feat(progress): show honest waits and delegated work

[rynfar]

Problem

Plan progress could not distinguish active work from a real dependency, so an agent waiting for the user looked like it was still working. Native delegated work was also visible only as separate agent activity, without a quiet summary beside the root plan.

Solution

• add a strict waiting plan state with waitingOn: user | delegates | external for managed Prime plans;
• render owner-specific wait language, including ! and Needs your input only when the user must act;
• derive one compact delegated-work aggregate from existing task events, scoped to the current plan turn and without extra model calls;
• recover the latest current-turn plan after restart, even when it falls outside the normal activity detail window;
• preserve same-turn dismissals until the waiting step or owner meaningfully changes.

The managed Prime tool enforces at most one active inProgress or waiting step. Existing provider-native plans keep their current compatibility. Web and desktop render this shared UI; mobile has no plan surface yet and is unchanged.
The rebased UI uses Pylon's shared Dot Matrix status language: warning for user-owned waits and waiting for passive dependencies.

Evidence

Before and after images will be attached to this PR conversation.

Verification

• 216 focused contract, server, recovery, session-logic, and component tests
• @t3tools/contracts typecheck
• t3 server typecheck
• @t3tools/web typecheck
• targeted lint and formatting on all 20 changed files
• git diff --check
• isolated seeded web/desktop preview reviewed by the maintainer

Closes #126

Built with GPT-5.6 Sol using Pylon's Prime Agent harness.

---

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

[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	13.2 KiB	13.3 KiB	+106 B (+0.8%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+12 B (+0.2%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.3 KiB	6.4 KiB	+94 B (+1.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	54.7 KiB	55.5 KiB	+822 B (+1.5%)	66.4 KiB	✅
Codex	Live turn messages	9	9	0 (0.0%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	−3 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−13 B (−0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+10 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	56.4 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: 92ec009 · PR result: daebe77 · 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: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

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

[rynfar]

Visual evidence

Combined before/after comparison from the isolated seeded web/desktop preview: