Invisible Text Forensics — AI skill for Claude Code
Agent Skill that finds and removes invisible Unicode in text: zero-width characters, AI watermarks, bidi overrides (Trojan Source), tag-character prompt injection and homoglyphs.
How to install Invisible Text Forensics
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Hiberius/invisible-text-forensics and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Invisible Text Forensics does
Agent Skill that finds and removes invisible Unicode in text: zero-width characters, AI watermarks, bidi overrides (Trojan Source), tag-character prompt injection and homoglyphs. The byte layer every humanizer skill misses. Zero dependencies.
Alternatives in AI
- Invisible Dots — Open-source, self-hosted alternative to OpenAI Dots, Meta Muse, Grok Bot, Manus Cue and Claude Cowork 31.8k ★
- Humanizer — A Claude Code skill that removes signs of AI-generated writing from text, making it sound more natural and hum 11.3k ★
- Pixel Agents — Compressed Reference — VS Code extension with embedded React webview: pixel art office where AI agents (Claude Code terminals) are an 6.5k ★
README
Invisible Text Forensics
**An Agent Skill that reads text at the code point level: finds and removes zero-width characters, bidirectional overrides, tag characters, variation-selector payloads and homoglyphs. Also watermarks documents so you can tell which copy leaked.**
[](LICENSE)   
npx skills add Hiberius/invisible-text-forensics
Works with Claude Code, Claude Desktop, Codex, Cursor, Windsurf, OpenClaw and anything else that reads a `SKILL.md`.
Why this exists
Humanizer and de-AI skills rewrite **style**: em dashes, "delve", tricolons. There are dozens of them and some have tens of thousands of stars. Not one of them reads the **bytes**.
Text carries characters that render as nothing and survive copy, paste, email, PDF extraction and git:
- a zero-width watermark that says which copy of your contract this is
- a run of tag characters spelling
IGNORE PREVIOUS INSTRUCTIONSinside a prompt - a bidirectional override making source code render differently from how it compiles
- a no-break space in a CSV header that quietly breaks a column lookup
This skill is the layer underneath the humanizers. Run a humanizer for the prose, run this for everything the prose is made of.
Try it in ten seconds
python3 scripts/itf.py scan document.md
document.md
CRITICAL U+202E line 12 col 18 Right-to-Left Override (RLO)
reorders rendering; Trojan Source (CVE-2021-42574) in source code
HIGH U+200B line 3 col 41 Zero Width Space (ZWSP)
invisible; carries watermarks and hidden payloads
MEDIUM U+0443 line 3 col 58 Homoglyph of 'y'
non-Latin letter that renders like ASCII
What i
Related Skills
Unicode Smuggling Guard
Detects invisible Unicode hidden in code, docs and AI agent files (CLAUDE.md, SKILL.md, MCP tool descriptions)
Clean Commit Skill
Keep AI noise out of your git history. A cross-platform agent skill that blocks Co-Authored-By bots, generator
Claude Sidebar Modifier
Disable or adjust the width of the sidebar functionality on Claude.ai
Human Writing
Skill bắt AI viết văn giống người, không dùng em dash và các ký tự Unicode lạ mà AI hay sinh ra. Thay toàn bộ
Machine Forensics Loop
Troubleshooting memory for AI coding agents: a persistent, per-machine knowledge base of what broke, what fixe
Coding Agent Forensics
Offline single-file forensic viewer for AI coding-agent sessions — Claude Code, Codex, Cursor, Gemini, Copilot
Related Agents
Watermark Text Cleaner
Dispatch to scrub invisible / hidden watermark characters from a TEXT file: zero-width spaces, bidi controls,
Glyph Reader
Stage 0 worker — reads rendered page regions and names the Unicode character each unresolved glyph of a broken
Character Developer
Builds psychologically complex characters with authentic voices, compelling arcs, and realistic relationships.