Humanize (Strip AI Writing Tells)

What it does

• Fires automatically on requests like "humanize this", "make it less AI-sounding", or when you ask why a draft reads robotic and generic.
• Rewrites against the nine signals detectors actually measure: perplexity, burstiness, hedge density, lexical repetition, structural markers, specificity, POS density, punctuation fingerprint, and the RLHF "helpful assistant" voice.
• Enforces seven hard rules: em-dash budget (max one per 300 words, zero under 300), no semicolons, straight quotes only, a banned-vocabulary list (delve, leverage, robust, furthermore...), no "not just X, it's Y" negation pivots, output-shape discipline (no preamble or changelog), and a measured sentence-length spread.
• Runs a literal pre-output gate where every tell is counted and quoted from the actual draft, plus a Signal I audit pass that catches rhetorical scaffolding (aphorism closers, anaphora, tricolons, chiasmus, either/or binaries).
• Optional writer-profile distillation: give it your past writing and it extracts 5-10 style hypotheses and matches your register instead of defaulting to terse prose.

Who it's for

• Bloggers, newsletter writers, and LinkedIn posters who draft with AI and want their own voice back.
• Marketing and PR teams stripping brochure language ("nestled in the heart of", "vibrant", "stands as a testament to") from launch copy and about pages.
• Anyone cleaning chat artifacts ("Great question!", "I hope this helps") out of documents before publishing.

Examples

1. Paste an AI-drafted launch post and say "humanize this" → you get back prose with em dashes and semicolons removed, a mix of sub-5-word fragments and 25-plus-word sentences, and concrete numbers replacing vague claims.
2. Supply three of your older essays and ask for a voice match → it profiles your sentence variance, verbal tics, and transition style, then rewrites in that register.
3. Feed it text with zero facts, names, or dates → it rewrites without inventing anything and appends a short note telling you which specifics to add so the specificity signal stops flagging it.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/harshaneel/humanize/HEAD/plugins/humanize/skills/humanize/SKILL.md

What I want: (describe your task here)

Install the skill I found on AI Skill Mart.
Copy the plugins/humanize/skills/humanize folder from the GitHub repo harshaneel/humanize into my ~/.claude/skills/harshaneel-humanize/.
When it's done, tell me in one line what this skill can do.