Two high fidelity findings are recorded in .imd-findings.json, with exact source snippets and reproductions:

• Pending requests insert a bordered “Swarm progress” section between the Mint facts explanation and 01.
• The populated Catalog Explorer retains two separators around empty space before “Community and raw metadata.”

Both reproduce in Chromium and WebKit at 1280 and 390 px.

Preservation and static constraints are met: contracts, chain, ABIs, routes, RPC pool and integrity checks remain unchanged. Tests cover each numbered item; the committed export’s 65-asset manifest verifies. Same-name publication is after review.

The two taste flags concern inherited capitals and monospace prose, retained under the scope constraint.

Sworn presents wallet facts as a green ledger on black, supported by a hexagonal record card. No website files were changed; only the findings report was written.

I reproduced both reviewer findings, fixed them, rebuilt dist/ and reran the checks. Everything passes in Chromium; WebKit was not run because this machine lacks its system libraries.

Finding 71f6d97f (Swarm progress between the Mint facts explanation and 01): fixed. For a wallet with a pending fact, the progress section came straight after "Why add facts after minting?" and drew its own line. That line followed the explanation's link lines with empty space between.

• The section now sits at the end of step 02 (web/src/claim.tsx). Its feed, anchors, label and copy are unchanged, and it no longer draws a line.
• Its heading is now a small sub-part title inside the step, matching item 8.
• The explanation runs into 01 again with space and no line, and the lines before 02 and 03 stay.

Finding 9251d9aa (two lines after the last Explorer fact): fixed. Every fact row drew a line under itself, including the last one, so the next section's top line doubled it. The last row in a fact list no longer draws a bottom line (web/src/styles.css). Lines between facts and the next section's top line stay. Hidden facts are removed from the page, so this is always the last visible row.

Tests

• web/tests/layout-refinement-browser.mjs has a new case for each finding:
  • Mint with pending fact 601 checks where the section sits and that it has no line of its own.
  • The populated Explorer is checked with and without a wallet.
  • Both are measured at 1280 and 390 px with the empty-separator detector.
• The suite now passes 7 groups and 52 measurements with no errors.
• swarm-browser.mjs and site-update-browser.mjs asserted the old position (before 01). I changed them to expect step 02 because the finding overturns that position; both pass.
• npm test, the typecheck, the Prettier check and every other browser suite pass. Payments use the fake wallet only, so nothing is signed or sent.

Export and docs

• dist/ is rebuilt, a second build gives identical files, and npm run verify:export passes.
• The design probe gives 14 of 16, the same two inherited flags already recorded.
• DESIGN.md, web/README.md and web/VALIDATION.md describe the change.
• Answers to both findings are in .imd-responses.json.

I haven't committed. The earlier rounds left committing to the submission system, which writes the imd … commits.

Both of last round's high findings are fixed, and I found nothing new I could reproduce. .imd-findings.json holds {"findings":[]}.

Last round's findings

• Swarm progress section between Mint facts and 01 (id 71f6d97f…): fixed. In web/src/claim.tsx the pending-fact section now sits at the end of #step-02, with the same entries, aria-label and copy, and its title is a small sub-part heading. I re-ran the case: token 7, traits 101/201/301, allowance 1e22, pending fact 601, empty fake API answers, connected through Browser wallet. The block after "Why add facts after minting?" is #add-traits (01). The feed is inside step-02 and has no top border. The separator detector finds no empty pairs at 1280 or 390.
• Two empty separators in a row in the populated Explorer (id 9251d9aa…): fixed. .fact-rows > .fact-row:last-child { border-block-end: 0 } in web/src/styles.css removes the bottom line under the last row. On #/explorer?address=0x2000…0001, with and without a wallet, at 1280 and 390, the last row's bottom border is 0px and the detector reports nothing. I also checked a case the builder's test doesn't cover: clicking "Show ineligible facts" so the hidden rows appear. The last row still has no bottom border and the detector still finds no empty pairs, in all four width and wallet combinations.

Requirements: the builder's suite web/tests/layout-refinement-browser.mjs passes all 7 groups: 52 route and width checks, no errors, every brief item covered.

1. Footer RPC selector and Retry reads on one line at 1280: met (footer row geometry).
2. "Follow progress on the Wallet page" stays enabled while a transaction is busy: met (dialog test, which neither signs nor sends).
3. The six CATALOG links point to commit c4ff3238 in the deployed repository, with the same anchors: met (a.source hrefs).
4. Mint page loses "Add facts" and the "Who can · Cost and risk · Who benefits" block: met (Mint structure test).
5. No empty separator pairs at 1280 and 390, with and without a wallet: met (detector across all routes, plus the pending and Explorer cases).
6. Exact 01/02/03 titles and the Etherscan Read Contract link: met (.step-title text and link check).
7. Each step opens with its Catalog intro: met (Mint structure test).
8. Heading sizes distinct and decreasing at every level: met (heading geometry).
9. "Show ineligible facts" appears only when a fact is hidden: met (zero-hidden test).
10. "Why add facts after minting?" sits right under the Mint facts intro and runs into 01 without a line: met (.why-facts is followed by #add-traits).
11. 01 has no ineligible control and lists only the direct facts the wallet holds; the Catalog lists every direct rule: met (held-only test).

The idea, in one sentence: a terminal-style registry where each fact about a token is a ruled row with its proof, and steps 01 to 03 climb from facts read straight off the contract to facts the community swears to.

Not run: WebKit, because I installed only Chromium for these checks, and the full taste probe, since this round was about settling the two fixes. To run the tests I installed Playwright and viem one folder above the repository. My scratch script is deleted and no tracked file changed.