docs/PACKAGING_VALIDATION.md

# Packaging validation — 20 October 2026

This record concerns preparation of the supplied ELYRION III release for the PureOne/ELYRION-III Hugging Face dataset repository. Scientific release 3.0 is unchanged; repository packaging is version 1.0.0. No remote repository creation or upload was performed during preparation.

| Check | Result |
|---|---|
| Original source checksums | 23/23 listed source hashes match |
| Original artifact retention | All 24 original file byte sequences retained; original README/checksum file relocated |
| Fresh main replay | 1,305/1,305 checks passed in a disposable copy |
| Fresh separate audit | 108/108 checks passed in a disposable copy |
| Figure replay | All five figures regenerated |
| Structured numerical export | 36 dose + 4 constants + 16 diffusion + 4 hazard rows exactly match the original ledger |
| Claim records | 12 records, unique IDs and one consistent string-field schema |
| Dataset card | Parsed locally with the official DatasetCard client and YAML parser; five explicit data configurations |
| Local card links | All relative targets checked |
| Uploader tests | 14/14 offline mock tests passed |
| Upload preview | Real command-line dry-run against the complete payload; no credentials or network |
| Payload inventory | SHA-256 and byte counts checked against PACKAGE_MANIFEST.json; no extra files |
| Unix launcher | Shell syntax check passed |
| Windows launcher | Statically reviewed; not executed on Windows |
| Live authenticated Hub upload | Not performed |

The main and separate scientific checks originated in the supplied research session. The additional source inspection and packaging checks were performed by AI agents during this preparation. They are not outside peer review.

The source replay details are in source_replay_audit.json. Uploader fixture-test details are included in the outer release folder as UPLOADER_VALIDATION.json. The outer VALIDATION.json records the complete-payload checks and their output.

The manifest detects modifications relative to its own contents; it does not authenticate authorship or prove the science. The data are synthetic model outputs and editorial claim summaries. No implementing chemistry, experimental feature measurements, independent priority assessment, formal theorem replay, or new reuse-license grant is supplied.

Repository preparation is complete. Physical implementation and historical novelty remain unresolved.