# MapleJockey pilot verification

17 November 2026 | v0.3 | Local synthetic scope

## Implemented flow

The standalone Canadian-branded workbench records scope and target, captures a retained file baseline, exports an unencrypted portable bundle, re-imports and restores files, compares bytes, performs selected file-reading checks, exposes failure simulations, rehearses rollback and exports metadata-only reports and assessment JSON. Source and plan changes invalidate earlier results. Pending asynchronous file reads cannot restore a stale result after scope changes.

## Verification summary

| Verification | Result | Boundary |
|---|---|---|
| Core engine and independent adversarial tests | 35 passed, 0 failed | Includes SHA-256 comparison with Node crypto, maximum file round-trip, corruption, attacker-rehashed content, missing/extra files, path attacks, Unicode/name collisions, malformed input, byte limits, unsupported formats and report-content separation |
| UI event/rendering harness | 18 passed, 0 failed | Uses actual engine and report functions with DOM stubs; includes guided and manual flows, pre-read limits, scope invalidation, pending-import cancellation, rollback and safe text preview |
| Existing assessment regression | 10 passed, 0 failed | Existing assessment core remains valid after companion navigation was added |
| Actual isolated disk exercise | 8 of 8 files match; 14 checks passed; 0 failed; 6 Not tested | Fresh directory restoration, disk readback, selected format/domain checks and unchanged source |
| Altered bundle exercises | 3 of 3 detected | Changed content, attacker-rehashed content and a missing file |
| Disk rollback | Passed | Only the newly created destination was removed; retained source hashes and calculation still match |
| Independent retained-source inspection | 8 hashes match Node crypto | Confirms current bytes and cleanup state, not cryptographic proof of the historical sequence |
| Native graphical browser | Unverified | Browser process launch is restricted; no layout, keyboard, native chooser or download claim |
| Live provider migration | Not tested | No vendor accounts, uploads, permissions, regions, key service or production recovery tested |

The latest run ID, timestamps, phase timings and source hashes are in results/disk-rehearsal.json. Elapsed timing covers export, disk restoration and independent byte/domain validation. It excludes source preparation, negative tests, rollback and repeated checks during report generation. Rollback has its own recorded timing. No supplier throughput or business recovery objective is inferred.

## Meaningful restored-data check

The restored metadata identifies the restored CSV. Four observations are parsed and used to calculate a total of 40. Accents, quoted fields, multiline content and a formula-like literal are preserved as text. Binary and SVG byte equality is verified, but their application behaviour stays Not tested.

## Corrections made during review

- Replaced a base64 pattern that could overflow the regular-expression stack at the maximum permitted file size.
- Kept all-unsupported functional findings Undetermined and separated them from integrity/timing findings.
- Escaped plan exclusions in readable reports and removed unsafe Markdown appendices.
- Cleared previous results before malformed/oversized import attempts and invalidated pending reads on scope/source changes.
- Cleared plain-text preview content on rollback, source invalidation, close and session discard.
- Made the timing boundary explicit and kept live-provider controls Unknown or Not tested.

## Evidence limits

Hashes establish agreement with a retained baseline, not source authenticity, safe content or legal control. Parsing is not complete application compatibility. The result file and assessment are editable and unsigned. Blank operator sign-off is not an approval. The actual fixture contains no private workload data.

The independent review is in review/INDEPENDENT_REVIEW.md. Exact test output is in results/automated-test-results.txt and ui/smoke-test-results.txt. The test scripts and disk exercise are reproducible from the package with an existing Node installation.

## v0.4 interface review

The compact layout, bundled Lato/Noto Sans fonts and stronger text contrast are included in the 17 November 2026 update. See `validation_v04/VALIDATION_REPORT.md` at the package root for the consolidated 151-check result, font checks, 27 colour-pair checks and remaining browser/provider gates. Earlier results above retain their original scope.
