Provenance is the product

How Goodwattle verifies provenance.

Before anything is offered to a buyer, we check who owns it, whether it has been public, what is third-party and what personal information it holds. This is the standard we apply.

A kookaburra perched on a vintage computer among golden Australian wattle.
Good things grow from good work.

The checks, in order

Seven checks before delivery.

  1. 1

    Ownership

    Who wrote it, under what employment or contract, and whether IP was assigned. We ask for the documents that show the chain.

  2. 2

    Never public, never licensed

    We compare the repository with public code hosts and ask the owner to attest that it has not been published or licensed for AI use.

  3. 3

    Third-party code

    Vendored dependencies, copied code and licence headers are identified with SPDX scanning, and third-party material is excluded.

  4. 4

    Secrets and personal information

    Keys, tokens, emails, phone numbers, tax file and IRD numbers and other identifiers are removed or replaced, and counts are logged.

  5. 5

    History and review

    Commits, contributors, history span and reviewed pull requests or changelists are counted from the repository itself, not self-reported.

  6. 6

    AI authorship

    The owner discloses any substantial AI-generated share, and we look for signs of it.

  7. 7

    Legal review

    Rights and scope are reviewed before delivery. Anything uncertain is left out.

The manifest

What we record.

Each asset is delivered with a manifest of these fields.

FieldWhat it records
Language mixShare of original source lines by language.
Lines of codeOriginal source, excluding vendored, generated and third-party code.
Reviewed PR countMerged pull requests or changelists with at least one human review.
History spanFirst and last commit dates, with gaps and rewritten history flagged.
Contributor countDistinct human authors after merging duplicate identities.
Test ratioTest files and lines relative to source.
Licence scan resultSPDX identifiers, licence headers and vendored code found, with exclusions listed.
PII scan resultSecrets and personal identifiers found and removed or replaced, with counts.
Provenance attestationsOwnership chain, never-public and no-prior-AI-licensing statements from the owner.

Out of bounds

What we never buy.

  • Code or data you don’t own or can’t license
  • Customer personal data
  • Health records and other sensitive information
  • Material a client, publisher or platform agreement forbids you to offer
  • Credentials and secrets (these are removed, never delivered)

For owners: nothing is uploaded until you choose to share it, and a detailed review only happens after confidentiality is agreed.

Try it yourself

Get a feel for it first.

Browser-only and illustrative. Nothing is stored or sent unless you choose to.

Owners and buyers

Provenance you can check.
Terms you can read.

Owners can start with a short enquiry. Buyers can send a brief.