Scanfolk digital artifact audit — version 1
Published: 2026-10-03
Method ID: scanfolk-digital-artifact-audit-v1
Record file: /evidence/scan-artifacts-20261003-v1.json

Purpose
Identify the exact displayed image bytes and payload, reconcile historical
records by SHA-256, and run a small reproducible digital decoder audit.
This is not a physical camera, print, distance or lighting benchmark.

Artifact identity
Each record contains the displayed asset URL, SHA-256, dimensions and expected
payload. Historical checks must match the tested file hash. A catalog preview
and its full source reference are different artifacts; reference results do
not certify the preview. Historical counts describe their original configured
matrix, not the four checks in this audit. Accepted historical gates may include
failed cases; acceptance does not mean every digital variant passed.
Botanical attempts include configured digital transforms and decoder attempts;
the hit count is not a real-device scan success percentage.

Audit procedure
Read the current image as RGBA. Composite transparency over solid white.
Resize to an exact square with Pillow LANCZOS. Decode with ZXing-C++ without
opening the destination. Four cases: RGB at 1024 px, grayscale at 1024 px,
green channel at 1024 px, RGB at 1536 px. Preserve every decoded string,
exact match, unexpected payload and failure. Compare payloads literally.
A decoder result is evidence for that case only. A wrong payload is a failure.

Provenance
Python/Pillow/ZXing-C++ package versions and UTC audit time are in the record.
Reproduction script: scripts/audit_seo_scan_evidence.py in the Scanfolk project.
Animal and botanical image hashes are compared with their preview manifests.
Abstraction reference hashes are compared with the abstraction manifest.
Cherry's current displayed artifact has fresh decoder checks; its older
reference matrix is not assigned to this displayed derivative.

Limits
This four-case audit does not rerun each historical production gate. Abstract catalog previews have their own case results in this audit; their
historical labels still identify the full reference as evidence owner. A new payload, regenerated
image, export derivative or changed file needs its own validation. Final
placement tests must record device/app, print size/material, distance, angle,
lighting and failures. No iOS/Android camera or print test was performed here.

Versioning
Keep this dated dataset immutable. Publish a new method/data version when the
procedure or artifacts change; retain failures and corrections. This method
does not support universal scan guarantees or a consumer scan-rate percentage.