fix(server): give Prime Agent scoped browser tools

[rynfar]

Problem

Prime Agent sessions honored the browser-access permission in Pylon, but neither daemon mode nor ACP compatibility mode received Pylon’s thread-scoped t3-code preview server. The permission therefore appeared enabled while native preview_* tools were unavailable.

Fix

• pass the existing McpProviderSession endpoint and bearer header into Prime ACP session/new requests
• attach the same server to daemon sessions through Prime Agent 0.8.0’s public owner-scoped MCP API before the first snapshot
• restore MCP ownership before publishing a reconnect resync, and fail closed if recovery is not authoritative
• release only Pylon’s owner and t3-code server during teardown
• revoke scoped credentials on explicit stop, failed teardown, stopAll, and authoritative autonomous session exit
• document the behavior and disabled-access path

Closes #77.

Verification

• vp test run apps/server/src/provider/Layers/PrimeAgentAdapter.test.ts apps/server/src/provider/prime/PrimeAgentDaemonAdapter.test.ts apps/server/src/provider/prime/PrimeAgentDaemonSessionRuntime.test.ts apps/server/src/provider/Layers/ProviderService.test.ts apps/server/src/mcp/McpSessionRegistry.test.ts — 220 passed
• vp test run apps/server/src/provider/Layers/CursorAdapter.test.ts apps/server/src/provider/Layers/GrokAdapter.test.ts apps/server/src/provider/Layers/ProviderService.test.ts — 81 passed
• vp run -F t3 typecheck — passed
• targeted lint on all nine changed TypeScript files — 0 warnings, 0 errors
• git diff --check origin/pylon...HEAD — passed
• real Pylon desktop client with an isolated runtime home and Prime Agent 0.8.0 — native preview_open, preview_type, preview_click, preview_snapshot, preview_recording_start, and preview_recording_stop completed against a local fixture; the final snapshot showed Pylon Browser Verification, pylon-browser-pass, and count 1; the saved MP4 was 37,395 bytes

Implemented by GPT-5.6 Sol using Prime Agent.

---

^{Need help on this PR? Tag @codesmith 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	13.4 KiB	13.4 KiB	+10 B (+0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+4 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.6 KiB	+6 B (+0.1%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.0 KiB	55.0 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.4 KiB	−6 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+2 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	−8 B (−0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	55.8 KiB	55.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 8320f07 · PR result: cd53159 · 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: 110.0 KiB

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