"Every prompt is a spell. Every spell, a decree of understanding."
RUNE transforms flat, ambiguous prompts into structured 8-layer prompts validated by Spinoza's philosophical framework — resulting in dramatically better AI responses.
v2.0 "Deus Sive Natura" — now with Spinoza Validator, 40+ grimoire templates, multi-agent swarm mode, and prompt lineage tracking.
Input: "Write a blog post about AI"
RUNE applies 8 layers:
L0 → Expert tech blogger persona
L1 → Domain context, target audience
L2 → Task definition, format spec
L3 → Constraints, ethics check
L4 → CoT thinking strategy
L5 → Tool selection
L6 → Spinoza QA validation
L7 → Output format, style, length
Output: Structured, high-quality prompt ready for any LLM
| Layer | Name | Purpose |
|---|---|---|
| L0 | System Core | Role, persona, behavioral rules |
| L1 | Context Identity | Domain knowledge, target audience |
| L2 | Intent Scope | Task definition, output format |
| L3 | Governance | Constraints, ethical boundaries |
| L4 | Cognitive Engine | Thinking strategy (CoT, ToT) |
| L5 | Capabilities Domain | Tools, integrations, capabilities |
| L6 | QA | Spinoza Validator — 4-pillar quality control |
| L7 | Output Meta | Format, style, length, language |
- Spinoza Validator — Conatus (agency), Ratio (logic), Laetitia (tone), Natura (flow)
- 40+ Grimoire Templates — coding, writing, analysis, creative, AI/ML
- 14 CLI Commands — cast, inscribe, duel, grimoire, validate, forge, swarm, lineage...
- Swarm Mode — multi-agent prompt evolution with tournament selection
- Oracle Engine — self-improving prompts with feedback loops
- Prompt Repetition — based on Google Research (arXiv:2512.14982)
npx clawhub@latest install neurabytelabs/rune-skill- Python 3.11+
- RUNE repo cloned locally
RUNE_API_KEYin~/.secrets
# 1. Clone RUNE repo
git clone https://github.com/neurabytelabs/rune ~/Documents/GitHub/rune
# 2. Install dependencies
cd ~/Documents/GitHub/rune && pip install -r requirements.txt
# 3. Add API key
echo "export RUNE_API_KEY=your_key" >> ~/.secrets# Enhance a prompt (default: inscribe)
echo "Explain quantum computing" | bash main.sh
# As argument
bash main.sh "Write a marketing email for my SaaS"
# Specific commands
bash main.sh cast "Design a REST API for a todo app"
bash main.sh validate "Check this prompt quality"
bash main.sh duel "Compare sorting algorithms"
bash main.sh grimoire
bash main.sh swarm "Evolve the best coding prompt"
# Pipe enhanced prompt into your AI
ENHANCED=$(echo "Analyze this code" | bash main.sh)
echo "$ENHANCED" | llm -m claude-3.5-sonnetMost people give AI a flat prompt and get mediocre output. RUNE applies prompt engineering best practices automatically — in under 2 seconds.
Built on Spinoza's philosophical framework by NeuraByte Labs.
"The more an action follows from the nature of the understanding agent, the more perfect it is." — Baruch Spinoza, Ethics IV
Mustafa Saraç · NeuraByte Labs
- RUNE Framework → github.com/neurabytelabs/rune — Core engine, full docs
- RUNE Playground → github.com/neurabytelabs/rune-playground — Browser demo
MIT