fix(mobile): resolve Antigravity account catalogs before sending

[rynfar]

Mobile could not safely apply Antigravity catalog membership checks: an unresolved antigravity-default choice would block first use, and the empty catalog after restart would block saved models. Resolve defaults and aliases only against the selected account, preserve its model options and unavailable concrete choices, and keep unchecked concrete restart sends usable.

A queued user request can discover that account's catalog once, with a 30-second bound. Delivery then re-reads the durable message, connection, binding, source epoch and catalog, and checks the catalog again after attachment preparation. Held requests retain their text, context, attachments and account; completed requests release discovery bookkeeping. Existing Pylon provider admission, continuation and Prime controls still apply. A fresh draft can discover an available Antigravity account without opening Settings, while unresolved marker rows stay out of the model picker.

Reopens the mobile omission in #479; refs #526. Adopted behavior comes from 06336460c9988f29c71e839c4c9c840c4552e077 and restart correction d487dfbf46be344e818725be70ee04be2436bfb4. Related 3faeee49ac67dfd9534369f1e1c627c0356b75ac remains the already-adopted web wizard change; it has no additional mobile hunks.

Validation: 140 tests across nine affected suites, followed by 30 tests across the two affected suites after correcting the test atom facade; mobile typecheck and scoped lint pass (React compiler warnings remain on existing screens). Tests cover account isolation, aliases/defaults, missing/legacy selections, restart and provider gates, catalog changes during attachment awaits, bounded discovery and bookkeeping cleanup, and existing outbox/context recovery. Cleanly rebased onto 8c9930c574; all 18 reviewed mobile task files were byte-identical through the rebase. Independent adversarial review approved the unchanged mobile diff.

No simulator, physical device or live Antigravity account was exercised, and no before/after images were captured: this task has no browser/device authorization. No native module, release or installation changes.

Implemented with GPT-6 in the Codex multi-agent harness.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
pylon-marketing	Ready	Preview	Sep 12, 2026 4:39pm UTC

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.9 KiB	14.1 KiB	+161 B (+1.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.2 KiB	7.2 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.7 KiB	6.8 KiB	+161 B (+2.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.0 KiB	58.8 KiB	+910 B (+1.5%)	66.4 KiB	✅
Codex	Live turn messages	8	10	+2 (+25.0%)	21	✅
Claude	Total thread wire	13.9 KiB	14.1 KiB	+140 B (+1.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.2 KiB	7.2 KiB	−4 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.7 KiB	6.8 KiB	+144 B (+2.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.8 KiB	59.7 KiB	+884 B (+1.5%)	66.4 KiB	✅
Claude	Live turn messages	8	9	+1 (+12.5%)	21	✅

Baseline: d6d2f8e · PR result: 26a20ef · 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: 115.6 KiB
• Claude decoded thread snapshot: 116.3 KiB

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