feat(belief-state): consume runtime decision records

[drewstone]

Summary

• read runtimeDecisionPoints embedded in benchmark corpus records when explicit decisions are not supplied
• validate malformed decision arrays/rows with diagnostics instead of fabricating points
• keep lifecycle-only benchmark rows blocked for belief claims
• update the runtime benchmark corpus test to prove record-embedded decisions feed Phase 0 measurement

Why

Agent Runtime benchmark rows now persist semantic decision points beside lifecycle runtimeEvents. Agent-eval should consume that artifact directly so the evidence path is corpus row -> runtime trajectory -> belief Phase 0 packet, with labels still explicit.

Verification

• pnpm exec vitest run tests/belief-state/runtime-benchmark-corpus.test.ts tests/runtime-trajectory.test.ts tests/belief-state/phase0-measurement.test.ts
• pnpm typecheck
• pnpm lint (passes; existing warnings in src/authenticity/index.ts and src/storyboard/code-edit.ts)

[tangletools]

⚠️ Review Interrupted — b2d91afa

The review runner stopped before publishing a final verdict: webhook_restarted.

State	Detail
Interrupted	webhook restarted

No review verdict was produced for this run. Trigger a fresh review on the current PR head if the PR is still open.

_{tangletools · #232 · model: kimi-for-coding · updated 2026-06-07T21:16:32Z}