8 Claude skills that make AI content sound like it came from a human with a point of view.
Built by Branded Mayhem Collective. Works in Claude.ai (web), Claude Code, Claude desktop, Cursor, and Codex CLI — anywhere SKILL.md is read.
AI-generated content has a voice. Unfortunately, it's the same voice for everyone: smooth, hedging, over-complete, structurally uniform, and emotionally flat. Detection tools catch it. Audiences feel it. Your content sounds like everyone else's content.
Eight skills that attack the problem from both sides:
- Create content worth reading — frameworks for LinkedIn, narratives, ideation, and production workflows that produce distinctive output
- Make AI content undetectable — a 16-pattern detection analysis system that breaks the statistical signatures AI detectors rely on, rewritten in YOUR voice
- Test before you ship — simulated audience focus groups that pressure-test your creative against realistic archetypes before it goes live
The secret weapon is the Voice Profiler — an interactive skill that interviews you about your writing style, analyzes your real writing samples, and builds a personal voice profile. Every piece of humanized content then sounds like YOU wrote it, not "a human."
| Skill | What It Does | Use When |
|---|---|---|
| linkedin-authority | LinkedIn posts using high-performing viral patterns: 5 core themes, 4 hook types, sentence rhythm rules | Writing LinkedIn content that gets engagement |
| story-spine | 6-beat narrative framework + emotional heat map for origin stories, case studies, founder essays | Any narrative that needs to hold attention |
| creative-thinking-ai | 10 classic creativity methods (SCAMPER, Six Hats, Mind Mapping, etc.) as structured AI workflows | Brainstorming, ideation sprints, concept stress-testing |
| mayhem-method-ai-use | Standardized AI-assisted creative production: prompt chaining, few-shot steering, structured output | Campaign briefs, social posts, any content deliverable |
| humanize | 16-pattern AI detection analysis + voice-profile-driven rewrite | Any AI-generated text that needs to sound like you |
| humanize-ig | Instagram-specific voice interrupter: lowercase, dense blocks, imperfect grammar, phone-typed feel | Instagram captions that don't read like AI |
| voice-profiler | Interactive interview that builds your personal voice profile from real writing samples | Run once before first /humanize use |
| ai-focus-group | Simulated panel of 5-7 audience archetypes generates structured reactions to any creative asset | Testing headlines, taglines, copy, concepts, or campaigns before launch |
This repo works on two surfaces — pick yours. The skills are surface-aware: voice-profile paths resolve correctly in both Claude Code and Claude.ai web automatically.
Claude Code (plugin install — primary path):
/plugin marketplace add Branded-Mayhem-Collective-LLC/content-creative-lab
/plugin install content-creative-lab@bmc-content-creative-lab
Installing the whole stack? Add Branded-Mayhem-Collective-LLC/contraband-marketplace and install with @contraband instead. Skills become available under the namespace /content-creative-lab:<skill>. Verify with /plugin list.
Claude.ai web / Claude Desktop:
Paste the repo URL into Claude.ai web and ask:
"Install the skills from https://github.com/Branded-Mayhem-Collective-LLC/content-creative-lab"
Claude fetches, registers, and asks permission. Alternatively download as .zip (GitHub Code → Download ZIP), unzip locally, upload skill folders via Settings → Customize → Skills. Note: voice profiles don't persist across Claude.ai web sessions — /humanize-kit:voice-profiler writes to /mnt/user-data/outputs/; download it and re-upload it at the start of each session where you want /humanize to use it.
For Cursor or Codex CLI, follow each tool's plugin/skill loading docs and point at this repo. SKILL.md is the open standard — same files, different loader.
/content-creative-lab:voice-profiler
Claude will interview you about your writing style and ask for 3-5 real writing samples. It generates ~/.claude/voice-profile.md — the file that makes /humanize sound like YOU instead of generic-human.
Takes about 5-10 minutes. Do it once. Update it whenever your voice evolves.
Write a LinkedIn post:
/linkedin-authority Write about why most brand strategies fail
Humanize any AI-generated text:
/humanize [paste text here]
Humanize for Instagram:
/humanize-ig [paste text here]
Run an ideation sprint:
/creative-thinking-ai Generate 10 content angles for a SaaS launch
Write a narrative:
/story-spine Write an origin story about starting my agency
Produce a campaign brief:
/mayhem-method-ai-use Create a 3-post LinkedIn campaign for [topic]
Test creative against a simulated audience:
/ai-focus-group Test this headline: "Your brand isn't boring. Your strategy is." Target: marketing directors at mid-size companies
The skills chain naturally:
creative-thinking-ai → Generate content angles
↓
linkedin-authority OR → Write the piece (LinkedIn, narrative, brief)
story-spine OR
mayhem-method-ai-use
↓
humanize OR → Break AI patterns, apply your voice
humanize-ig
↓
ai-focus-group → Test creative against simulated audience panel
- Ideate with
creative-thinking-ai— get 10 angles, stress-test with a second method, shortlist to 3 - Write with the format-specific skill —
linkedin-authorityfor LinkedIn,story-spinefor narratives,mayhem-method-ai-usefor structured deliverables - Humanize with
/humanizeor/humanize-ig— breaks AI detection patterns and rewrites in your voice profile - Test with
/ai-focus-group— simulate audience reactions before publishing to catch friction points, comprehension gaps, and resonance issues
The /voice-profiler skill runs an interactive interview:
Phase 1: Collect Samples
- You paste 3-5 pieces of real writing you're proud of
- Claude analyzes: sentence rhythm, opening/closing patterns, vocabulary fingerprint, grammar personality, emotional range, persuasion style
Phase 2: Guided Questions
- Claude asks targeted questions to fill gaps the samples didn't reveal
- Words you hate, phrases that are distinctly yours, sign-off style
Phase 3: Synthesis
- Claude plays back what it found: "Here's what I see in your voice..."
- You correct anything that's off
Output: A structured voice-profile.md file saved to ~/.claude/voice-profile.md that includes:
- Core identity description
- Sentence rhythm patterns
- Opening/closing patterns
- Vocabulary preferences and banned words
- Persuasion style
- Grammar personality
- Real writing samples (for reference)
- Platform-specific overrides (LinkedIn, email, Instagram)
Both /humanize and /humanize-ig use the same 16-pattern detection analysis:
Tier 1 — Statistical Patterns (4 patterns) Low perplexity, low burstiness, limited semantic diversity, smooth token probability
Tier 2 — Deep Learning Patterns (8 patterns) Structural uniformity, hedging, generic abstraction, mechanical transitions, emotional flattening, over-completeness, list-heavy structure, preamble/summary
Tier 3 — Document-Level Signals (4 patterns) Consistent register, balanced paragraphs, perfect grammar, absence of voice, symmetrical structure
Each pattern scored 0-3. Total score determines rewrite mode:
- 0-15% → Surgical (touch only flagged sentences)
- 16-50% → Moderate (rewrite flagged sections)
- 51%+ → Full Rewrite (rebuild from scratch, keep core message)
The rewrite applies your voice profile systematically — vocabulary, rhythm, persuasion style, emotional range — so the output doesn't just avoid detection, it sounds like you actually wrote it.
Do I need the voice profile to use humanize? No, but the output will be generic-human instead of YOUR-human. The voice profile is what makes this package worth buying. Spend the 10 minutes.
Can I have multiple voice profiles?
Yes. Save them as different files (voice-profile-linkedin.md, voice-profile-email.md) and tell Claude which to load. The default location is ~/.claude/voice-profile.md.
Does this guarantee passing AI detection? No. AI detection is probabilistic and evolving. But breaking 16 statistical patterns while writing in an authentic human voice profile makes detection significantly harder. The goal is authenticity, not evasion.
Can I use these skills with other Claude skills?
Yes. They're standalone .md files that chain with anything. The Brandprint Engine, for example, can feed its output through /humanize for client-facing copy.
How do I update my voice profile?
Run /voice-profiler again. It'll detect your existing profile and offer to update or start fresh.
content-creative-lab/
├── README.md
└── skills/
├── linkedin-authority/SKILL.md # LinkedIn post framework
├── story-spine/SKILL.md # Narrative framework
├── creative-thinking-ai/SKILL.md # 10 ideation methods
├── mayhem-method-ai-use/SKILL.md # AI production workflow
├── humanize/SKILL.md # AI pattern interrupter
├── humanize-ig/SKILL.md # Instagram voice interrupter
├── voice-profiler/SKILL.md # Voice profile builder
└── ai-focus-group/SKILL.md # Simulated audience testing
MIT. Free and open — use it, fork it, ship client work with it. Outputs you produce with these skills are yours. See LICENSE.
Questions? Email support@brandedmayhem.com