[unlockdigital]

This brings the approved checkpoint-A upstream fixes into the Unlock Digital fork while preserving its dependency patches, account routing, privacy controls, and migration bridge. It updates bundled browsers, provider modes and Codex recovery, diagnostic redaction, and delegation/Claude Stop behavior. Forward-read safeguards preserve newer credential-reference settings and unsupported schedules during rollback.

The Claude SDK remains pinned with a reviewed backport for its embedded Zod v4 leading-bound array checks. The dependency guard validates resolved security floors and the installed SDK bundle hash. CI uses native pnpm for the advisory threshold. The fork web job has a twenty-minute budget for dependency setup, the full suite and production boot; individual test deadlines stay at their existing values.

PDF reload coverage runs as three focused scenarios. It preserves the page, zoom, scroll offsets, fit-width mode and reset on another file path within the normal test budget.

Validation: checkpoint-wide changed-file checks pass, including 1,519 tests across 28 files and typechecks for 39 affected packages. The final test-only correction check passes all 135 tests in the affected file, format, lint, knip and affected typechecks. Frozen installation and macOS arm64 packaging pass. Packaged server files match the build; the package retains the Unlock Digital preview identity and protocol. The two credential-collision findings were independently reviewed and corrected. Full PR CI passed on bc4d5ae, including all 5,295 web tests and production boot: https://github.com/unlockdigital/t3code/actions/runs/41916926254. The final thread-transfer report remains within all enforced ceilings.

Qualification: six workspace advisories remain documented separately from the dismissed reference-only alerts. Legacy SDK Zod v3 reachability remains unresolved. The production restart prompt is unchanged; only offline eval preparation is included. Native visual verification was stopped at the user's command-line-only direction. Current live application data and the original dirty checkout are preserved. Checkpoint B and managed client policies remain outside this change.

Implemented with Codex (GPT-6.1 Sol), with scoped Luna, Terra, and Sol implementation and independent reviews.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.6 KiB	21.6 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.1 KiB	+6 B (+0.1%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.3 KiB	1.3 KiB	+6 B (+0.5%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 0104640 · PR result: bc4d5ae · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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