Local name-similarity screen
The supplied post describes cloned tokens and split attention. This screen finds simple name matches; it does not discover live bots, prove fraud, or fetch liquidity.
The source uppercases names and removes everything outside letters A to Z and digits. Lead NOXA and candidates NOXA2, N0XA, NOX and ORBIT yield two flags, NOXA2 and NOX, because either normalized name can start with the other. Bars compare four candidates with two flags at one hundred twenty pixels per name. The digit zero in N0XA is not converted to letter O. These simple string flags are review prompts, not proof of fraud, token ownership or a real cloned deployment. If the lead contains only punctuation, its normalized form is an empty string. Every candidate starts with the empty string, so three entered nonempty candidates produce three flags. Bars share one hundred fifty pixels per name and are equal at four hundred fifty pixels. Similarly, a punctuation only candidate normalizes to empty and passes the reverse prefix test against a nonempty lead. This source edge case is preserved and disclosed; there is no semantic similarity model or live market evidence. Save stores six properties: lead token, candidate list, flags, screening status, source boundary and note. Export uses that saved screen when present, but replaces its note with the currently entered note. Bars count six properties against one property refreshed at export, at eighty pixels per property. Editing names without saving does not change the stored screen. Reload does not restore the visible inputs even though the saved record remains available to export. Reset removes that record and reloads. No blockchain or liquidity service is contacted.