FromMinutes(5) and TimeSpan.FromMinutes(2) for container commands. The setting is effectively dead configuration.

Pass settings.QualityGateTimeoutSeconds into RunFinalQualityGateAsync and apply it as the overall deadline, splitting proportionally across the clone and scan steps, or use it as the timeout for the container runner request.

RequiredGitHubChecks null dereference when saving settings

High

src/OpenClaw.Orchestrator.Infrastructure/Services/IntegrationSettingsService.cs

UpdateAsync passes request.MergeGate.RequiredGitHubChecks (type string?) directly to UpsertAsync without null guard. A null value will cause a NullReferenceException at runtime when a user clears the field.

Change the call to: await UpsertAsync(MergeGateRequiredGitHubChecks, request.MergeGate.RequiredGitHubChecks ?? "", updatedByUserId, now, cancellationToken);

Completed MergeGateRun can be overwritten if job re-enqueues

Medium

src/OpenClaw.Orchestrator.Infrastructure/Services/WorkerJobProcessor.cs

RunMergeGateAsync finds existing runs with FirstOrDefaultAsync without filtering on status. If a Completed gate run exists and the job somehow re-fires, it will reset the completed run's status to Running, losing the audit record.

Add a status filter to the lookup: `.FirstOrDefaultAsync(x => ... && x.Status != MergeGateStatus.Completed)`. If only Completed runs exist, create a new run for audit trail purposes.

Dashboard mergeGates metric counts all completed-via-merge-gate workflows

Medium

web/features/workflows/workflow-dashboard.tsx:62

The filter `workflow.mergeGateStatus` is truthy for any workflow in Completed status (DtoMappings sets mergeGateStatus = "Completed" for Completed workflows). The metric grows unboundedly with completed workflows, not just active/blocked gates.

Filter only on active gate states: `workflow.status === 'MergeGateRunning' || workflow.status === 'MergeGateBlocked'`.

no-fake scan regex matches legitimate code (stub, placeholder in package names)

Medium

src/OpenClaw.Orchestrator.Infrastructure/Services/WorkerJobProcessor.cs

The quality gate runs `grep -i -E '\b(fake|stub|placeholder|todo: fake)\b'`. The word 'stub' matches gRPC stubs, network stub libraries, and test fixtures with legitimate names. This will produce false-positive failures on many real codebases.

Restrict the pattern to explicit markers such as `// FAKE:`, `// STUB:`, `// TODO: fake`, `// PLACEHOLDER:` rather than bare word matching. Alternatively, add a `.openclaw-no-fake-ignore` mechanism similar to .gitignore.

PullRequests.State migration default '' diverges from entity default 'Open'

Medium

src/OpenClaw.Orchestrator.Infrastructure/Persistence/Migrations/20260426094953_AddMergeGate.cs

The migration adds State with defaultValue: "" but PullRequestRecord.State defaults to "Open". Existing PR rows after migration will have empty string state, breaking any string comparisons against "Open" or "open".

Change defaultValue: "" to defaultValue: "Open" in the migration, or add an UPDATE statement in the Up() method to backfill existing rows.

workflow.currentGate references non-existent TypeScript field

Medium

web/features/workflows/workflow-detail-client.tsx:861

workflow-detail-client.tsx renders `latestMergeGate?.phase ?? workflow.currentGate`. WorkflowDetail has no currentGate property; this evaluates to undefined at runtime, silently rendering nothing in the fallback branch. Dead code that could confuse future readers.

Remove `?? workflow.currentGate`. If a phase fallback is needed, use a static string like `?? "Unknown"` or omit the line entirely.

GitHubApiSnapshot artifact type defined but never written

Low

src/OpenClaw.Orchestrator.Domain/Enums/ArtifactType.cs

ArtifactType.GitHubApiSnapshot was added to the enum and is referenced in artifactStepType() in the frontend, but no code path in WorkerJobProcessor writes an artifact of this type. The plan mentioned storing sanitized API responses for debugging.

Either implement writing sanitized GitHub API responses (PR details, checks result) as GitHubApiSnapshot artifacts in RunMergeGateAsync, or remove the type and frontend mapping until implemented.

No new Playwright E2E tests for merge gate UI

Low

web/tests/e2e

The AC requires an e2e test covering the flow against mocked GitHub API. The 4 passing Playwright tests are pre-existing. No new assertions cover the merge gate panel, settings UI, or dashboard metric added by this PR.

Add Playwright tests for: (1) merge gate panel renders on workflow detail when status is MergeGateRunning, (2) merge gate settings section saves and reflects correctly, (3) dashboard metric shows correct count for active gates.

Live logs

100 entries / offline

10:00:06smoke|■■■■■■■■ | 10% of 170.4 MiB

10:00:06smoke|■■■■■■■■■■■■■■■■ | 20% of 170.4 MiB

10:00:07smoke|■■■■■■■■■■■■■■■■■■■■■■■■ | 30% of 170.4 MiB

10:00:07smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 40% of 170.4 MiB

10:00:08smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 50% of 170.4 MiB

10:00:08smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 60% of 170.4 MiB

10:00:08smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 70% of 170.4 MiB

10:00:09smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 80% of 170.4 MiB

10:00:10smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 90% of 170.4 MiB

10:00:10smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■| 100% of 170.4 MiB

10:00:12smokeChrome for Testing 147.0.7727.15 (playwright chromium v1217) downloaded to /workspace/repo-10-danubii-openclaw-orchestrator/web/node_modules/playwright-core/.local-browsers/chromium-1217

10:00:13smokeDownloading FFmpeg (playwright ffmpeg v1011) from https://cdn.playwright.dev/dbazure/download/playwright/builds/ffmpeg/1011/ffmpeg-linux.zip

10:00:19smoke| | 0% of 2.3 MiB

10:00:19smoke|■■■■■■■■ | 10% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■ | 20% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■ | 30% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 40% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 50% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 60% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 70% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 80% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 90% of 2.3 MiB

10:00:19smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■| 100% of 2.3 MiB

10:00:19smokeFFmpeg (playwright ffmpeg v1011) downloaded to /workspace/repo-10-danubii-openclaw-orchestrator/web/node_modules/playwright-core/.local-browsers/ffmpeg-1011

10:00:19smokeDownloading Chrome Headless Shell 147.0.7727.15 (playwright chromium-headless-shell v1217) from https://cdn.playwright.dev/builds/cft/147.0.7727.15/linux64/chrome-headless-shell-linux64.zip

10:00:26smoke| | 0% of 112 MiB

10:00:27smoke|■■■■■■■■ | 10% of 112 MiB

10:00:27smoke|■■■■■■■■■■■■■■■■ | 20% of 112 MiB

10:00:28smoke|■■■■■■■■■■■■■■■■■■■■■■■■ | 30% of 112 MiB

10:00:28smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 40% of 112 MiB

10:00:28smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 50% of 112 MiB

10:00:29smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 60% of 112 MiB

10:00:29smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 70% of 112 MiB

10:00:29smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 80% of 112 MiB

10:00:30smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ | 90% of 112 MiB

10:00:30smoke|■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■| 100% of 112 MiB

10:00:31smokeChrome Headless Shell 147.0.7727.15 (playwright chromium-headless-shell v1217) downloaded to /workspace/repo-10-danubii-openclaw-orchestrator/web/node_modules/playwright-core/.local-browsers/chromium_headless_shell-1217

10:00:31smoke> /workspace/repo-10-danubii-openclaw-orchestrator/web/node_modules/.bin/playwright test

10:00:43smokeRunning 4 tests using 2 workers

10:00:51smokestderr: [WebServer] [36m[browser][39m [31mUncaught Error: Hydration failed because the server rendered text didn't match the client. As a result this tree will be regenerated on the client. This can happen if a SSR-ed Client Component used:

10:00:51smokestderr: [WebServer]

10:00:51smokestderr: [WebServer] - A server/client branch `if (typeof window !== 'undefined')`.

10:00:51smokestderr: [WebServer] - Variable input such as `Date.now()` or `Math.random()` which changes each time it's called.

10:00:51smokestderr: [WebServer] - Date formatting in a user's locale which doesn't match the server.

10:00:51smokestderr: [WebServer] - External changing data without sending a snapshot of it along with the HTML.

10:00:51smokestderr: [WebServer] - Invalid HTML tag nesting.

10:00:51smokestderr: [WebServer]

10:00:51smokestderr: [WebServer] It can also happen if the client has a browser extension installed which messes with the HTML before React loaded.

10:00:51smokestderr: [WebServer]

10:00:51smokestderr: [WebServer] https://react.dev/link/hydration-mismatch

10:00:51smokestderr: [WebServer]

10:00:51smokestderr: [WebServer] ...

10:00:51smokestderr: [WebServer] <SegmentViewNode type="page" pagePath="workflows/...">

10:00:51smokestderr: [WebServer] <SegmentTrieNode>

10:00:51smokestderr: [WebServer] <WorkflowDetailPage>

10:00:51smokestderr: [WebServer] <AppShell>

10:00:51smokestderr: [WebServer] <div className="min-h-scre...">

10:00:51smokestderr: [WebServer] ...

10:00:51smokestderr: [WebServer] <div className="mx-auto ma...">

10:00:51smokestderr: [WebServer] <header>

10:00:51smokestderr: [WebServer] <WorkflowDetailClient workflow={{id:"workfl...", ...}}>

10:00:51smokestderr: [WebServer] <div className="grid min-h...">

10:00:51smokestderr: [WebServer] <section className="rounded-lg...">

10:00:51smokestderr: [WebServer] <PanelHeader>

10:00:51smokestderr: [WebServer] <div className="p-3">

10:00:51smokestderr: [WebServer] <div>

10:00:51smokestderr: [WebServer] <div className="grid grid-...">

10:00:51smokestderr: [WebServer] <div>

10:00:51smokestderr: [WebServer] <div className="rounded-md...">

10:00:51smokestderr: [WebServer] <div className="flex items...">

10:00:51smokestderr: [WebServer] <div className="min-w-0">

10:00:51smokestderr: [WebServer] <div>

10:00:51smokestderr: [WebServer] <div className="mt-1 flex ...">

10:00:51smokestderr: [WebServer] <span>

10:00:51smokestderr: [WebServer] + Queued 24h 30m 51s · Apr 25, 09:30 AM UTC

10:00:51smokestderr: [WebServer] - Queued 24h 30m 50s · Apr 25, 09:30 AM UTC

10:00:51smokestderr: [WebServer] ...

10:00:51smokestderr: [WebServer] [39m

10:00:51smokestderr: [WebServer] [31m at <unknown> (https://react.dev/link/hydration-mismatch)

10:00:51smokestderr: [WebServer] at span (<anonymous>)

10:00:51smokestderr: [WebServer] at <unknown> (features/workflows/workflow-detail-client.tsx:561:25)

10:00:51smokestderr: [WebServer] at Array.map (<anonymous>)

10:00:51smokestderr: [WebServer] at WorkflowDetailClient (features/workflows/workflow-detail-client.tsx:548:24)

10:00:51smokestderr: [WebServer] at WorkflowDetailPage (app/workflows/[id]/page.tsx:54:9)[39m

10:00:51smokestderr: [WebServer] 559 | <div className="truncate text-sm font-medium text-zinc-100">{splitS...

10:00:51smokestderr: [WebServer] 560 | <div className="mt-1 flex flex-wrap items-center gap-x-2 gap-y-1 te...

10:00:51smokestderr: [WebServer] > 561 | <span>{stepTimelineAnchor(step)}</span>

10:00:51smokestderr: [WebServer] | ^

10:00:51smokestderr: [WebServer] 562 | {timing ? (

10:00:51smokestderr: [WebServer] 563 | <>

10:00:51smokestderr: [WebServer] 564 | <span className="size-1 rounded-full bg-zinc-700" />

10:01:00smoke····

10:01:00smoke 4 passed (28.4s)

10:01:01smokeSmokeTest completed with exit code 0.

10:01:01workerQueued ClaudeReview.

10:01:02workerWorker claimed ClaudeReview job 3ac2de03-d0a3-4c03-9a6e-a92bee1274c1.

10:04:36claudeClaudeReview completed with exit code 0.

Feedback loop

13 review findings

Implementation feedback