Preserve deterministic NodeAgent workbook recovery

[HomenShum]

Summary

• preserve deterministic workbook plans across runtime compaction
• stop stale follow-on plans after verified deterministic writes instead of delegating into repeated provider calls
• accept only validated completed Excel cache finalization as recalculation proof
• add multi-sheet and external-recalculation regressions

Proof

• clean
  pm run floor: 350 files, 2,489 tests passed
• SpreadsheetBench V2 Debugging/01_01: 5 steps, 0 model tokens, 22 verified formula repairs, overall 0.576332 baseline -> 0.580125
• free-provider quota is no longer required for this deterministic task

Honest limits

• task remains
  eeds_repair without fully completed Excel recalculation; stable-pending is not promoted to passed
• full 321-task model-edit-plan receipt coverage remains incomplete

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
noderoom	Ready	Preview, Comment	Jul 16, 2026 9:48pm

[github-actions]

Scaffold Handoff — For Your Coding Agent

Your coding agent (Codex, Claude Code, etc.) should apply the accepted
scaffold proposals below. Do NOT touch any immutable files.

Immutability Check

Mode: advisory

✅ No immutable files were modified in this branch.

Changed Files

• docs/eval/OFFICIAL_BENCHMARK_READINESS.md
• docs/eval/OFFICIAL_BENCHMARK_TASK_COVERAGE.md
• docs/eval/OPENROUTER_CONVEX_BENCHMARK.md
• docs/eval/agent-improvement-loop.md
• docs/eval/agent-improvement-loop.svg
• docs/eval/agent-improvement-loop/20260716T214714Z.json
• docs/eval/agent-improvement-loop/latest.json
• docs/eval/agent-workspace-sandbox-smoke.json
• docs/eval/algorithm-artifact-smoke.json
• docs/eval/bankertoolbench-official-contract.json
• docs/eval/docker-sandbox-probe.json
• docs/eval/eval-runs.jsonl
• docs/eval/halo-convex-context-telemetry.json
• docs/eval/halo-self-improvement-smoke.json
• docs/eval/halo-variant-selection.json
• docs/eval/official-benchmark-readiness.json
• docs/eval/official-benchmark-task-coverage.json
• docs/eval/openrouter-convex-benchmark.json
• docs/eval/professional-catalog-proofs.json
• docs/eval/professional-proof-ledger.json
• docs/eval/spreadsheetbench-chart-visual-probe.json
• docs/eval/traces/credit/20260716T214722695Z-eea1f191_dirty.c04d38c98d5641d1/cascade-healthy.json
• docs/eval/traces/credit/20260716T214722695Z-eea1f191_dirty.c04d38c98d5641d1/delta-incomplete.json
• docs/eval/traces/credit/20260716T214722695Z-eea1f191_dirty.c04d38c98d5641d1/mapping-correct.json
• docs/eval/traces/credit/20260716T214722695Z-eea1f191_dirty.c04d38c98d5641d1/mapping-misbind.json
• docs/eval/traces/credit/20260716T214722695Z-eea1f191_dirty.c04d38c98d5641d1/summit-stressed.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L1_read_scripted.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L2_edit_scripted.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L3_conflict_scripted.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L4_blocked_scripted.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L5_large_range_scripted.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L6_long_horizon_scripted.json
• docs/eval/traces/ladder/20260716T214722289Z-eea1f191_dirty.e309e2a67caf3b2e/ladder_L7_resume_scripted.json
• src/eval/spreadsheetBenchNodeAgentBridge.ts
• tests/spreadsheetBenchNodeAgentBridge.test.ts

Needs Adversarial Review — Do NOT Apply Yet

These proposals passed the reject check but have not been approved by
an adversarial reviewer. A human or frozen LLM judge must approve them first.

• scaf-001 (AGENTS.md): Add explicit instruction for step spreadsheetbench-runner-fixture: Step spreadsheetbench-runner-fixture failed — scaffold may need explicit instruction or evidence assertion.
• scaf-002 (AGENTS.md): Add explicit instruction for step convex-boundaries: Step convex-boundaries failed — scaffold may need explicit instruction or evidence assertion.

Safety Boundary

Agent may improve the scaffold.
Agent may NOT weaken the proof gate.

Immutable files (never modify):

• scripts/proofloop.mjs
• scripts/agent-improvement-loop.ts
• tests/harnessChangeEval.test.ts
• .github/workflows/
• src/eval/evalTrustPolicy.ts
• src/eval/architectureBudget.ts
• evals/evalStore.ts

Scaffold files (safe to modify):

• AGENTS.md
• CLAUDE.md
• proofloop/scenarios/*.yaml
• proofloop/rubrics/*.yaml
• proofloop/subagents/*.md
• proofloop/adapters/*.js
• .proofloop/memory.jsonl
• src/nodeagent/models/prompts/systemPrompt.ts