nanoru
https://reg34.smp.ne.jp/regist/is?SMPFORM=nhrh-tfohq-e442ee78a4c692b49d3514c224b3805f
Goal: reach the confirmation screen (入力内容の確認)
Run imperial-hotel-2, 2026-11-07 17:19:12 UTC
Public page. Nothing was submitted.

One name was shown changed on the confirmation screen: Letters with accents.

The kanji-name baseline reached confirmation with the name shown exactly; this one reached it with the name displayed differently. See the evidence.

Comparison

Same form, same goal, same contact details; the agent was told to make the same fixed non-name choices (date rule, first options, message) in every case. The observed non-name rows on the confirmation screen matched the baseline's row for row in every case (measured; this compares the rows the screen shows, not hidden inputs or the actions taken). Names other than the baseline are name-format test cases, not verified identities. Recorded run, not live.

Name formatOutcomeWhat the form did to the name
Kanji name, baseline
山田 太郎
Confirmation screen reached
Confirmation screen reached showing 山田 太郎 and ヤマダ タロウ; safe checkpoint, not submitting. (13 actions)
No measured problem. Final screen
Letters with accents
Nguyễn Thị Hương
Reached, but the name was shown changed
Confirmation screen reached.
moreThe name is shown as HTML entities (Nguyễn Thị Hương), so the diacritics are garbled. The katakana reading displays correctly. (name changed: review screen recognised, but a typed name value is shown changed in its review row) (13 actions)
NameSei (review row) altered
review row shows the name as HTML entities (double-escaped): review row "氏名" shows "Nguyễn Thị Hương" (nanoru measurement)

Examples test name formats. Except for the creator's name, they are sample inputs, not research participants.

Evidence

Detail: the name field and the error
Nguyễn Thị Hương: the same screenshot, magnified around the measured field, nothing edited. Nguyễn Thị Hương, full screenshot
Baseline, final screen
Baseline 山田 太郎: reached the review screen. Baseline, full screenshot

Site defect, supported by evidence. Proposed by Claude from measured facts, not yet reviewed by a person.

Field
氏名 (review row)
Entered, NameSei
Nguyễn
Entered, NameMei
Thị Hương
Shown on confirmation
Nguyễn Thị Hương
The page said
(no error text)

The values above were measured by code. What follows was written by Claude from them.

What happened
The confirmation screen's 氏名 row displayed the surname as "Nguyễn" instead of "Nguyễn", while the baseline's row showed 山田 太郎 correctly. This is nanoru's own measurement comparing the typed value with the review row text; the site showed no error message.
Why it matters
A person with a name of this shape, such as a Vietnamese name with diacritics, would see their name corrupted on the review screen before sending. The record the site keeps may carry the same corruption.
Basis
Page 27 item 3 asks name fields to accept Latin letters; it does not explicitly address diacritics or how a name is displayed, so the corrupted display is only inconsistent with this guidance by analogy. The guidance is for Tokyo Metropolitan Government forms and is not law. (analogy, for review; GovTech Tokyo Input Form Guidance v1.1, p.27 item 3 (accept Latin letters), by analogy)
Cause
Reproduced from the displayed text, not confirmed in source: the name appears to be double-escaped, with numeric character references such as ễ escaped again so they show literally. One possible trigger is a non-UTF-8 form encoding that makes the browser send characters outside the charset as references, which the server then escapes again.
Proposed repair
Serve and submit the form as UTF-8 (meta charset and accept-charset="UTF-8"), and HTML-escape the name exactly once when rendering the confirmation row. Keep the existing maxlength=64 and any server-side validation unchanged so invalid input is still rejected.
The repair brief as text

Run details

山田 太郎: 13 actions, reached the review screen
  1. fill n1 山田 family name before · after
    • NameSei: accepted
  2. fill n2 太郎 given name before · after
    • NameMei: accepted
  3. fill n3 ヤマダ katakana family before · after
    • NameSeiKana: accepted
  4. fill n4 タロウ katakana given before · after
    • NameMeiKana: accepted
  5. fill n7 test@example.com email before · after
    • email: accepted
  6. fill n8 test@example.com confirm email before · after
    • email:cf: accepted
  7. fill n9 1500001 postal code before · after
    • ZipCode: accepted
  8. fill n10 東京都渋谷区神宮前1-1-1 address before · after
    • RegidentialAd: accepted
  9. fill n11 09012345678 phone before · after
    • Phone: accepted
  10. fill n15 nanoru による自動入力テストです。送信はしません。 required message before · after
    • message: accepted
  11. click n17 agree to privacy policy before · after
  12. click n18 go to confirmation before · after
  13. done n1 Confirmation screen reached showing 山田 太郎 and ヤマダ タロウ; safe checkpoint, not submitting. before · after
Nguyễn Thị Hương: 13 actions, reached the review screen, name shown changed
  1. fill n1 Nguyễn family name before · after
    • NameSei: accepted
  2. fill n2 Thị Hương given name before · after
    • NameMei: accepted
  3. fill n3 グエン family kana before · after
    • NameSeiKana: accepted
  4. fill n4 ティ フオン given kana before · after
    • NameMeiKana: accepted
  5. fill n7 test@example.com email before · after
    • email: accepted
  6. fill n8 test@example.com email confirm before · after
    • email:cf: accepted
  7. fill n9 1500001 postal before · after
    • ZipCode: accepted
  8. fill n10 東京都渋谷区神宮前1-1-1 address before · after
    • RegidentialAd: accepted
  9. fill n11 09012345678 phone before · after
    • Phone: accepted
  10. fill n15 nanoru による自動入力テストです。送信はしません。 message before · after
    • message: accepted
  11. click n17 agree to privacy policy before · after
  12. click n18 go to confirmation before · after
  13. done n1 Confirmation screen reached. The name is shown as HTML entities (Nguyễn Thị Hương), so the diacritics are garbled. The katakana reading displays correctly. before · after
    • NameSei (review row): altered review row shows the name as HTML entities (double-escaped): review row "氏名" shows "Nguyễn Thị Hương" (nanoru measurement)
    • NameMei (review row): altered review row shows the name as HTML entities (double-escaped): review row "氏名" shows "Nguyễn Thị Hương" (nanoru measurement)