# SYNTHETIC ReleaseProof judge walkthrough

SYNTHETIC inputs and outcomes only. Report-generation recovery is not clean native QC, deployment verification, human approval or production release. G0 OPEN / NOT APPROVED; G1 CLOSED.

Open index.html first for the observed attempt sequence and available artifact links. If the entry page could not be written, use this Markdown and receipt.json; absent files remain unavailable.

Observed demo outcome: SYNTHETIC_DEMONSTRATION_PASSED.

1. Inspect valid-input/packet.json and the separate valid-input/expected.json; receipt.json records their exact byte counts and SHA-256 digests. All provider-like identities are invented fixture data, separate from recorded CI execution identity when present.
2. Compare negative-packet.json: only required validation_exits_raw is omitted; its provenance is retained. Inspect negative.stdout, negative.stderr and the negative attempt in receipt.json. Accepted rejection requires real exit 2, the missing-field diagnostic and an empty fresh negative-output directory. Spawn errors fail the demo.
3. If available, inspect recovery-packet.json. The intended single correction restores validation_exits_raw using the original bytes. The receipt records recovery_byte_identity_established only after the runner verifies the byte relationship; do not infer it from a partial run. The recovery is a separate invocation into recovery-output, with separate stdout/stderr and receipts.
4. Open recovery-output/review.html and review.json. Successful generation requires exit 0, matching valid JSON/HTML, five visible synthetic High findings, FAILED_WITH_FINDINGS, incomplete synthetic provenance, pending human release and deployment.authorized=false.
5. Check receipt.json outcome and recorded CI execution identity. This generated demonstration contains no CI execution identity. No identity is inferred from the build host. A failed runner exits nonzero and retains its failure receipt; never treat an incomplete attempt as successful evidence.