Can Latin1 encoding turn an invisible marker into a control byte?

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. Destination observations run independently on the exact input and cleaned strings in the recorded Node runtime. We make no detector-score or statistical-watermark removal claim.

The encoder changes more than byte length

Our A-marker-B fixture produces UTF-8 hex 41e2808b42. Node Latin1 encoding instead produces 410b42 and decoding those bytes as Latin1 yields A, a vertical-tab character and B. After deletion, both encoders produce 4142. The accented café reference remains unchanged by cleanup but has distinct UTF-8 and Latin1 byte sequences; its Latin1 roundtrip preserves the accented letter. U+0100 is also retained, yet Latin1 emits 00 and returns a NUL character. The plain AB reference matches in both encodings. These paired hex records distinguish selected deletion from lossy conversion of characters the cleaner leaves intact.

This is Node output encoding, not browser decoding

We invoke Buffer.from on JavaScript strings with explicit utf8 and latin1 names, then show a Latin1 roundtrip. The test does not load a file, use TextDecoder, set an HTTP charset or inspect a spreadsheet export. Our UTF-16 and BOM articles examine incoming bytes; this question concerns bytes generated after a cleaned string is sent to an encoder. Node defines its Latin1 behavior, and another environment using a similarly named encoding needs its own test. The controls were deliberately selected for readable hex. They do not characterize arbitrary supplementary characters or establish recovery of information after a lossy conversion.

Keep the encoding contract with the original string

Choose the byte encoding required by the destination rather than expecting invisible-character deletion to make every retained character representable. Preserve the original Unicode string and compare generated bytes or a roundtrip when investigating an export discrepancy. In this matrix, deleting one selected marker repairs only that authored marker case; the U+0100 loss remains. A changed control byte can be caused by encoding rules even when the cleaner reports no selected characters. These measurements provide reproducible byte examples and a clear boundary, not a guarantee of lossless export, a file-format validator or evidence that Claude produced a particular byte sequence.

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