Skip to content

ci(nightly-judge): stop republishing stale verdicts on failed runs - #254

Merged
HomenShum merged 2 commits into
mainfrom
fix/nightly-judge-no-stale-artifacts
Sep 11, 2026
Merged

ci(nightly-judge): stop republishing stale verdicts on failed runs#254
HomenShum merged 2 commits into
mainfrom
fix/nightly-judge-no-stale-artifacts

Conversation

@HomenShum

Copy link
Copy Markdown
Owner

A maintainer downloading the nightly-judge artifact expects verdicts the judge just produced. The judge step needs GOOGLE_GENERATIVE_AI_API_KEY, which is absent from this repo's Actions secrets, so it fails honestly on every scheduled run (37 of 37 since 2026-08-05). The upload step then ran under if: always() and shipped the git-tracked docs/eval/gemini-media-judges/ folder (last changed 2026-07-04) as a fresh per-run artifact. This change uploads only after a successful judge run.

Not changed: the judge still fails red without the key. Whether to provision the key (Gemini spend on every nightly run) or leave the lane NOT_RUN is an owner decision recorded in the portfolio ledger.

🤖 Generated with Claude Code

The judge step needs GOOGLE_GENERATIVE_AI_API_KEY and fails honestly without it,
but the following upload step ran with if: always() and shipped the git-tracked
docs/eval/gemini-media-judges/ folder as a per-run artifact. Every one of the 37
scheduled failures since 2026-08-05 therefore published the committed verdicts
from 2026-07-04 as if they had just been produced. Upload only after a successful
judge run. The missing-key failure itself stays red and unchanged.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@vercel

vercel Bot commented Sep 11, 2026

Copy link
Copy Markdown

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

Project Deployment Actions Updated
noderoom Ready Ready Preview Sep 11, 2026 9:07am UTC

Request Review

@github-actions

github-actions Bot commented Sep 11, 2026

Copy link
Copy Markdown

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

⚠️ Immutable proof files changed: .github/workflows/nightly-judge.yml

This is advisory because the check is not running in strict scaffold-repair mode.
If this PR is applying accepted scaffold proposals, rerun with --strict-immutability
and reject the PR unless the immutable changes are removed.

Immutable files guarded during scaffold repair:

  • 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

Changed Files

  • .github/workflows/nightly-judge.yml
  • 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/20260911T090637Z.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/20260911T090646021Z-8be6f21b_dirty.d59b505e339e2a3c/cascade-healthy.json
  • docs/eval/traces/credit/20260911T090646021Z-8be6f21b_dirty.d59b505e339e2a3c/delta-incomplete.json
  • docs/eval/traces/credit/20260911T090646021Z-8be6f21b_dirty.d59b505e339e2a3c/mapping-correct.json
  • docs/eval/traces/credit/20260911T090646021Z-8be6f21b_dirty.d59b505e339e2a3c/mapping-misbind.json
  • docs/eval/traces/credit/20260911T090646021Z-8be6f21b_dirty.d59b505e339e2a3c/summit-stressed.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L1_read_scripted.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L2_edit_scripted.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L3_conflict_scripted.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L4_blocked_scripted.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L5_large_range_scripted.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L6_long_horizon_scripted.json
  • docs/eval/traces/ladder/20260911T090645545Z-8be6f21b_dirty.324592b8e1292fd5/ladder_L7_resume_scripted.json

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

@HomenShum
HomenShum merged commit 6f1af0f into main Sep 11, 2026
11 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant