refactor(server): resolve self invocations through one shared helper

[juliusmarminge]

#12747 and #12791 each fixed one place where the server spawned itself with process.execPath plus argv[1], which breaks under the single-executable, and #12747 then reached for node:sea directly inside the ACP adapter. The rest of the server already answers "are we the packaged binary" through the HostProcessIsExecutable reference and resolveNodeExecutable, so this leaves three ways of asking the same question.

This adds resolveSelfInvocation next to resolveNodeExecutable in @t3tools/shared/nodeRuntime: one Effect that returns { command, entrypoint } from HostProcessExecutablePath, HostProcessIsExecutable and HostProcessArguments, plus selfInvocationArgs to build the argv. The ACP adapter takes the resolved value as an option instead of reading the process, and the three adapter factories (Grok, Antigravity, ACP registry) resolve it in their Effect layer. AcpAdapterV2.ts no longer touches process.execPath, process.argv, or node:sea.

The /bin/sh cgroup wrapper from #12791 is unaffected: it no longer spawns Node at all.

Verified: packages/shared nodeRuntime.test.ts (2 new cases, 11 total), apps/server AcpAdapterV2.test.ts / AcpRegistryAdapterV2.test.ts / GrokAdapterV2.test.ts / AntigravityAdapterV2.test.ts (125), vpr typecheck in both packages, lint on touched files. The existing self-contained bridge test now provides HostProcessIsExecutable instead of mocking node:sea.

Claude Fable 5 via Claude Code.

🤖 Generated with Claude Code

---

[macroscopeapp]

Import this local Effect service module as a namespace rather than adding another service tag as a named import. Please use import * as HostProcess from "./hostProcess.ts" and reference its tags through that public namespace (for example, yield* HostProcess.HostProcessArguments), updating the existing host-process references in this file accordingly.

^{Posted via Macroscope — Effect Service Conventions}

[juliusmarminge]

This file already imported HostProcessEnvironment, HostProcessExecutablePath, HostProcessIsExecutable and HostProcessPlatform by name before this PR (same on main). Adding HostProcessArguments to that list matches the file; switching the whole file to a namespace import is unrelated churn for this change.

[macroscopeapp]

Effect Service Conventions found one issue in packages/shared/src/nodeRuntime.ts: the newly consumed local Effect service tag should be accessed through a namespace import. The inline review comment includes the expected fix; it requires updating multiple references, so no self-contained diff is feasible.

^{Posted via Macroscope — Effect Service Conventions}

[macroscopeapp]

Approvability

Verdict: Approved at d010b48

Macroscope's review found this PR approvable — This PR centralizes existing ACP self-invocation logic and corrects packaged-binary handling without changing schemas, defaults, or production infrastructure. The runtime effect is narrowly limited to launching the existing MCP bridge and is covered by focused tests.

^{You can add or adjust custom eligibility rules. Learn more.}

[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	—	4.9 KiB	—	6.8 KiB	✅
Codex	Thread snapshot wire	—	3.7 KiB	—	4.9 KiB	✅
Codex	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Codex	Live turn WebSocket decoded	—	20.4 KiB	—	29.3 KiB	✅
Codex	Live turn messages	—	2	—	8	✅
Claude	Total thread wire	—	4.9 KiB	—	6.8 KiB	✅
Claude	Thread snapshot wire	—	3.7 KiB	—	4.9 KiB	✅
Claude	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Claude	Live turn WebSocket decoded	—	20.8 KiB	—	29.3 KiB	✅
Claude	Live turn messages	—	2	—	8	✅

Baseline: unavailable · PR result: d010b48 · 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: 106.1 KiB
• Claude decoded thread snapshot: 106.4 KiB

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