# Sample Reproducible Proof Recipe

The dashboard demo uses wholly synthetic, illustrative values and is not
customer or benchmark evidence. The checks below use redacted public artifacts
and the legacy quarantine registry; they do not promote raw JSON, JSONL, or
screenshots into current verified proof.

Run commands from the repository root.

## Consistency And Summarizer Checks

```sh
node benchmarks/world-best/claim-consistency-smoke.js
node benchmarks/world-best/public-benchmark-page-smoke.js
node benchmarks/world-best/summarize-live-codex-api-regression.js
node benchmarks/world-best/quality-safe-proof-fixture-smoke.js
```

Expected labels:

```text
claim consistency smoke passed
public benchmark page smoke passed
live Codex API summarizer adversarial regression passed
quality-safe proof fixture smoke passed
```

## Legacy Artifact Boundary

The historical JSON and JSONL keep their original schema and bytes. Do not read
field names in those files as verification or CostLayers attribution. Inspect
the hash-bound quarantine policy instead:

```sh
node -e 'const r=require("./proof_artifacts/legacy-artifact-registry.json"); console.log(r.registry_status, r.policy.current_verified_proof, r.policy.promotion_status, r.artifacts.length)'
```

Expected labels:

```text
active_quarantine false prohibited <registered-artifact-count>
```

## Separate Provider-Avoidance Smoke

```sh
jq -r '"provider_smoke_requests=\(.counters.requests)", "provider_smoke_called=\(.counters.provider_called)", "provider_smoke_avoided=\(.counters.provider_avoided)", "provider_smoke_quality_safe_rows=\(.counters.quality_safe_rows)", "certificate_sha256=\(.certificate_sha256)"' proof_artifacts/world-best-live/provider-avoidance-scale-smoke.json
```

Expected labels:

```text
provider_smoke_requests=12
provider_smoke_called=1
provider_smoke_avoided=11
provider_smoke_quality_safe_rows=11
certificate_sha256=dfd6c486e46f6a4c5542e2c871583bc894373f205d2772f36184a30f00c93435
```

This is a repeated-request gateway smoke, not the legacy mixed-arm benchmark,
a customer invoice export, or a rank proof.

## Public Context Matrix

```sh
jq -r '"context_rows=\(.summary.result_count)", "context_ok=\(.summary.ok_count)", "context_quality=\(.summary.quality_gate_ok_count)", "certificate_sha256=\(.certificate_sha256)"' proof_artifacts/world-best-public-10x3/bench-suite-report.json
```

Expected labels:

```text
context_rows=120
context_ok=120
context_quality=120
certificate_sha256=abaa111b2d716ec4a301630f8c7195a24b5891f2be0d15f9b9d19d7f9dcdc60a
```

## Recompute Quarantine Hashes

```sh
node benchmarks/world-best/claim-consistency-smoke.js
```

The smoke recomputes every file-byte hash in
`proof_artifacts/legacy-artifact-registry.json`, requires exact registry coverage
for legacy JSON/JSONL and screenshots, and rejects direct promotion links from
current public proof surfaces. Expected label:

```text
claim consistency smoke passed
```

Still missing external proof: a real customer invoice export, a complete live
task matrix, and independent review with sufficient raw evidence.