# Local Merkle validation receipt

Date: 2026-10-25. D30-F05-A05. Video-ready private draft, no remote publication. Raw command output: local-validation.json (retained under canonical package tests/evidence). Actual browser results: visual-qa.json (retained under canonical package tests/evidence).

- 143 complete independent expected build/verification outputs agree with actual Python and TypeScript execution. Every output field, proof hash and trace node is compared, not just status.
- 11 Python test methods pass. Supplemental coverage includes independent stack roots for sizes 1–65, parent literal full-G roots/paths, NFC/NFD/leg-order byte mutations, strict containers and line terminators, semantic ordering, no input aliases, native crypto failure and the opaque-byte trust boundary.
- Both runtimes execute the full 4096-entry limit, checking an independently computed repeated-leaf root, 8191 postorder nodes and a 12-sibling proof. The 16384-byte entry boundary and limit rejection are included in full literal fixtures.
- Actual native browser WebCrypto runs at 390, 768 and 1440 pixels: 15 scenarios per width, 45 runs, 249 guided reveal states compared with the standalone compiled core, and 429 complete independent expected outputs.
- The browser asserts the one-sibling correction path, two-sibling funding path, 5/7-leaf odd splits, wrong bytes/index/root/tree-size/scope, short path rejection, missing trust, empty tree, repeated bytes and one empty byte entry.
- All controls are at least 44 pixels. Begin/Back/Step/Play/Pause/Reset, speed, keyboard arrows, timer stopping on scenario change, terminal disabling and reduced-motion single-step Play pass. Full exact input hex remains inspectable.
- Three native SVGs and Mermaid parse/render pass. Actual visual inspection covers all full native figures, workflow, phone top/full lab. Exact hashes wrap inside cards without clipping. The mobile top uses readable controls and an informative completed correction proof.
- Zero browser console errors, network requests or horizontal overflow. Full package validator: zero errors and zero warnings. Guided-playground validator: pass.

Screenshots: 390-pixel top (retained under canonical package tests/evidence), 390-pixel full (retained under canonical package tests/evidence), 768-pixel full (retained under canonical package tests/evidence), 1440-pixel full (retained under canonical package tests/evidence), tree (retained under canonical package tests/evidence), paths (retained under canonical package tests/evidence), assurance boundary (retained under canonical package tests/evidence), workflow (retained under canonical package tests/evidence).

The independent oracle generator imports no production algorithm. Its initial malformed-input request alias was corrected by copying the source entry array; this isolated later mutations without changing the independent mathematical expectations or relaxing the production core. Parent before-core G framing bytes and roots remain pinned unchanged.

The pure core validates opaque hex and proof math only. A pass does not establish financial source validity, canonical G framing, complete/current history, signer authority, non-equivocation or external retention. Parent integration must actually validate/encode source history and bind retained commitments separately.

## Visitor integration

Root visitor integration pending. Local package QA does not claim production website publication or live external assurance. Parent integration will append isolated visitor and source-adapter receipts.