fix(prime): preserve applied refinement notices across synchronization

[rynfar]

Applied Prime refinements emit both an outcome and a model-facing notice. Pylon recognized only the outcome, so a valid notice could trigger a protocol violation or make the next snapshot fail transcript continuity.

Recognize refinement_notice with its own private transcript identity. Preserve existing outcome and harness-digest fingerprints; both refinement message types remain absent from client projections. Reconciliation continues to reject changed, reordered, missing or incorrectly attributed history.

Fixes #550.

Validation: reproduced the decoder failure before implementation; 531 focused tests passed across event decoding, adapter synchronization, restart replay, runtime-event projection and session runtime. After a type-narrowing correction, the 22 affected decoder/restart tests passed again. Server package typecheck, targeted lint, formatting and diff checks pass. Tests cover outcome-plus-notice ordering, valid session/prompt attribution, invalid provenance, live tool snapshot recovery, restart suffix replay, changed/reordered history and private-content exclusion.

Audit: remaining built-in native message/summary transitions are recorded in #552. They remain unsupported by transcript reconciliation pending scoped proofs; this change does not accept arbitrary custom messages. This is server-side Prime normalization shared by web, desktop and mobile in local/remote modes; no contracts, other providers, persistence schema, release or live userdata changes. It does not repair old recovery proofs persisted by a version that dropped these messages.

Model: GPT-6. Harness: Codex.

---

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

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
pylon-marketing	Ready	Preview	Sep 14, 2026 4:50pm UTC

[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	14.0 KiB	13.9 KiB	−146 B (−1.0%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.2 KiB	7.2 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.7 KiB	−142 B (−2.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.8 KiB	58.0 KiB	−866 B (−1.4%)	66.4 KiB	✅
Codex	Live turn messages	9	8	−1 (−11.1%)	21	✅
Claude	Total thread wire	14.1 KiB	13.9 KiB	−153 B (−1.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.2 KiB	7.2 KiB	+3 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.9 KiB	6.7 KiB	−156 B (−2.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	59.7 KiB	58.8 KiB	−928 B (−1.5%)	66.4 KiB	✅
Claude	Live turn messages	10	8	−2 (−20.0%)	21	✅

Baseline: d773507 · PR result: 43473f4 · 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: 115.6 KiB
• Claude decoded thread snapshot: 116.3 KiB

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