feat(mobile): allow several new-task drafts per project

[juliusmarminge]

Stacked on #10260 (which is stacked on #10245).

Problem

Mobile keyed new-task drafts by project (new-task:<environmentId>:<projectId>), one slot per project. Starting a new task in a project always landed in whatever draft already sat there, so there was no way to hold two ideas for the same repo. Web mints a draft id per draft and never had this limit. With #10260 listing drafts in the thread list, the single slot became visible.

Fix

Drafts are keyed new-task:<draftId> and carry their project in a project stamp (environmentId, projectId, createdAt). Everything that used to read the project out of the key now reads the stamp:

• Migration. Legacy project-keyed entries are rewritten to id keys with the stamp filled in when the draft file loads, for live drafts and for drafts archived by sign-out, so nothing on users' devices is lost.
• New Task mints a fresh draft. Each entry into the composer binds to its own draft, whether it came from the compose button, a project row, or the provider's project fallback. Tapping a Draft row resumes that draft by id through a new draftId route param; the screen waits for draft hydration and for the project list before deciding a draft is gone.
• Switching projects or environments retargets the same draft (text and model/mode choices follow; branch and worktree are cleared since they belong to the old repo) instead of copy-if-empty across keys. copyComposerDraftContent* is removed.
• Uploads and sign-out resolve a draft's environment from the stamp (composerDraftEnvironmentId now takes the draft), so id-keyed drafts upload attachments and survive relay sign-out.
• A rejected pending task restores into its own message-keyed draft stamped with the project, rather than merging into whatever the user is typing.
• Drafts list newest first by their own createdAt instead of a render-time stamp.

Verification

apps/mobile: vp test run across the draft store, list model, outbox drain/removal/editor, attachment upload queue, and thread-list suites — 239 passed. vpr typecheck passes. Lint on touched files shows only pre-existing warnings.

New tests: legacy key migration (live and archived), minting several drafts per project, retarget keeps text and clears workspace selection, environment lookup from the stamp, rejected-task restore into a stamped draft.

Before / after — starting a fresh New Task in a project that already has a draft

After — three drafts for the same project in the thread list

One migrated from a legacy project key, one already id-keyed, and one typed live and retargeted across environments. Tapping any row opens that draft.

Captured on an iPhone 17 Pro simulator with the dev client against a disposable backend; the "before" is the head of #10260 with the same seeded state. Captures were taken before the stack was rebased onto current main (which retinted the Pending pill and reworked carryDraftContentTo); the rebase changed no draft behaviour shown here.

Surfaces

• Entry points: compose button, project row "new thread", Draft rows (v1 and v2 lists, Home and iPad sidebar), pending-task restore.
• Clients: mobile only. Web already has per-draft ids; server untouched.
• Reverse state: Discard from the row menu, or clear the composer; both drop the stamp so the row disappears.

Model: Claude Fable 5 · Harness: Claude Code in T3 Code

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	7.0 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.5 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	57.0 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	8	—	21	✅
Claude	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	7.0 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.5 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	57.8 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	8	—	21	✅

Baseline: unavailable · PR result: c1157a3 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.9 KiB
• Claude decoded thread snapshot: 114.6 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR changes mobile draft identity and persistence from one project slot to multiple ID-keyed drafts, adding migration, route resumption, project retargeting, attachment scoping, and recovery behavior across several production paths. The cross-cutting state and persisted-data changes have a broad runtime surface and warrant human review.

Not approved because:

• Monthly spending limit reached (workspace setting). Approvability relies on correctness review in order to determine eligibility

^{Review your spending limits in Billing settings. You can add or adjust custom eligibility rules. Learn more.}

[cursor]

Bugbot is paused — on-demand spend limit reached

Bugbot uses usage-based billing for this team and has hit its on-demand spend limit.

A team admin can raise the spend limit in the Cursor dashboard, or wait for the next billing cycle to continue.

[RealJFX]

Super cool