Skip to content

Roadmap status

Ownership tracking lives in GitHub Issues, not on this page

Every open item (status Proposal or In progress) has a matching GitHub issue, and that issue's Assignees — not this dashboard or any file in the repo — are the single source of truth for who, if anyone, is working on it. Browse issues labeled roadmap-tracking: no:assignee for the unclaimed backlog, assignee:<user> for one person's plate. See BE-0109 for how the sync works.

Live view of every roadmap (BE) item, grouped by category — each category showing the share of its outstanding items already implemented, and each card its own status. Regenerated from item metadata on every docs build, so it always reflects the committed roadmap. Categories with no outstanding work are grouped separately under Completed. Categories start collapsed to a progress overview — click a heading to expand it, toggle the status chips on and off, or type in the search box to narrow the cards by id, title, topic, or status. Switch between the card grid and a sortable table with the Cards / Table toggle — the table lists every item as a row with sortable Created and Updated columns, and the search and status filters narrow both views alike. A third view, Map, draws the relationships the items themselves record as a transit map: one line per topic, its items as stops along it in id order, and a curve joining every pair that stands in a Related, Origin, or Superseded by relation. Point at a stop to light up what it connects to and see a card with its title, a short excerpt from its Introduction, and a clickable link to each item it connects to; the Zoom and Full size controls above the map scale it up and expand it to fill the browser window. Each card links to its full proposal on GitHub. This dashboard is the only status view — for what a roadmap item is and how to add one, see roadmaps/README.md (both languages).

In progress

Platform support

90%65/72 implemented
BE-0005Implementedidb_companion version monitoringIssue
BE-0006Implementedidb element-tree normalization accuracyIssue
BE-0007ImplementedAndroid backendIssue
BE-0008ImplementedFlutter supportIssue
BE-0010ImplementedUpdate the scope statementIssue
BE-0019ImplementedXCUITest backendIssue
BE-0020ImplementedMulti-backend evidence fallbackIssue
BE-0037ImplementedWebView / hybrid supportMagicPodIssue
BE-0041ImplementedWeb (Playwright) backendIssue
BE-0054ImplementedWeb backend completion (rich capabilities & parallel runs)Issue
BE-0076ImplementedSelectable browser engines & cross-browser compatibility matrix (web backend)Issue
BE-0087Implementedidb action timing robustness (settle before actuation)Issue
BE-0088ImplementedOverlap the Simulator boot with the buildIssue
BE-0105ImplementedSingle-snapshot element query for XCUITestIssue
BE-0207ImplementedMake the XCUITest runner channel robust to transient timeoutsIssue
BE-0208ImplementedAndroid on-device e2e in CI (emulator via KVM)Issue
BE-0210ImplementedAndroid on-device actuation fidelityIssue
BE-0211ImplementedAndroid device control (setLocation, clipboard)Issue
BE-0218ImplementedStabilize the E2E Simulator gate: namespace-aware readiness and a bounded actuation timeoutIssue
BE-0221ImplementedGuarantee shared showcase scenarios run unchanged on AndroidIssue
BE-0223ImplementedReach every Android tab by driving the tab bar over adbIssue
BE-0227ImplementedWeb swipe scroll fidelity (mode-aware scroll dispatch)Issue
BE-0228ImplementedWeb device mode (desktop / mobile emulation)Issue
BE-0231ImplementedHarden the E2E first-wait against Simulator settling flakeIssue
BE-0232ImplementedMulti-touch gestures on the adb driver (pinch / rotate)Issue
BE-0233Implementedadb clipboard on-device fidelityIssue
BE-0234ImplementedSpeed up adb scenario runs (uiautomator dump bottleneck)Issue
BE-0240ImplementedCapability-aware automatic actuator selection for iOS (idb/XCUITest transparency)Issue
BE-0245ImplementedResident UI Automator server for adb readsIssue
BE-0269ImplementedSpeed up the system-alert guard's intervention during wait stepsIssue
BE-0271ImplementedStructural parity across platform E2E workflowsIssue
BE-0279ImplementedAlign required E2E checks across every backendIssue
BE-0281ImplementedAdd real on-device actuation coverage to the iOS CIIssue
BE-0283ImplementedNetwork-capture assertions for the Android backendIssue
BE-0287ImplementedXCUITest runner-channel resilience under multi-touch actuationIssue
BE-0289ImplementedMake the XCUITest channel re-resolve a stale actuation handle before failingIssue
BE-0290ImplementedMake XCUITest the default iOS backend and retire idbIssue
BE-0291ImplementedReuse the XCUITest runner across scenarios to amortize cold startupIssue
BE-0292ImplementedBundle the XCUITest runner so testRunner is optionalIssue
BE-0310ImplementedUse iOS accessibility screen-change notifications to make the readiness and settle waits more accurateIssue
BE-0312ImplementedDerive XCUITest actuation handles from element identity so an unchanged screen keeps its handles validIssue
BE-0315ImplementedMake the reactive alert guard deterministic and native, reusing BE-0316's SpringBoard pathIssue
BE-0319ImplementedMake the XCUITest cold runner spawn diagnosable and self-healingIssue
BE-0320ImplementedMake the iOS system-alert button selector deterministic under a non-English Simulator localeIssue
BE-0323ImplementedRecover the XCUITest cold launch when the runner crashes during the readiness gateIssue
BE-0334ImplementedGive the on-device conformance suite the infrastructure-fault recovery the run pipeline already hasIssue
BE-0342ImplementedGive the on-device suites' lease a teardown that reaches the runnerIssue
BE-0344ImplementedRepair the Simulator between XCUITest cold-spawn retry attemptsIssue
BE-0350ImplementedGive the on-device conformance and fault-injection suites the video and deviceLog evidence every scenario-driven CI job already getsIssue
BE-0353ImplementedForce device recovery on a backend-crash retry and cap total crash-recovery time per runIssue
BE-0354ImplementedDetect a wedged XCUITest session fast and escalate a repeated crash retry to a replacement deviceIssue
BE-0357DeferredDrop XCUITest's non-hittable duplicate accessibility node when exactly one is tappableIssue
BE-0358ImplementedCarry the device OS version as a parsed fact, so a cross-version run is not scored as flakyIssue
BE-0359ImplementedWait for the Simulator to finish booting before installing and launching the appIssue
BE-0360ImplementedRe-ask the runner's liveness while waiting out a mid-run crashIssue
BE-0362In progressGive the runner's HTTP server queues an explicit quality of serviceIssue
BE-0363ImplementedBound every simctl call with a timeout so a wedged CoreSimulator fails with a named causeIssue
BE-0367ImplementedCollect the layered diagnostics an Android CI failure needsIssue
BE-0368ImplementedResolve a content-identical duplicate node pair to one live element instead of failing the lookupIssue
BE-0369ImplementedExtend the iOS system-alert prompt/choice table to the paste-consent promptIssue
BE-0372ImplementedHarden the runner's HTTP server against the connection failures a driver timeout producesIssue
BE-0373ImplementedOffer the one reachable named descendant when a container refuses a tapIssue
BE-0374ProposalFail the scenario when a crash retry's device prep times out, instead of retrying onto the wedged deviceIssue
BE-0376ProposalQuarantine a device whose capture pipeline stalled, even when the scenario failed on its own assertionIssue
BE-0378ImplementedStop a wedged-Simulator timeout from failing the on-device conformance suite as a verdictIssue
BE-0381In progressGenerate the XCUITest runner's HTTP server from an OpenAPI contract instead of hand-rolling itIssue
BE-0389ImplementediOS native TipKit tip guardIssue
BE-0396ImplementedRead SFSafariViewController's element tree from the process that draws itIssue
BE-0397ImplementedRequire the tip container before the TipKit guard dismisses a popoverIssue
BE-0399ImplementedAnswer an interrupting system alert only by the scenario's own policyIssue
BE-0405ProposalIdentifierTool: a standalone Android identifier libraryIssue
BE-0406ProposalDeclare system alerts by prompt alone, and answer them during a handleSystemAlert stepIssue

Driver & backend architecture

90%18/20 implemented
BE-0009ImplementedCross-platform abstractionsIssue
BE-0042ImplementedPlatform-aware backend registry & selectionIssue
BE-0057ImplementedRename the config `apps` key to `targets`Issue
BE-0082ImplementedPreflight capability check before a runIssue
BE-0114ImplementedDriver conformance suite for backend-agnostic behaviorIssue
BE-0118ImplementedUnify the wait_for polling contract across driversIssue
BE-0126ImplementedSplit Effective into per-platform configsIssue
BE-0128ImplementedPreflight-gate device-control steps by capabilityIssue
BE-0141ImplementedBring backend lifecycle into the type systemIssue
BE-0212ImplementedSplit the coarse deviceControl capability into per-operation tokensIssue
BE-0270ImplementedDriver conformance for the adb backend on-deviceIssue
BE-0280ImplementedExtend the driver conformance contract to unactuated Driver operationsIssue
BE-0299ImplementedClose the read-race gap in mid-scenario reads and idb's settle timingIssue
BE-0305ImplementedReal-device fault-injection coverage for driver resilience pathsIssue
BE-0332ImplementedApply a read-lag barrier to every read path that mistakes a stable tree for a current oneIssue
BE-0339In progressResolve and actuate on the device so an Android gesture never aims at a stale coordinateIssue
BE-0349ImplementedVerify tappability before acting, with a bounded scroll safety netIssue
BE-0351ImplementedGate the adb settle poll's fast path on a key it actually proved stableIssue
BE-0355ImplementedSurface each element's real Z position via an opt-in app-side SDKIssue
BE-0364RejectedAdd a control channel so bajutsu can hide the touch visualization for a single captureIssue
BE-0365In progressGive bajutsu a channel into the running app so an in-app capability can be toggled mid-scenarioIssue

Device-cloud execution

57%4/7 implemented
BE-0235ImplementedAWS Device Farm batch submitterIssue
BE-0236ImplementedDevice-cloud provider abstractionIssue
BE-0237ProposalFirebase Test Lab / Device Streaming adapterIssue
BE-0238ImplementediOS device-cloud executionIssue
BE-0288In progressiOS device-signing build for the batch routeIssue
BE-0303ProposalReal-grid verification of the XCUITest live device-cloud routeIssue
BE-0336Implementedserve-driven Device Farm dispatch with bounded per-scenario fan-outIssue

Scenario authoring features

92%24/26 implemented
BE-0030ImplementedParameterized shared stepsMagicPodIssue
BE-0031ImplementedData-driven scenariosMagicPodIssue
BE-0032ImplementedSecret variablesMagicPodIssue
BE-0033ImplementedScenario variables + light control flowMagicPodIssue
BE-0034ImplementedTags / labels + selective runsMagicPodIssue
BE-0035ImplementedDevice-control steps (background, status-bar override)MagicPodIssue
BE-0036ImplementedHTTP utility stepMagicPodIssue
BE-0046ImplementedOTP & email side-channel stepsMagicPodIssue
BE-0052ImplementedDevice-state primitives: timezone, clipboard, shakeMagicPodIssue
BE-0157DeferredShake device primitiveMagicPodIssue
BE-0158DeferredTimezone device primitiveMagicPodIssue
BE-0186ImplementedMailbox provider registry for the email stepIssue
BE-0265ImplementedText-editing steps: select, clear, delete, copyIssue
BE-0276ImplementedDeclarative per-scenario permission state (simctl privacy / pm grant)MaestroIssue
BE-0314ImplementedDeterministic interrupt handlers for unpredictable interstitial screensIssue
BE-0316ImplementedExplicit mid-flow step for iOS permission-prompt alertsIssue
BE-0317ImplementedRename the dismissAlerts guard to alertHandling to match its grant-or-dismiss behaviorIssue
BE-0326ImplementedThe `scroll` action: scroll until an element appearsIssue
BE-0327ImplementedRename the alertHandling guard to systemAlertHandling to name what it handlesIssue
BE-0329ImplementedDecide from observed element motion whether `scroll`'s region stopped or overshotIssue
BE-0356ImplementedAdd a deterministic pickerWheel value-setting step for XCUITestIssue
BE-0377ImplementedRandom and datetime value generationIssue
BE-0382ImplementedLet the reactive system-alert guard answer each covered prompt by its own ruleIssue
BE-0392ImplementedIndependent before/after lifecycle hooks for scenariosIssue
BE-0400ImplementedMake a scroll step travel the distance it asks for and let authors choose itIssue
BE-0401ImplementedConsolidate the systemAlertHandling DSL into one key per answer pathIssue

Verification & coverage

91%21/23 implemented
BE-0001ImplementedDeterministic runner (M1)Issue
BE-0028ImplementedGuard against over-matching evidence rulesIssue
BE-0029ImplementedVisual-regression assertionsBothIssue
BE-0040DeferredAI assertionsMagicPodIssue
BE-0048ImplementedBehavioral / protocol assertionsMaestroIssue
BE-0049ImplementedDeterminism / flakiness auditMaestroIssue
BE-0050ImplementedE2E coverage mapMaestroIssue
BE-0165ImplementedSelectable perceptual compare engines for visual regressionIssue
BE-0171ImplementedElement-scoped visual assertions and selector-based maskingIssue
BE-0282ImplementedReal-backend network capture, mock, and assertion coverage in CIIssue
BE-0285ImplementedVerify scenario-authoring features on real backendsIssue
BE-0298ImplementedReal concurrent-device verification of parallel device-pool isolationIssue
BE-0306ProposalReal OAuth handshake verification for serve's GitHub loginIssue
BE-0308ImplementedReal-model verification of the system-alert guardIssue
BE-0309ImplementedReal-Postgres CI lane for the serve database layerIssue
BE-0321ImplementedTest impact analysis (affected-step selection from a change)Issue
BE-0341ImplementedCapture per-step report evidence before the step actsIssue
BE-0345ImplementedRecord the concrete coordinate and gesture geometry each step actually actuatedIssue
BE-0346ImplementedAnchor step and network timestamps to the recording's confirmed startIssue
BE-0348ImplementedRecord video, step, and network timestamps as absolute wall-clock timeIssue
BE-0370ImplementedFinish a cancelled run gracefully so it is recorded as a failed runIssue
BE-0371ImplementedDraw a marker at each touch the app receives so recordings show where a gesture landedIssue
BE-0403ImplementedPair a step's screenshot with the element tree that describes the same screenIssue

Authoring experience

97%30/31 implemented
BE-0002ImplementedAI authoring loop & evidence (M2)Issue
BE-0011ImplementedLocal web UI (`bajutsu serve`)Issue
BE-0012ImplementedAction-capture recordIssue
BE-0013ImplementedScenario GUI editorIssue
BE-0014ImplementedDemarcation from the existing AI recordIssue
BE-0044ImplementedScenario provenance (`from:` — step ↔ natural-language origin)BE-0002 follow-up (provenance was deferred there as "remaining light")Issue
BE-0060ImplementedDownload / export a run report as a zipIssue
BE-0068ImplementedRegenerable reports (render from stored run data)Issue
BE-0072ImplementedResponsive serve Web UI (small-screen & touch layout)Issue
BE-0095ImplementedInteractive crawl graph (draggable nodes + realign)Issue
BE-0098ImplementedUnified authoring surface in serveBE-0014Issue
BE-0102ImplementedAggregate run-stats dashboardIssue
BE-0178ImplementedMulti-action record turns (batch intra-screen actions)Issue
BE-0179ImplementedHuman-in-the-loop handoff during record (pause / hand off / resume)Issue
BE-0180ImplementedCrawl history viewer in the Web UIIssue
BE-0182ImplementedHuman value entry during record (OTP / random / one-off values)Issue
BE-0185ImplementedHuman takeover step during record (CAPTCHA / biometrics / unresolvable gestures)Issue
BE-0191ImplementedPluggable theme system for the serve Web UI (visual tokens and swappable transitions)Issue
BE-0192ImplementedVision-on-demand in record (attach a screenshot only when it adds information)Issue
BE-0193ImplementedRight-size the record screenshot with a deterministic client-side downscaleIssue
BE-0194ImplementedLean the record turn payload (compact the element tree, token-budget controls, per-category usage)Issue
BE-0225ImplementedConfig project hub in serve (register, list, switch, run)Issue
BE-0226ImplementedCross-project metrics comparison dashboardIssue
BE-0241ImplementedDrill down from the Stats dashboard to the runs behind itIssue
BE-0262ImplementedLive step-picking and target-scoped runs in the Author editorIssue
BE-0263ImplementedBring the Author view into the tiling layoutIssue
BE-0267ImplementedReuse cost-ordered actuator selection in serve capture and enrichIssue
BE-0273ImplementedView a scenario's contents from the Replay tab (raw YAML + structured steps)Issue
BE-0295ImplementedReal-model verification of the record and crawl propose loopsIssue
BE-0330ProposalShip a Claude Code skill that drafts and self-validates scenarios from sourceIssue
BE-0398ImplementedSplit the comparison view's row click into read-only navigation and an explicit activate actionIssue

Hosting the web UI

85%23/27 implemented
BE-0015ImplementedPublic hosting of the web UIIssue
BE-0016ImplementedSelf-hosting of the web UIIssue
BE-0051ImplementedServe hardening for hosting (auth, input validation)Issue
BE-0055ImplementedOperational logging for the hosted serveIssue
BE-0070DeferredLive in-progress run artifacts across the worker splitIssue
BE-0090ImplementedGovern and sandbox command execution from uploaded bundle configsIssue
BE-0106ImplementedPost-completion worker model (eliminate Redis dependency)Issue
BE-0108ImplementedRestrict config sources to upload and Git when hostedIssue
BE-0110ImplementedEvidence upload to object storage via URIIssue
BE-0127ImplementedSplit the serve operations god-moduleIssue
BE-0129ImplementedBound serve scope and keep host concerns out of shared configIssue
BE-0160ImplementedCredential-free worker uploads via presigned URLsIssue
BE-0166ImplementedCapability-routed job queuesBE-0016Issue
BE-0167ProposalControl-plane scale-out behind a load balancerBE-0016Issue
BE-0168ProposalSelf-hosted high availability and single-point-of-failure hardeningBE-0016Issue
BE-0169ImplementedServe metrics and observability endpointBE-0016Issue
BE-0170ProposalWeighted-fair cross-org job dispatchBE-0016Issue
BE-0173ImplementedSlim Linux web-worker container imageIssue
BE-0190ImplementedOrg-scoped crawl history on the server backendReview of BE-0180Issue
BE-0204ImplementedGCS support for server-side object storageIssue
BE-0239ImplementedDeletable runs and reports in the serve Web UIIssue
BE-0244RejectedDeploy the hosted web UI serviceIssue
BE-0277ImplementedEmbed the commit hash into self-hosted Docker images for the version badgeIssue
BE-0313ImplementedGitHub org membership and Team-based RBAC for serveIssue
BE-0352ImplementedAdmin GitHub Team env var bypasses the org-membership sign-in gateIssue
BE-0375ImplementedDatabase-backed org lifecycle and membership management for serveIssue
BE-0395ImplementedChoose the active org in serve when a login belongs to severalIssue
BE-0404ImplementedCollapse the project layer into the org and the targetIssue

Configuration sourcing

86%12/14 implemented
BE-0063ImplementedLoad config (and its scenario tree) from a Git repository + refIssue
BE-0073ImplementedUpload a config + scenarios + app-binary bundle as a zip and run it from the web UIIssue
BE-0119ImplementedVersion the scenario schema for cross-version readsIssue
BE-0177ImplementedPer-target config defaults for run-behavior settingsIssue
BE-0187ImplementedView the loaded config in the serve Web UI (raw YAML, structured tree, Git provenance)Issue
BE-0224ImplementedGranting private-repository access for the GitHub config sourceIssue
BE-0242ImplementedResolve config-declared paths relative to the config fileIssue
BE-0243ImplementedPersist uploaded zip config bundles to object storage for hosted serveIssue
BE-0268ImplementedUpload config, scenarios, and app binary as independent content-addressed artifacts composed per runIssue
BE-0302ProposalReal-repository fetch verification for the config sourceIssue
BE-0307ProposalReal integration test for the GitHub App config-source token flowIssue
BE-0324ImplementedRead hosted scenarios from the bound config's local cache tree instead of object storageIssue
BE-0325ImplementedReuse the active composition when uploading only the legs that changedIssue
BE-0393ImplementedPer-org config memory, restored into each sessionIssue

Contributor workflow

98%53/54 implemented
BE-0043ImplementedConflict-resistant file flow (generated indexes, modular files, git hygiene)Issue
BE-0061ImplementedCollision-proof BE-ID allocation (atomic reservation + auto-repair)Issue
BE-0065ImplementedDocstring standard & generated API referenceIssue
BE-0067ImplementedCode-quality gate hardening (CI fidelity, security lint, supply-chain)Issue
BE-0069ImplementedExecutable contributor guardrails (procedures as commands)Issue
BE-0074ImplementedStandardize the BE item template (EN / JA)Issue
BE-0078ImplementedStatus-driven roadmap folders (proposals / deferred / in-progress / implemented)Issue
BE-0089ImplementedMerge-time BE-ID allocation on mainIssue
BE-0093ImplementedPublic project website & documentation portal (GitHub Pages)Issue
BE-0094ImplementedGenerated roadmap status dashboard on GitHub PagesIssue
BE-0096ImplementedKeep docs links to roadmap items from rotting on promotionIssue
BE-0100ImplementedProgress tracking and cross-item relations in the BE templateIssue
BE-0103ImplementedRight-size the model and reasoning effort per development taskIssue
BE-0109ImplementedGitHub Issues as the ownership tracker for open roadmap itemsIssue
BE-0112ImplementedEnforce core / contract / periphery layer boundaries in the gateIssue
BE-0113ImplementedRealign DESIGN.md with the current implementationIssue
BE-0117ImplementedCover the rest of the CLI command layer, then ratchet the coverage floorIssue
BE-0122ImplementedLegible GitHub Actions workflow and job namesIssue
BE-0139ImplementedLink the roadmap dashboard and item files to their tracking issueIssue
BE-0149ImplementedClose the roadmap-placeholder format-guardrail gapIssue
BE-0156ImplementedKeep roadmap-item PR labels in sync with TopicIssue
BE-0159ImplementedFlatten roadmap items into one directory (retire status-driven folders)Issue
BE-0162ImplementedRoadmap status-filter skill for AI sessionsIssue
BE-0203ImplementedClaude Code as the automated PR code reviewerIssue
BE-0213ImplementedTerminology glossary and documentation structure reviewIssue
BE-0214ImplementedWeb-only beginner tutorial (no Xcode/Simulator required)Issue
BE-0216Implementedpropose-and-build: author a BE proposal and its implementation together in one PRIssue
BE-0217ImplementedHarden the automated PR review prompt with research-backed policyIssue
BE-0219ImplementedText search on the roadmap status dashboardIssue
BE-0222ImplementedScheduled daily workflows that refresh the roadmap and docs separately, each opening its own review PRIssue
BE-0230ImplementedHands-free implement-review loop: auto-PR and pr-followup polling in implement-beIssue
BE-0266ImplementedContributor workflow tutorial: a hands-on guide to ideation / implement-be / propose-and-buildIssue
BE-0278ImplementedA unified technical-writing skill for BE items and prose docsIssue
BE-0284ImplementedConsolidate duplicated documentation norms under single sources of truthIssue
BE-0286ImplementedLink glossary terms to their definitions on first useIssue
BE-0311ImplementedSortable table view for the roadmap dashboardIssue
BE-0322ImplementedScenario-scoped E2E filtering (fire only the affected on-device jobs)Issue
BE-0328ImplementedBlock secrets from being committed with gitleaksIssue
BE-0335ImplementedRelationship graph view for the roadmap dashboardIssue
BE-0337ImplementedTransit-map layout for the roadmap dashboard's relationship viewIssue
BE-0338ImplementedScenario-key the iOS codegen and visual E2E jobsIssue
BE-0343ImplementedAdd a companion-PR workflow so Claude review's wording-only findings never need a code PR's CI cycleIssue
BE-0347ImplementedBound the CI review cycle with a local Fable-plan, Sonnet/Opus-fix loop requested on demandIssue
BE-0366ImplementedAdd a Rejected roadmap status, distinct from DeferredIssue
BE-0379ImplementedTrack BE work in a live progress Artifact via a dedicated low-cost skillIssue
BE-0380ImplementedA fix-issue skill ships small GitHub issues without a roadmap itemIssue
BE-0383ImplementedRequire a verifiable outcome in every BE item's MotivationIssue
BE-0384ImplementedShip a record-issue skill that files minor bugs and improvements as GitHub IssuesIssue
BE-0385ImplementedTurn the coverage floor into a continuous ratchetIssue
BE-0386In progressEnforce a cyclomatic complexity ceilingIssue
BE-0387ImplementedBroaden ruff's lint rule selectionIssue
BE-0388ImplementedType-check the test suite under mypyIssue
BE-0390ImplementedManage agent skills with APM from a single per-skill sourceIssue
BE-0391ImplementedGuard worktree cleanup against live sessions and undelivered branchesIssue

Miscellaneous / on hold

0%0/1 implemented
BE-0027Deferred`mockServer` (external mock)Issue

Completed 14

Autonomous crawl

100%5/5 implemented
BE-0038ImplementedAutonomous crawl exploration (App Explorer style)Autify VAXIssue
BE-0064ImplementedParallel crawl across multiple simulatorsUser request (crawl efficiency)Issue
BE-0066ImplementedWeb crawl (Playwright backend)Issue
BE-0077ImplementedParallel web crawl across multiple browsersUser request (parallel web crawl)Issue
BE-0181ImplementedResumable crawl continuation (Web UI + full-frontier resume)Issue

codegen coverage

100%10/10 implemented
BE-0003Implementedcodegen, traces, network & CI (M3)Issue
BE-0025ImplementedCoordinate swipe generationIssue
BE-0026ImplementedShrink unsupported syntaxIssue
BE-0062ImplementedPlaywright codegen targetIssue
BE-0083ImplementedUnify the codegen emitters behind a shared scenario walkIssue
BE-0085ImplementedShrink unsupported web (Playwright) codegen syntaxIssue
BE-0209ImplementedAndroid codegen emitter (Espresso / UI Automator)Issue
BE-0293ImplementedReal-compile verification for the Playwright (TypeScript) codegen targetIssue
BE-0294ImplementedReal-compile verification for the UI Automator (Kotlin) codegen targetIssue
BE-0297ImplementedExpand XCUITest codegen's real-compile coverage to the full DSL surfaceIssue

Self-healing triage

100%7/7 implemented
BE-0004ImplementedSelf-healing triage (M4)Issue
BE-0021ImplementedAI triage (root-cause summary, fix suggestions)Issue
BE-0022Implemented`update` (minimal-diff proposals = applying structured fixes)Issue
BE-0023ImplementedGuards against "making tests laxer"Issue
BE-0039ImplementedSelf-healing limited to "propose + opt-in apply"BothIssue
BE-0220ImplementedFlaky-test suggestion and cross-run fix proposals from DB run historyIssue
BE-0296ImplementedReal-model verification of the triage --ai diagnosis pathIssue

doctor / onboarding

100%3/3 implemented
BE-0024Implementeddoctor / onboardingIssue
BE-0164ImplementedConfig-aware environment installerIssue
BE-0304ImplementedReal-environment verification for the onboarding gate (doctor, preflight, provision)Issue

Integration & automation

100%3/3 implemented
BE-0017ImplementedMCP serverIssue
BE-0018ImplementedReturn evidence as MCP resourcesIssue
BE-0301ImplementedReal wire-protocol round-trip test for the MCP serverIssue

Integration with external services

100%2/2 implemented
BE-0099ImplementedWebhook notifications for run resultsIntegration with external servicesIssue
BE-0161ImplementedExport run results in Common Test Report Format (CTRF)Issue

AI provider configuration

100%17/17 implemented
BE-0047ImplementedAI data sovereignty (provider-agnostic, redacted AI path)MaestroIssue
BE-0053ImplementedAmazon Bedrock as a pluggable AI providerIssue
BE-0097ImplementedAI data sovereignty for the crawl guide and serve-spawned AI pathsIssue
BE-0101ImplementedLegible Claude-using / Claude-free split with a zero-config non-AI pathIssue
BE-0104ImplementedVendor-neutral AI backend interfaceIssue
BE-0111ImplementedMake the AI SDK an optional extra so the deterministic gate installs AI-freeIssue
BE-0163ImplementedReplace the Claude Code CLI authoring backend with an `ant`-CLI OAuth AI providerIssue
BE-0175ImplementedSign in to the `ant` provider from the serve Web UIIssue
BE-0176ImplementedRevive Claude Code as an AiBackend adapter with file-based visionIssue
BE-0183ImplementedPer-provider AI settings in the serve Web UIIssue
BE-0184ImplementedPersist serve AI provider settings across restartsIssue
BE-0188ImplementedConfigurable AI output language for record and crawlIssue
BE-0215ImplementedExplicit CLAUDE_CODE_OAUTH_TOKEN credential for the claude-code providerIssue
BE-0229ImplementedResolve serve AI provider settings per organization at runtimeIssue
BE-0300ImplementedReal-API contract smoke lane for the AI backend adaptersIssue
BE-0394ImplementedAdd a none provider that disables every AI path, including the vision alert guardIssue
BE-0402ImplementedDrop the AI-vision fallback from run's reactive system-alert guardIssue

AI usage and cost observability

100%2/2 implemented
BE-0195ImplementedVisualize AI token usage and cost in the serve Web UIIssue
BE-0196ImplementedRecord AI token usage and cost as an attributed, persistent ledgerIssue

Surfacing CLI features in the serve Web UI

100%10/10 implemented
BE-0137ImplementedGenerate native test code from the serve Web UIIssue
BE-0138ImplementedInline scenario validation in the serve editorIssue
BE-0145ImplementedDeterminism audit in the serve Web UIIssue
BE-0146ImplementedE2E coverage map in the serve Web UIIssue
BE-0147ImplementedTriage failed runs in the serve Web UIIssue
BE-0148ImplementedDoctor readiness panel in the serve Web UIIssue
BE-0272ImplementedShow bajutsu's running commit/version in the serve Web UI headerIssue
BE-0275ImplementedA projects management page in serve (a top-level view, not a modal)Issue
BE-0318ImplementedShow the running server's configuration and bundled iOS test-runner status in a serve Settings tabIssue
BE-0340ImplementedUpload a scenario file directly into ReplayIssue

Security hardening

100%19/19 implemented
BE-0115ImplementedAuthenticate the in-process iOS network collectorIssue
BE-0116ImplementedTighten UDID validation against argument injectionIssue
BE-0120ImplementedTokenize secrets in recorded scenario YAMLIssue
BE-0121ImplementedUnconditional CSRF and Host-allowlist defenses for serveIssue
BE-0123ImplementedRoute composite-action inputs through env indirectionIssue
BE-0124ImplementedTighten config-source owner and repo validationIssue
BE-0125ImplementedRestrict the claude-code authoring agent toolsIssue
BE-0130ImplementedRedact sensitive network headers and cookies by defaultIssue
BE-0131ImplementedRestrict run-artifact file permissionsIssue
BE-0133ImplementedPin the actionlint installer by SHAIssue
BE-0136ImplementedWrite-once secrets store for serveIssue
BE-0144ImplementedClose the auto-merge stale-approval raceIssue
BE-0151ImplementedWarn when screenshots and video may capture on-screen secretsIssue
BE-0152ImplementedKeep literal TOTP seeds out of run artifactsIssue
BE-0153ImplementedEncode-aware secret redactionIssue
BE-0154RejectedRun roadmap-promote from the base SHAIssue
BE-0155ImplementedPass idb input text via stdin to keep secrets out of argvIssue
BE-0174ImplementedContain scenario component and data refs within the suite rootIssue
BE-0274ImplementedSet scenario-declared secrets from the serve Web UIIssue
BE-0331ImplementedRedact at the artifact boundary so no writer can emit a plaintext secretIssue

Dogfood fixtures (demo apps)

100%3/3 implemented
BE-0045ImplementedDogfood showcase apps (UIKit × SwiftUI, accessibility-paired)DogfoodingIssue
BE-0079ImplementedConsolidate the demo & dogfood apps onto the showcase suiteDogfoodingIssue
BE-0107ImplementedReach every showcase tab by navigation, not a launch-env shortcutDogfoodingIssue

Dogfood fixtures (web UI)

100%3/3 implemented
BE-0058ImplementedDogfood the serve Web UI (web-backend regression net)DogfoodingIssue
BE-0059ImplementedBring up the target server for a run (`launchServer`)DogfoodingIssue
BE-0189ImplementedGate the serve Web UI dogfood in CIDogfoodingIssue

CI / build infrastructure

100%2/2 implemented
BE-0333ImplementedForce every run-path file to be either gated or explicitly excluded in the E2E relevance filterIssue
BE-0361ImplementedCollect the layered diagnostics an iOS CI failure needsIssue

Codebase quality & technical debt

100%34/34 implemented
BE-0092ImplementedExtract the crawl coordinator into a classIssue
BE-0132ImplementedDeduplicate crawl screenshot helpersIssue
BE-0134ImplementedEliminate serve-to-CLI flag-mirror driftIssue
BE-0135ImplementedResolve top-level module naming debtIssue
BE-0140ImplementedDeduplicate Claude client initializationIssue
BE-0142ImplementedCover the CLI command layerIssue
BE-0143ImplementedDecompose the run command god-functionIssue
BE-0150ImplementedFail cleanly on a malformed scenario in `trace --explain` and `audit`Found while scoping BE-0117 (CLI command-layer coverage)Issue
BE-0172ImplementedDecompose the run-path step loop and per-scenario runnerIssue
BE-0197ImplementedEven out the Environment protocol shape for a third platformIssue
BE-0198ImplementedSplit the JobRegistry out of ServeStateIssue
BE-0199ImplementedShare the doctor screen probe between CLI and serveIssue
BE-0200ImplementedMake the run-id format a single named contractIssue
BE-0201ImplementedConsolidate the duplicated replay helpers of record and enrichIssue
BE-0202ImplementedSplit serve.js into section files without a build stepIssue
BE-0205ImplementedDecompose the crawl CLI command like runIssue
BE-0206ImplementedSplit serve job state from job executionIssue
BE-0246ImplementedClarify the module taxonomy for talking to ClaudeIssue
BE-0247ImplementedMove the serve frontend to ES modulesIssue
BE-0248ImplementedContinue decomposing ServeState into auth and provider-settings managersIssue
BE-0249ImplementedRemove the dead Claude-client wrapper orphaned by the backend seamIssue
BE-0250ImplementedSplit assertions into a package and thread evaluation contexts as one EvalContextIssue
BE-0251ImplementedHoist duplicated driver helpers into drivers.base and unify small constantsIssue
BE-0252ImplementedSplit config into a package and group Effective into sub-recordsIssue
BE-0253ImplementedUnify the serve dual-backend route tables behind a declarative registryIssue
BE-0254ImplementedExtract a shared CoordinateTreeDriver base for idb and adbIssue
BE-0255ImplementedDeduplicate codegen identifier and regex helpers into codegen_commonIssue
BE-0256ImplementedSplit platform_lifecycle into a package and route device resolution through the Environment seamIssue
BE-0257ImplementedPackage the enforced architecture layers as directoriesIssue
BE-0258ImplementedRoute serve step-artifact reads through the ArtifactStore seamIssue
BE-0259ImplementedReuse the settled query snapshot across assert and extract stepsIssue
BE-0260ImplementedConsolidate the duplicated CLI command bring-up and add a neutral DeviceErrorIssue
BE-0261ImplementedRound-trip Author YAML edits through the serializerIssue
BE-0264ImplementedReturn JSON errors for uncaught serve handler exceptionsIssue