feat(server): auto-nest and settle provider-native subagents

[jimprince]

Problem

Claude Task-tool subagents were left loose and unsettled in the sidebar: 24 were reported today, and new ones keep arriving. They have no app run, so the existing settle-on-complete policy could not see their completion.

Fix

Native children inherit their organization parent from the thread that spawned them. The existing lifecycle settles explicit successful native completion, reopens automatically settled work when the parent resumes the provider task, and honors per-thread opt-outs. Failed or cancelled tasks, pending decisions, background work and live children keep their guards; manual choices stay authoritative. create_threads and t3_thread_launch remain top-level, and canaries are untouched. Native child conversations remain read-only; messaging goes through the parent, and reopening follows a new provider task_started generation.

Approved by Brad on 2026-10-11. This PR is based on product/stack-20261012 (PR #87), so its diff contains only the separately droppable fork-sidebar-hygiene-native concern. All 156 lower patch objects remain unchanged.

Checks

Exact-tip candidate CI: https://github.com/jimprince/t3code/actions/runs/38162199592 (source 2e16d59).

Eight affected package typechecks and the final server repeat passed. Focused native/seed/purpose suites passed (30 tests), as did native boundary and launch/MCP suites (61 tests). One optional copied-state fixture was skipped locally. vp check, isolated 157-patch stack validation, documentation and 244 release-note entries passed.

Five native mutations were seen red: creation seeding, successful completion, generation comparison, resume reopening and the background-work guard. The real Claude adapter → ingestor → EventSink → lifecycle path passes after restoration.

Not included

Historical backfill, automatic settlement of failed/cancelled work, default nesting for ordinary launch tools, a new scheduler or monitor, or changes to the Harness and cleanup patches below this concern.

Built by Claude Sonnet 5.5, Claude Opus 5.5 and GPT-6.1 Sol agents via T3 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	—	5.0 KiB	—	6.8 KiB	✅
Codex	Thread snapshot wire	—	3.8 KiB	—	4.9 KiB	✅
Codex	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Codex	Live turn WebSocket decoded	—	20.9 KiB	—	29.3 KiB	✅
Codex	Live turn messages	—	2	—	8	✅
Claude	Total thread wire	—	5.0 KiB	—	6.8 KiB	✅
Claude	Thread snapshot wire	—	3.8 KiB	—	4.9 KiB	✅
Claude	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Claude	Live turn WebSocket decoded	—	21.2 KiB	—	29.3 KiB	✅
Claude	Live turn messages	—	2	—	8	✅

Baseline: unavailable · PR result: c61e372 · Source CI: failure

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.