Does the cleaner remove every Unicode format character?

Measured inputs and outputs

These are locally constructed test strings, not evidence that Claude inserts these characters. We executed the saved homepage script snapshot with a minimal DOM harness and compared exact strings. The invisible-checkbox state and dash mode appear in each row. The observations below test this page's specific question. File API and event-lifecycle tests include an additional stage described in the analysis. We make no detector-score or statistical-watermark removal claim.

Category membership and selection are separate

The observer uses Unicode property escapes to list format characters and nonspacing marks in each input. U+200B appears in the format list and is removed. U+200D and U+200E also appear in that list, but remain after cleanup. U+034F appears in the nonspacing-mark list and is removed. Thus the homepage selection is neither all Cf characters nor only Cf characters. The table reports actual property membership from the recorded runtime alongside the independent cleaner output. This question concerns coverage of a character category, not how removing a marker changes the meaning of a user-supplied regex pattern.

A category does not decide whether editing is correct

Unicode general categories are descriptive properties, not a universal deletion policy. A retained joiner or direction mark may serve a valid text function. Conversely, a selected mark can be present even if a Cf-only scan would not list it. The five-entry homepage map names its chosen characters directly. We do not infer a complete set of invisible glyphs from either category: rendering depends on sequences, fonts and context. The tests contain only these four synthetic examples and do not implement the full Unicode default-ignorable property or a document-format validator.

Inspect named characters in their sequence

Choose an inspector that identifies the specific code points relevant to your destination and retain the original. If you use a category-based search, label what that search covers and verify whether the character you care about belongs to it. Avoid replacing the homepage map with a broad category deletion without a text policy and sequence review. Compare the resulting value in the actual application after a targeted edit. The saved property lists and removal results explain these examples; neither a no-findings status nor an empty Cf list proves absence of every invisible character or any statistical text watermark.

Reproduce this test

Save reproduce.cjs and tested-app.js in the same folder. Run the command below with Node.js. The harness prints its runtime, script SHA-256 and every measured row. Compare those rows with the original record. Using a newer script or runtime creates a new experiment; retain the version information with your rerun.

node reproduce.cjs