About humanizer
humanizer is a writing skill for cleaning AI-generated text and reviewing it against context-specific guidance. It combines mechanical character cleanup with flags for phrases that need a human rewrite.
Key features
- 01Normalize punctuation and invisible characters
- 02Flag overused wording and clichés
- 03Use guides for README, email, Slack, and commits
- 04Run cleanup on text or files
Use cases
- 01Polish an AI-written README
- 02Check an email for filler phrases
- 03Clean text before sharing in Slack
- 04Review commit messages for style
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/humanizerin Claude Code to confirm it loaded.
Bundled files
References and scripts installed along with SKILL.md.
references/commit-guide.mdemail-guide.mdphrases.txtreadme-guide.mdslack-guide.mdscripts/humanizeDetails
From the project READMEHumanizer
Clean AI-generated text by normalizing characters, flagging overused phrases, and following context-appropriate writing guidelines.
Workflow
- Identify context — determine what you're writing: README, email, Slack message, commit message, or general documentation
- Load the guide — read the appropriate reference file (see table below)
- Write following the guide — apply tone, structure, and length guidelines as you draft
- Run the script — pipe final text through
humanizefor mechanical cleanup and phrase flagging - Review flags — rewrite any flagged phrases naturally; don't just delete them, replace with plain language
Context Guides
| Context | Reference | Key principle |
|---|---|---|
| README | references/readme-guide.md | Direct and factual, show the command first |
| references/email-guide.md | State the ask in the first sentence | |
| Slack | references/slack-guide.md | Terse — 1-2 sentences max |
| Commit | references/commit-guide.md | Imperative mood, explain why not what |
| General | Apply common sense from all guides | Prefer clarity over formality |
CLI Reference
| Command | Description |
|---|---|
echo "text" | humanize | Clean text via stdin |
humanize file.md | Clean text from file |
humanize --report file.md | Verbose output with replacement counts |
humanize --help | Show usage |
Exit codes: 0 = clean, 1 = error, 2 = phrase flags found (text still cleaned)
What the Script Fixes Automatically
- Em dashes → space-hyphen-space
- Smart quotes → straight quotes
- Unicode ellipsis → three periods
- Non-breaking/invisible spaces → regular spaces
- Invisible Unicode watermark characters → removed
- Bullet characters → ASCII hyphens
What Gets Flagged (Not Auto-Fixed)
The script flags phrases that need human judgment to rewrite:
- Red-flag words: delve, tapestry, seamless, robust, comprehensive, etc.
- Hedging filler: "it's important to note", "it's worth noting"
- Cliche openers/closers: "I hope this finds you well", "feel free to reach out"
- Overused transitions: furthermore, moreover, additionally
Full pattern list: references/phrases.txt
Comments
Markdown supported