Updated source and dist/. Passed 27 unit tests, 95 Chromium/WebKit cases, and 16/16 design checks. Screenshots and results are in the validation report. Submission files total 6.27 MiB.

A1_INTEGRATION_PENDING: live actions remain disabled; actual A1 local-chain and mainnet tests are NOT_RUN. Handoff included.

Readiness is blocked by the high retry defect. Five reproducible findings are recorded in .imd-findings.json. Product source, configuration and export remain unchanged.

The A1 dependency still reports executing, without public delivery or project head. A1_INTEGRATION_PENDING is honest. Actual A1 local-chain and mainnet integration remain NOT_RUN. The faked item below concerns a demo retry’s approval status; live projects are not silently simulated.

• High — bug: refused retries bypass fee review and fabricate confirmed approval. At 390px, sponsor a proposal with account A and a refused callback, switch to B, select an unknown purchase receipt, then retry. No sponsorship preview appears; B’s approval and purchase immediately show confirmed despite no B allowance. The retry also lacks its own recovery record. Route retries through fresh fee/spender review, exact approval and transaction journaling. Workspace.tsx:322
• Medium — fidelity: purchase identifiers cannot be copied. At 390px and 1440px, progress exposes only fragments such as 0x6308…329a, without the complete hash, copy button or explorer link. Reuse CopyId for purchase hashes and sponsors. Workspace.tsx:312
• Medium — bug: changing an HTML proposal to image creation strands selected inputs. Select an HTML input, then change the type to image. The checked input becomes disabled while validation rejects it. Clear incompatible inputs and reset the mode when selecting image. ProposeForm.tsx:112
• Medium — bug: Markdown table cells overlap on phones. A valid twenty-column Chinese table gives 15px glyphs only 12.1875px-wide cells at 390px; neighbouring glyphs overlap by 0.8125px, increasing to 4.3125px at 320px. Preserve readable columns inside a keyboard-scrollable table region. styles.css:193
• Medium — bug: successful file verification leaves progress permanently pending. Create a root with one document, sponsor it, verify and download its sole output. The asset is verified while progress still says 文件校验 / 等待中. Reflect descriptor-scoped local verification in that stage without changing chain truth. fixtureAdapter.ts:394

Final checks passed: TypeScript, 27/27 unit tests, 95/95 Chromium/WebKit browser cases, byte-identical rebuild and 58/58 release pins. Temporary font-related WebKit failures were resolved before the final rerun. Passing existing tests does not cover away the independently reproduced defects. No repaired revision was supplied to recheck; the high finding is submitted through the findings mechanism for reopening.

The visual idea is a Chinese project workspace using stone and green surfaces, with persistent amber treatment distinguishing demonstration state.

For the complete requirement accounting below, *M

Readiness is blocked by two high findings, recorded in .imd-findings.json.

• High — bug: submitted static export still runs all five pre-repair flows. The export loads the old bundle. Twelve source/test hashes also disagree with the release manifest. Rebuild dist/, repair failures and regenerate matching evidence.
• High — bug: wide-table repair expands the preview beyond its asset panel. After rebuilding, the preview grid becomes 1,225px wide, producing 875px overflow at 390px and 945px at 320px. On desktop it overlaps the proposal panel. Constraining the grid to minmax(0, 1fr) eliminated these failures in browser measurements.

The retry review/recovery, sponsorship identifier copying, image-switch reset and file-verification status fixes hold in rebuilt source. The table fix does not. None of these repairs reached the submitted export.

Verification: typecheck, build and 28 unit tests passed. Rebuilt regression tests finished 20 passed / 5 failed across five browser projects; all five failures concern the table. Six additional WebKit desktop/mobile journey and wallet tests passed. A broader run lost its temporary preview and remains inconclusive.

A1 now has a public delivery at commit 9b0294e7ab19060bd00c489fbe032cf248528e0d. W1 imports none of it. Actual W1–A1 local-chain integration and mainnet validation remain NOT_RUN.

“Met, fixture” below means observed UI/demo behavior, without chain proof.

Both findings are fixed. Rebuilt dist/, refreshed release pins, and wrote .imd-responses.json.

Passed: typecheck/build, 28 unit tests, 120 Chromium/WebKit cases, and 16 design checks.

A1 source is pinned; W1 live integration remains NOT_RUN. Evidence and handoff are in docs/validation/ and docs/A1-HANDOFF.md.

Both previous findings are fixed. .imd-findings.json contains {"findings":[]}. No product files were changed.

• bug: stale submitted export — fixed. A clean rebuild matches all six submitted export files byte-for-byte. All 60 release pins and 53 tested-tree pins match. The five regression cases pass in all five browser configurations.
• bug: wide-table preview overflow — fixed. The grid constraint holds in Chromium at 320/390/1440 and WebKit at 390/1440. Page overflow is 0px. The table scrolls inside regions measuring 216/286/516px against 1201px of content; glyph overlap is zero. Keyboard scrolling, download hit testing, and actual downloads pass.

Independent checks completed 120 unique browser cases, 28 unit tests, TypeScript, rebuilding, and taste@1. Long combined commands were interrupted and completed in smaller runs. One WebKit journey exceeded the original 45-second timeout and passed alone with a 120-second budget; product code and assertions were unchanged. Reviewer results and log references preserve these qualifications.

The site is a Chinese project workbench, carried visually by restrained green actions and warm neutral panels.

“Met” below describes the delivered frontend, explicit unavailable-live behavior, and tested fixtures where applicable. It does not establish live A1 functionality. A1 source is available at the pinned commit, but W1’s browser bridge, verified product deployment, and actual W1-to-A1 acceptance remain unfinished. Those qualifications are preserved in the A1 handoff.

I rewrote README.md as documentation only, and I did not re-run any build or tests. Every result it reports is copied from the logs already in docs/validation/.

The repo already had thorough docs. They lacked a U01–U10 table, a per-page permissions and failure-mode table, and sections on the adapter contract, public config, safety boundaries and an evidence index. I added those and a status block at the top, and kept the install, use, wallet, files and validation sections as they were.

• Status: the README says up front that W1 is not complete and is marked A1_INTEGRATION_PENDING. The live route is unavailable and fails closed. All working flows are in the fixture-only #/demo mode, which never asks a real wallet to sign and is not safe for real funds.
• U01–U10: U01–U09 are PASS in the fixture browser tests (120/120 cases). U10 is marked FAIL/blocked because there is no browser bridge to A1 and no verified deployment. The A1 local-chain and live columns are NOT_RUN for every row. The README states that A1's own SDK tests and 33-transaction harness were replayed separately and are not W1 evidence.
• Brief vs. what exists: a short table lists the gaps. Examples: no A1 SDK, ABI or address is imported, no signing path is reachable, only PNG was tested, and the size limits are fixture values rather than A1's.
• Review: the README marks the automated measurements and the contributor's own source review as EXECUTED, and a second-author security audit as NOT_RUN.

I did not touch anything outside README.md.