Repository navigation
Conversation
Scratch ("No project") is a plain folder, not a Git repository, so a
scheduled task saved with "Create a new worktree" (the editor default, and
what MCP schedule_task stores for unbound tasks) failed every run with
"Workspace preparation failed".
ThreadLaunchService.launch now treats a worktree request on Scratch as a
root launch, up front, so the per-thread Scratch folder, preparation and
retries all see root. ScheduledTaskService.upsert, which both the web
editor and MCP go through, stores such tasks as root. Both ask the new
ManagedProjectFolders.isScratchProject.
github-actions
Bot
added
vouch:unvouched
![macroscopeapp[bot]](https://avatars.githubusercontent.com/in/900172?s=60&v=4){kind=small}
ApprovabilityVerdict: Approved at Macroscope's review found this PR approvable — This is a narrowly scoped server bug fix for scheduled and direct launches targeting the non-Git Scratch project. It preserves Git-project worktrees, adds coverage for legacy and new task paths, and does not alter schemas, product defaults, deployment, or static-analysis configuration. You can add or adjust custom eligibility rules. Learn more. |
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configuration
📒 Files selected for processing (11)
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review. 📝 WalkthroughWalkthroughThe managed-folder service now identifies Scratch projects. Thread launches and scheduled-task upserts convert requested worktree strategies to root strategies for Scratch projects. Tests cover Scratch and Git project behavior and provide the added service dependency. ChangesScratch workspace handling
Priority: ⬇️ Low Estimated code review effort: 3 (Moderate) | ~20 minutes Change: Bug fix · Severity of issue fixed: Low Suggested reviewers:
|
| Check name | Status | Explanation |
|---|---|---|
| Description check |  |
The description covers the problem, implementation, scope, linked issue context, verification results, limitations, and agent details. It provides sufficient evidence for the server-side fix, although… |
| Title check | |
The title is concise, uses the conventional commit format, and clearly identifies the main change: fixing scheduled tasks for No project in a Scratch folder. |
| Linked Issues check | |
The directly linked issue is #17309. The PR addresses the reported Scratch ("No project") failure. ThreadLaunchService.launch converts a Scratch worktree request to root, so the launch uses a pe… |
| Out of Scope Changes check | |
The changed production files support #17309. ManagedProjectFolders.isScratchProject supplies the shared Scratch check, and runtimeLayer supplies the new dependency. The changed tests update servic… |
✨ Finishing Touches
🧪 Generate unit tests (beta)
- Create a new PR
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Comment @coderabbitai help to get the list of available commands.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
This branch has not been deployed
1 participant
Fixes #17309
Problem
@bzbetty reported that a scheduled task for "No project" fails on start with "Workspace preparation failed" when worktree creation is left enabled.
@Alb11747 reproduced it end to end over MCP:
schedule_taskstores unbound tasks with a worktree strategy, and the run failed inremoteExistson the Scratch folder before the provider started.@juliusmarminge traced it on
mainand asked for a server-side fix that coerces worktree strategies on Scratch torootboth on upsert and at launch, so existing tasks and MCP-created tasks stop failing.Scratch ("No project") is a plain folder, not a Git repository. Only
rootlaunches get a per-thread Scratch folder; aworktreelaunch always rangit.createWorktreeagainst it and failed.Change
ThreadLaunchService.launchtreats aworktreerequest on Scratch asroot, once, up front. A root request allocates a per-thread Scratch folder, and the launch persists that folder on the run asexisting_worktree, so preparation and replays of the launch use it. Tasks already stored with a worktree strategy now run.ScheduledTaskService.upsertstores a Scratch task'sworktreestrategy asroot. The web editor and MCPschedule_task/update_scheduled_taskboth save through it.ManagedProjectFolders.isScratchProject(projectId). It reuses the existing Scratch root check thatfolderForThreadalready relied on.Coercion rather than rejection: on Scratch, a root request allocates a per-thread Scratch folder, which is what a worktree was meant to give. Rejecting would only push the same fix onto the user or agent, and no existing validation error covers this case.
Relation to open PRs:
rooteither way, so the two compose.Verification
43f8a8de17)failed, and a probe showedWorkspace preparation failed during provision worktree: Failed to resolve the VCS driver for this Git command.ThreadLaunchService.test.ts,ManagedProjectFolders.test.ts,src/scheduledTasks/*,Scheduler.integration.test.ts,ClaudeAutomaticDelivery.integration.test.tsvp run typecheck(apps/server)error TS/warning TSThreadLaunchService.test.ts, on a line this PR does not touch)worktreeon Scratch is released (starting) in its own Scratch folder, stored asexisting_worktree, nocreateWorktreeworktreeon a Git project completes worktree creation and is released with itsworktreestrategyworktreeis stored asrootworktreeworktreeworktreeruns (runNow), is released in its own Scratch folder, stored asexisting_worktreefailedin workspace preparationisScratchProjectis true for Scratch, false for other projects and when Scratch is unavailableThis is a server-only change with no UI, so there are no screenshots.
The tests mock Git to fail on the Scratch root with the same
GitCommandErrorthe real path raises ("Failed to resolve the VCS driver for this Git command."), and to succeed on the Git project. They wait on the run's release or failure event, with no polling.Limitations / not checked:
retryPreparation) still reuses its storedworktreestrategy and fails again. Running the task again creates a corrected launch.Implemented with Claude Opus 5.5, verified with GPT-6 Astra, coordinated by Claude Fable 5.1 in Claude Code.