Skip to content

smykla-skalski/humanize

v2.5.2MIT

Make text sound natural by removing AI writing patterns. For commit messages, PR descriptions, review comments, docs, changelogs, and any text that sounds robotic or AI-generated.

GitHubOpen Website

humanize

Identifies and removes signs of AI-generated writing from text, then rewrites it using proven composition principles.

The plugin is a portable Agent Plugin with one Agent Skill, so the same package works in Claude Code, Codex, Copilot CLI and opencode.

Two complementary sources:

  • Detection: Wikipedia's Signs of AI writing guide (WikiProject AI Cleanup) - 24 patterns across content, language, style, communication, and filler categories
  • Composition: William Strunk Jr.'s The Elements of Style (1918) - active voice, concrete language, omitting needless words, sentence variety, emphasis placement

Installation

Claude Code (Copilot CLI is the same with copilot in place of claude):

claude plugin marketplace add smykla-skalski/sai
claude plugin install humanize@sai

Codex:

codex plugin marketplace add smykla-skalski/sai
codex plugin add humanize@sai

opencode, or any agent that reads Agent Skills, loads skills/humanize/ directly:

ln -s /path/to/sai/plugins/humanize/skills/humanize ~/.config/opencode/skills/humanize

Local checkout: claude --plugin-dir /path/to/sai/plugins/humanize/

Usage

In Claude Code and Copilot CLI use /humanize, in Codex $humanize, or ask in plain words ("humanize this PR description"). Where the agent has no subagent tool, the pattern scan runs inline instead of in a subagent; the output is the same.

/humanize path/to/file.md
/humanize path/to/file.md --score-only
/humanize path/to/file.md --dry-run
FlagPurpose
(positional)File path to humanize
--score-onlyReport detected patterns without rewriting
--dry-runOutput rewritten text to chat instead of saving

What it detects

24 AI writing patterns in five categories:

  1. Content (1-6): significance inflation, notability claims, superficial -ing analyses, promotional language, vague attributions, formulaic challenges
  2. Language (7-12): AI vocabulary, copula avoidance, negative parallelisms, rule-of-three, synonym cycling, false ranges
  3. Style (13-18): em dash overuse, boldface overuse, inline-header lists, title case, emoji decoration, curly quotes
  4. Communication (19-21): chatbot artifacts, knowledge-cutoff disclaimers, sycophantic tone
  5. Filler (22-24): filler phrases, excessive hedging, generic positive conclusions

License

MIT