# Verification record

Candidate: `budget-unrecorded-navigation-1`. Scope: isolated local contribution,
not a Hub release. Checked on 2026-11-07 UTC.

## Provenance and license

Downloaded the source ZIP from the link on
https://free4allhub.com/tools/budget/:

https://free4allhub.com/tools/budget/budget-workbench-source.zip

Its SHA-256 is
`64e665456f2f1579aa3b1aeb88eb70cdf387c153341037095edc4d138c7289f7`.
The exact eight-entry archive is included as `baseline-public.zip`. It matched
the current Hub build's downloadable ZIP and an earlier saved public download.
The archive's license is MIT with an explicit workbench-only scope. Its original
copyright/permission notice is retained byte-for-byte. Original style and core
tests are also unchanged. No spreadsheet package or unrelated Hub asset was added.

## Reproduced automated checks

- Original ZIP extracted into a clean directory: both module syntax checks and
  all 19 original tests passed without installing packages.
- Independent navigation contract tests were written before inspecting the
  implementation: all 9 passed. Coverage includes exhaustive small circular
  search arrangements, exact blanks versus known/signed zero, invalid nonblank
  values, frozen input, malformed rows/cursors, 200-row bounds and invalid
  unrelated draft fields.
- A separate source/artifact review confirmed the exact candidate/source match,
  retained MIT scope and truthful return instructions. The patch applied to a
  fresh baseline with zero fuzz and produced a byte-identical candidate tree.
- Portable candidate archive: both module syntax checks and all 28 tests passed
  from a fresh extraction.
- Two independent output builds are byte-identical, including the source ZIP,
  standalone tree, patch and manifest. They match the included `output/` tree.
- Review found one standalone-footer mismatch during preparation; the final
  source restores the original standalone footer and adds only the navigation
  markup there. No missing source-download link was introduced.
- See `reproducibility-tests.txt` for the replay output. Timing values are
  incidental and are not part of deterministic build output.

## Browser interaction checks

The candidate source was served through the supported local preview and checked
in the available desktop Chromium browser, with fictional built-in figures only.
The tested app/core/page/index/style bytes match the delivered candidate source.

Passed:

1. Fictional example loads. The new control reports one blank actual and focuses
   entry 2. Actual net remains partial at USD 1282.75.
2. JSON downloads immediately before and after focus navigation have identical
   bytes. Both hashes are
   `6c7b2debdbc422bbdbd16b59c5d74e2616775a27e5b29a4ac83e004715d630e4`.
3. Recording known zero disables the control when no blank actuals remain.
4. Clearing two actuals enables the control with the correct remaining count.
   Repeated activation wraps in row order; Enter activation also moves focus.
5. An unrelated invalid planned amount (`bad`) stays unchanged while navigation
   works. Totals remain hidden and JSON download remains disabled.
6. Currency review and incoming starter review disable navigation. Canceling
   preserves invalid edits and restores navigation.
7. Adding/removing entries resets the navigation position and chooses the next
   existing blank field. No stale row-index failure was observed.
8. Actual downloaded JSON opens in a fresh tab after review/explicit Open.
   Navigation finds its original missing actual. Changing it to `200.00`, saving,
   then reopening that second file in another fresh document preserves the
   revision and complete net USD 1482.75. Navigation is disabled with no blanks.
9. A 360 CSS-pixel iframe checks the same standalone file. Usable document width
   was 345 pixels with its scrollbar; scroll width was also 345, so no horizontal
   overflow was observed. Adding a row and navigating focused its blank actual,
   with the visible focus ring and field fitting the narrow layout.

`browser-downloads/` contains the two identical before/after fictional files and
the revised fictional file. The revised file hash is
`d0dfc6a052c53f255c06a48951974b33b1daf852e71d70ff948fca8f0694e1d2`.

## Limits and remaining release review

The initial separate headless browser attempt could not launch in this
environment. It did not produce a passing test run; the browser checks listed
above were then performed through the supported preview/browser route.

The transient asynchronous file-read lock was inspected in source, not held open
and asserted in a browser. File-review replacement/cancellation uses the same
lock path, but only starter-review cancellation and fresh-file acceptance were
exercised here. No comprehensive CSV browser interaction, 200-row browser stress
run, physical phone, touch keyboard, screen reader, other browser engine, real
beginner session, performance benchmark, financial accuracy or professional
security/accessibility certification is claimed. Original CSV/core behavior
remains covered by the shipped tests.

The full hosted Hub wrapper, project bundle adapter and wider release suite were
not integrated or rerun as part of this isolated kit. The owner must coordinate
that review before any live integration. Public acceptance, deployment, a real
outside contribution and an established intake endpoint remain unclaimed.