NOCLAU GUIDE
Claude Text Formatting Guide
Review copied assistant text, inspect characters, and preserve document meaning before publishing.
Copied assistant text can require several kinds of review before publication: factual editing, document formatting, and character cleanup. These are separate jobs. noclau helps with selected character-level issues in text copied from Claude or any other source. It does not validate facts, reconstruct a rich-text document, identify authorship, or remove a verified proprietary watermark. This workflow helps you preserve the substance of a draft while investigating the specific formatting problem that brought you to the cleaner.
Scroll the table sideways to see every column.
| Symptom | First action | Outside the cleaner |
|---|---|---|
| Exact identifier mismatch | Inspect character sequence | Destination validation |
| Missing heading style | Check paste format | Editor heading controls |
| Excessive dashes | Review sentence structure | Substantive rewriting |
| Uncertain factual claim | Verify source | Fact checking and attribution |
Name the actual problem first
Start with a concrete description such as a heading that wraps unexpectedly, a product code that does not match, or punctuation that differs from your house style. Avoid beginning with the assumption that every copied passage contains a hidden marker. A page, editor, clipboard conversion, or downstream application may affect the text. The source name alone does not reveal where a particular character entered the copying path.
Keep the original response and a copy of the destination text. If the source looks correct but the destination does not, compare a small passage from each. Reproduce the problem with a neutral example before sharing it with support. This preserves the evidence you need while avoiding disclosure of a private conversation, a client brief, or another person's information.
Separate characters from rich-text structure
A plain-text textarea receives characters, not the complete visual structure of an editor. Headings may arrive as ordinary lines, lists may retain visible bullet symbols, and links may lose their hidden destinations. A cleaner cannot infer the intended heading level or rebuild a table from appearance alone. Check the publishing system's paste options and format controls when structure is the problem.
Markdown syntax is also text. If a copied passage contains heading hashes, asterisks, or fenced code, noclau leaves those characters alone. It does not render Markdown or convert it to document formatting. Decide whether your destination expects Markdown, rich text, or plain text before running any character rule. A mismatch between those formats is not solved by deleting invisible characters.
Inspect before you alter the passage
Use the checker if you want a report while preserving the input. It identifies selected invisible characters, unusual spaces, dash variants, and controls. The Analyze text action produces an unchanged result, making it a useful first step when you are unsure which edits are justified. Record the code points and the surrounding words rather than describing the report simply as positive or negative.
For example, a finding of U+200B inside an invented identifier may explain an exact-match failure. A finding of U+200D inside an emoji may be expected. Both are detected characters, but they require different decisions. No finding is a score for authorship. An empty report also does not prove that the text is factually correct, stylistically polished, or free of unsupported character issues.
Apply conservative cleanup to a copy
When you have identified an accidental separator, open the cleaner and start with conservative mode. It removes zero-width spaces, BOM characters, and soft hyphens, while preserving joiners, special spaces, and dashes. Check whether every removed character is appropriate for the passage. Conservative mode is a limited rule set, not a language-aware judgment about which marks matter.
The default example includes a hidden separator after Sample and an em dash later in the sentence. Running conservative cleanup removes the hidden separator and keeps the dash. That example demonstrates one explicit transformation. It does not claim that Claude produces that sequence, or that removing it changes how another service classifies the text. Use the example to learn the controls before applying them to your own copy.
Handle spaces and punctuation as separate decisions
If the actual problem is an unusual space, enable Normalize spaces only after reviewing why that space is present. The rule replaces supported special spaces with ordinary spaces. It does not collapse repeated ordinary spaces, trim the document, or rebuild paragraphs. A no-break space around a number or name may have been intentional in the destination layout.
Dash handling is independent. Keep dashes for a neutral character cleanup. Use the dedicated dash workflow when the publication requires a punctuation change, and read every affected sentence afterward. Automatically removing a dash between two words can join them together. If the sentence needs a comma, colon, or new sentence, make that editorial change explicitly in your writing tool.
Check code, links, and quoted material separately
Code blocks need their normal review and tests. The cleaner preserves tabs and ordinary line breaks, but that does not make it a parser or a code validator. A hidden mark in a string may be intentional test data. Retain the original and use an editor that exposes characters when the surrounding syntax matters. Do not run an unexplained bulk replacement across a codebase.
For links, inspect the actual destination in the publishing editor, not just the visible label. For quotations, verify the quoted wording and preserve required attribution. Character cleanup does not supply permissions to reuse material or establish that a citation supports a claim. These checks remain necessary even when every proposed character edit is harmless.
Accept the result in its final context
Copy the result into the destination and repeat the operation that originally failed. Check searching or matching if that was the issue, then read the full paragraph for word boundaries and punctuation. View the published preview to confirm headings, lists, and line wrapping. If a problem appears only after saving, inspect that stage of the workflow rather than assuming the cleaner missed a character.
For a reusable process, keep a short checklist with the source format, chosen rules, and destination test. Preserve an original until the review is complete. The result should be a passage whose content you have checked and whose edits you can explain. That is a stronger publishing workflow than relying on a tool name or a zero-findings counter.