Skip to content

README.md

Latest commit

 

History

History
109 lines (81 loc) · 5.06 KB

README.md

File metadata and controls

109 lines (81 loc) · 5.06 KB
Raw

Invisibles

Paste text, see every hidden character in it with its name and code point, then click once to strip them. Zero-width spaces, non-breaking spaces that look like real ones, direction controls, and the odd characters that ride along when text moves between editors, documents, chat apps, and websites.

Ko-fi License Tests

Web app: cig13zs.github.io/invisibles. Or install the browser extension so it's one click from any tab, offline.

A single invisible character can break CSV and JSON imports, stop strings from matching, or throw off code search. Most editors do not draw these characters, so the problem often appears somewhere downstream.

Paste tools are not always clear about how they handle input. This repository is deliberately auditable: it has no server, analytics, or remote processing. The web app is one HTML page and one small script, and it works offline.

There are two builds sharing one engine (core.js): the web app and a side-panel extension. The extension requests only Chrome's sidePanel permission and no host access, so it cannot read the pages you visit.

What it looks for

Category Examples Why it matters
Zero-width ZWSP U+200B, ZWJ U+200D, word joiner, BOM U+FEFF Invisible; break search, passwords, code
Look-alike spaces NBSP U+00A0, narrow NBSP U+202F, ideographic space Look like a space, aren't one
Bidi controls RLO U+202E, isolates U+2066..U+2069 Can reorder or hide text (Trojan Source)
Soft hyphen U+00AD Invisible unless the line wraps
Tag characters U+E0000..U+E007F Invisible; smuggle hidden text or prompts
Variation selectors U+FE00..U+FE0F Invisible; can carry hidden data
Control chars C0/C1, line and paragraph separators Non-printing; break parsers

Look-alike spaces get replaced with a normal space so words don't run together. Everything else is removed. There's an optional toggle that also swaps fancy punctuation (curly quotes, em dashes, ellipsis) for plain ASCII when a target system does not handle typographic characters reliably.

Browser extension

Opens directly as a Chrome Side Panel. Click the toolbar icon, and it docks cleanly on the side of your browser without floating in front of your page content. Paste, clean, copy. The panel states before input that pasted text stays in the panel and is never stored or uploaded. It declares no remote permissions and no host access, so it cannot read any page you visit, and it works completely offline.

Load it unpacked in Chrome, Edge, Brave or Opera:

  1. Download the latest zip from Releases and unzip it.
  2. Open chrome://extensions and turn on Developer mode.
  3. Click Load unpacked and pick the extension folder.
  4. Click the Invisibles icon to open the sidebar and paste your text.

Use it as a library

core.js has no dependencies and runs in the browser (window.Invisibles) or in Node (require).

const I = require('./core.js');

I.scan('he​llo').total;      // 1
I.scan('he​llo').findings;   // [{index:2, cp:8203, hex:'U+200B', name:'Zero-Width Space', category:'zero-width'}]
I.clean('he​llo');           // 'hello'
I.clean('a b');              // 'a b'  (NBSP becomes a normal space)
I.normalizePunctuation('“a”—b');  // '"a"--b'  (opt-in)

scan() and clean() never throw, even on non-string input.

Files

core.js        classify() + scan() + clean() + normalizePunctuation()
core.test.js   node core.test.js
index.html     the web app. No framework, no build step
extension/     MV3 side panel, sidePanel permission only
scripts/       deterministic Chrome Web Store package builder
store-assets/  listing copy, reproducible screenshots, and submission notes
tools/         deterministic icon generator

node core.test.js
node site.test.js
python tests/package.test.py
node store-assets.test.js

Limits

It flags characters, not intent. A non-breaking space is sometimes deliberate, so the reveal view shows you exactly what and where before you strip anything.

The punctuation normalize is opt-in and lossy, since em dashes become --. It's off by default.

It can't tell you whether text was written by an AI. It only shows the characters that are actually there.

More tools

  • Toolfold, selected local developer utilities in one PWA
  • Plusminus, measurement uncertainty budgets from instrument specifications
  • Rinse, inspect and remove location metadata from a photo

MIT licensed. Ko-fi if it saved you a headache.