[matsvarn]

Worktree setup previously modified credential links and let agents start while dependency installation was still running. Add a repeatable bash scripts/setup.sh that uses the frozen native pnpm lockfile through Vite+, warms the checkout's web cache, and leaves credentials and databases alone. The automatic Unix action waits for completion; preserve the existing Windows action as manual.

The fork's checks also previously waited indefinitely for unavailable Blacksmith runners. Select GitHub-hosted Ubuntu and macOS for this fork, preserve upstream Blacksmith labels, and guard the Blacksmith mirror setup. Check jobs, assertions and timeouts stay intact.

Add a manual scripts typecheck and dev-runner test action. Existing dev runner state and port isolation remain unchanged, and development guidance explains per-environment action import and commit-based transfer between machines.

Validation: Mac setup passed three times, including final setup behavior, from outside the checkout without source or lockfile changes or credential links. The exact Check action passed with 74 dev-runner tests; seven project-file tests passed separately. Scripts typecheck passed with existing advisory diagnostics, and the final three-action file passed the native schema. Linux setup passed twice after the separate host build-tools install, leaving source clean and credentials untouched. Scripts typecheck and 81 focused tests passed on Linux. Targeted formatting, shell syntax, workflow YAML and APT-helper skip checks passed. Full hosted CI run 37326522696 passed all nine jobs on the final head, including the desktop build and full test lanes; the advisory native fingerprint check passed. T3 action import and automatic hook execution remain unverified. No installed T3 application or live user state was changed. Vite+ was installed separately on codingbox without shell-profile changes; global pnpm10 remains in place.

Model and harness: GPT-6 through Codex desktop.

[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	—	13.5 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	7.1 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.5 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	56.3 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	10	—	21	✅
Claude	Total thread wire	—	13.5 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	7.1 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.5 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	57.0 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	9	—	21	✅

Baseline: unavailable · PR result: e0573b2 · 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.