Create a reusable Claude Code and Antigravity skill that scores AI-written copy, diagnoses what feels synthetic, rewrites it in a real voice, and improves after each run.
A review-and-rewrite skill that catches AI-sounding copy, applies a scoring rubric, rewrites drafts in a chosen voice, and updates itself from feedback.
Quick view
Check the audience, core tools, and access before you start the detailed steps.
Best for
Core tools
5 tools used across the workflow.
Access
Why this works
Prompting AI to sound human often fails because the instruction is too vague. A humanizer skill works better when it scores the draft first, names the patterns that feel artificial, and then rewrites against a specific voice baseline. The feedback loop matters because every awkward draft teaches the skill a new anti-pattern to catch next time.
Expected results
Review speed
2-4 hours saved per batch
The skill handles first-pass scoring, diagnosis, and rewrite suggestions so editors focus on judgment and final polish.
Copy quality
Scores plus diagnosis
The workflow explains why a draft feels artificial instead of only producing a different-sounding rewrite.
Voice consistency
Source-sample grounded rewrites
Real writing samples and anti-examples make the output more specific than generic tone instructions.
Skill improvement
Rule updates after real feedback
The skill compounds because repeated reviewer corrections become explicit instructions.
Preview the open steps, then unlock the remaining implementation details and prompts.
Step-by-step workflow
The first 2 steps are open. Pro unlocks the remaining steps, copy-paste prompts, pro tips, tool-by-tool setup guidance, and implementation details.
45-60 min
45-60 min
Create a Google Docs folder with 5-10 writing samples that actually sound like the person or brand you want to emulate. Include LinkedIn posts, emails, landing page sections, internal notes, or transcripts if they capture natural phrasing. Add 5 examples of AI-written copy that feels wrong and annotate why: too polished, too symmetrical, too generic, too many transitions, no domain specificity, or no real opinion. The skill needs both positive and negative examples to learn what to preserve and what to avoid.
A voice source folder with real writing samples, weak AI examples, and annotation notes.
The best voice samples are often less polished than you expect. Clean corporate copy may be approved, but it rarely teaches rhythm, personality, or judgment.
45 min
45 min
Create a scoring rubric with four 1-10 categories: AI likeness, authenticity, reader value, and domain credibility. For each category, define what scores 1, 5, and 10 look like, and include common failure patterns. Store the rubric in Google Docs and add a Google Sheets tracker with columns for draft type, score, diagnosis, rewrite decision, reviewer feedback, and rule added. QA check: a reviewer should be able to score the same draft within one point of the skill.
A copy review rubric and score tracker that the skill can apply consistently.
Do not make the score only about sounding human. A casual but empty rewrite can score low on reader value or domain credibility even if it sounds less robotic.
Pro workflow preview
Previewing 2 of 8 steps
Get the remaining 6 steps, copy-paste prompts, pro tips, tool-by-tool setup guidance, and ongoing workflow and prompt releases.
$9/month
Related workflows
Continue with workflows that share a similar GTM motion, category, or tool stack.
Build a brand voice QA system with voice rules, banned phrase checks, review prompts, scoring, and an approval workflow for AI-generated GTM content.
A launch asset factory that turns one approved narrative into coordinated draft launch assets routed through PMM and stakeholder review.
A positioning refresh memo with customer language, message pillars, objection themes, and copy recommendations.
