fix(prime): preserve transcript continuity after harness refinement

[rynfar]

Prime Agent can append a refinement_outcome custom message after a background memory update. Pylon rejected its live message events and dropped it from snapshots while retaining the native message count, causing a later tool snapshot to fail with “transcript continuity could not be verified after synchronizing.” This matches the reported September 14 failure: background refinement at 14:52:53 UTC, followed by snapshot rejection at 14:53:52 UTC.

Decode refinement outcomes into a private transcript identity containing only a timestamp and SHA-256 digest. Live events and snapshots now retain the same message, while the existing refinement event continues to own client projection. The existing harness-digest restrictions and rejection of unrelated custom messages remain intact. Snapshot reconciliation still rejects changed history.

Validation: reproduced the decoder failure before the fix; 524 tests passed across Prime daemon event decoding, adapter, runtime-event mapping, and session runtime. Server package typecheck, targeted lint, formatting, and diff checks passed. Regression coverage verifies background refinement followed by tool-result snapshot recovery and rejection of altered refinement history.

This is confined to the Prime adapter boundary: no wire contract, persistence schema, or other provider behavior changes. Web, desktop, and mobile share the corrected server path in local and remote connections. Live userdata was inspected read-only; this PR does not repair previously persisted recovery proofs or deploy a running-app update.

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 3:02pm UTC

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	14.0 KiB	13.9 KiB	−114 B (−0.8%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.2 KiB	7.2 KiB	+5 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.7 KiB	−119 B (−1.7%)	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	−166 B (−1.2%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.2 KiB	7.2 KiB	+5 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.9 KiB	6.7 KiB	−171 B (−2.4%)	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: 0bef062 · PR result: 96421cc · 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: 116.4 KiB

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