Humanizer Skill for Claude

Run detect.py on a .txt or .md draft to find AI-writing patterns and get an AI probability score.

Use transform.py to humanize the draft, then run compare.py to check the before/after improvement.

• You need a humanizer skill for Claude, ChatGPT, GPT-4, or other LLM drafts before publishing or submission.
• You work on essays, editorial drafts, documentation, or humanities skills assignments where AI-sounding phrasing hurts credibility.
• You want a systematic, pattern-based audit that identifies specific AI markers before manual editing.
• You need batch processing for dozens of .txt or .md files and want before/after AI probability scores.

• You need a real-time browser or in-editor rewrite tool; this is a CLI-based, file-in / file-out workflow.
• The text is highly technical, legal, medical, or domain-specific enough that automated phrase replacement could change meaning; manual review is still required.

detect.py checks for AI-writing signals across critical, high-signal, medium-signal, and style categories, including citation bugs, chatbot artifacts, overused AI vocabulary, filler phrases, vague attribution, curly quotes, and em dash overuse. Each issue includes an overall AI probability rating.

Use this humanizer skill Claude workflow after drafting with Claude, ChatGPT, GPT-4, or another LLM. It keeps scan, rewrite, and comparison steps file-based and repeatable for individual writers or editorial teams.

transform.py applies safe fixes for the most reliable patterns: stripping markdown artifacts, removing chatbot opener or closer sentences, replacing copula-avoidance constructions with plain equivalents, simplifying filler phrases, and reducing punctuation signatures.

compare.py runs detection on both the original and transformed versions side by side, so you can verify score improvement before saving the cleaned result.

Detection terms and replacement rules live in scripts/patterns.json, and the scripts support stdin plus shell loops. You can scan or rewrite folders of .txt and .md files while tuning the pattern library for your writing standards.

Run detect.py on essays, reports, and humanities skills assignments to surface AI markers before submission. Then use transform.py for mechanical fixes and review flagged vocabulary by hand.

Editors can batch-scan AI-drafted articles and send Medium-or-higher files through transform.py before human review, reducing time spent on repetitive cleanup.

LLM drafts often contain promotional filler and structural clichés. This humanizer skill removes the obvious patterns so a human writer can refine the message faster.

AI-generated docs often include markdown residue, chatbot sign-offs, and punctuation signatures. transform.py cleans those patterns in bulk across .md files.