Authentic AI-Assisted Authorship: Turn one brand interview into 12+ months of content that's 90-95% indistinguishable from the original creator.
Created by Zach Lloyd | Black Sheep Systems | @blacksheepsystems
The AAA System solves the #1 problem in AI content creation: authenticity.
Most AI content fails because:
- Everyone uses the same prompts from the same YouTube videos
- They skip the strategic work (brand discovery, voice analysis)
- They treat AI like a microwave: "Make me content"
- Result: 40-50% authenticity. Your audience can tell.
The AAA System is different:
- 23 modular skills for the complete content lifecycle
- Voice Meta v2.0 captures what you say AND what you'd NEVER say
- Validated at 90-95% authenticity in blind tests
- Complete pipeline from generation to publishing
1 Brand Interview → 100 Topics
1 Topic → 19 Content Pieces
100 × 19 = 1,900 Pieces
= 12+ MONTHS of content from ONE session
# Clone the repository
git clone https://github.com/Zeek2Fit/aaa-authority-acceleration.git
cd aaa-authority-acceleration
# Copy skills and commands to your Claude Code environment
cp -r .claude/skills/* ~/.claude/skills/
cp -r .claude/commands/* ~/.claude/commands/ls ~/.claude/skills/
# Should show 23+ skill directories
ls ~/.claude/commands/
# Should show 23 command files (aaa-workflow.md, brand-discovery.md, etc.)Tip: After installation, type
/aaaand press Tab - Claude Code will autocomplete available AAA commands.
/aaa-workflow
You'll be guided through the complete journey:
- STEP 0: CAPTURE - Gather content samples
- STEP 1: ANALYZE - Voice DNA + Disgust Mapper + Brand Discovery
- STEP 2: ARCHITECT - Authority positioning + Topics
- STEP 3: ACTIVATE - Multi-platform content generation
Natural Language Input → Skill Auto-Activation → Content Generation
↓ ↓ ↓
"New client Sarah" /aaa-workflow suggested Voice DNA extracted
↓
/content-to-airtable
↓
Airtable Queue
↓
curl webhook → n8n → Kit Draft
| Layer | Purpose | Output |
|---|---|---|
| Layer 0: Brand Discovery | Strategic positioning (125 ERISE variables) | Complete Brand Profile |
| Layer 1: Voice Meta | Voice DNA + Disgust Mapper (negative boundaries) | Voice Meta Profile |
| Layer 2: Topic Generation | Combinatorial creativity | 50-100 strategic topics |
/aaa-workflow- Master orchestrator/deep-brand-intake- Content capture/brand-discovery- 39-question strategic interview/voice-dna- 5-layer voice extraction/disgust-mapper- Negative boundary mapping/generate-topics- Combinatorial topic generation/topic-to-matrix- 1 topic → 19 pieces
/authority-engine- STP analysis + competitor discovery/competitor-analysis- Full content analysis/competitor-hooks- Viral hook extraction/competitor-structures- Format analysis/competitor-topics- High-engagement topics/competitor-gaps- White space opportunities
/trust-signal-generator- Orchestrator/trust-case-studies- Client transformation stories/trust-frameworks- Proprietary methodologies/trust-contrarian- Thought leadership takes/trust-metrics- Transparency content/trust-deep-dives- Long-form authority/trust-provenance- Behind-the-scenes
/content-to-airtable- Push to content queue/content-calendar- Bulk scheduling/content-review- Human-in-the-loop approval/session-handoff- Context management
We rejected RAG (Retrieval Augmented Generation) in favor of complete context loading:
| RAG (What Others Do) | AAA System |
|---|---|
| Chunks content into snippets | Loads COMPLETE digital footprint |
| Searches for "relevant" pieces | 500k-800k tokens in one shot |
| Stitches together fragments | Uses Gemini's 1-2M context window |
| 40-50% authenticity | 90-95% authenticity |
RAG can't see that you always open with a question. RAG can't detect your signature three-beat rhythm. We load everything.
The system integrates with Airtable for content management:
Base ID: See SETUP.md for configuration
| Table | Purpose |
|---|---|
| AAA Clients | Client profiles |
| AAA Topics | Topic library |
| AAA Content Pieces | Content queue |
The included n8n workflow automates publishing:
Airtable (Status: Approved) → Webhook → n8n → Kit Draft → Airtable (Published)
Currently Supported:
- Email → Kit (ConvertKit)
- Twitter → X API (bring your key)
- WordPress → WP REST API (bring credentials)
- LinkedIn → LinkedIn API (bring token)
The tools/extraction/ folder contains Python scripts for gathering content:
| Tool | Purpose |
|---|---|
youtube_transcript_extractor.py |
Download YouTube video transcripts |
kit_email_extractor.py |
Extract Kit (ConvertKit) email broadcasts |
podcast_transcript_extractor.py |
Get podcast transcripts from Buzzsprout |
gemini_voice_synthesizer.py |
Analyze 500k+ tokens with Gemini 2.5 |
For Twitter/X: See tools/extraction/README.md for alternatives (API is paid).
Quick start:
pip install youtube-transcript-api requests beautifulsoup4
python tools/extraction/youtube_transcript_extractor.pySee tools/extraction/README.md for complete AI-friendly usage instructions.
- Claude Code - This system runs in Claude Code
- Airtable MCP - For content queue integration
- Kit Account - For email draft creation
- n8n Instance - For automation (optional)
- Python 3.9+ - For extraction scripts
Zach Lloyd - Founder of Black Sheep Systems
Black Sheep Systems helps small business owners implement AI automation that actually works. The AAA System was built to solve the authentic content problem once and for all.
- Claude Code
- Anthropic's Skills architecture
- Gemini 2.5 for voice synthesis
License: MIT - Use freely, attribution appreciated.