Prompt Forge
The system behind prompts that make AI think differently
Build prompts that rewire how AI thinks. 6-phase system with 8 meta-cognition layers and 5-expert adversarial review.
What's Inside
Stop feeding AI instructions. Start engineering how it thinks.
- Install one skill and get a 6-phase system that builds, stress-tests, and delivers prompts you couldn't write by hand
- 8 meta-cognition layers force AI to decompose problems, argue with itself, and map failure modes before output
- Every prompt reviewed by a 5-expert adversarial panel and must hit 95/100 before delivery
What's Included
Everything you need to get results fast.
6-Phase Workflow
Extraction, Architecture Selection, Draft Generation, Expert Panel Review, Optimization Loop, and Delivery. You answer questions. The skill handles the rest.
8 Meta-Cognition Layers
Inversion, First Principles, Competing Perspectives, Failure Mapping, Self-Interrogation, Constraint-First, Recursive Depth, and Cross-Domain Transfer. Stackable based on task complexity.
5-Expert Adversarial Panel
Every prompt evaluated across technical precision, Claude optimization, creative exploitation, practical robustness, and cognitive augmentation. Must score 95+ to ship.
Technique Library (15+ Methods)
Chain of Draft, Tree of Thought, XML Architecture, Contrastive Examples, Emotional Prompting, and more. Each with when-to-use guidance and ready XML templates.
Anti-Pattern Detection
20+ failure patterns built into the quality check. Wall of Text, Polite Suggestion, Default Delegator, Generic Expert. Documented with fixes so your prompts avoid every trap.
3 Archetype Blueprints
Complete structural templates for Agent System Prompts, Task Prompts, and Mega-Prompts. Full XML scaffolding ready to customize.
Frequently Asked Questions
What exactly is Prompt Forge?
A Claude Code skill with 6 reference files. Install it in your .claude/skills/ directory and it handles prompt engineering for you. 6-phase workflow: interviews you, picks architecture, drafts, stress-tests through 5 experts, iterates to 95+ score, delivers. Works for agents, tasks, and mega-prompts.
Do I need to know prompt engineering to use this?
No. The skill does the engineering. You answer questions about what you need and the skill handles technique selection, meta-cognition layers, architecture choices, and quality testing. You get the result without understanding the process.
Will this work with ChatGPT / GPT / Gemini?
The skill runs in Claude Code, but the prompts it produces work with any AI. The techniques are effective across all modern LLMs. Claude gets a slight edge from Claude-native optimization, but the output is portable.
What's the difference between this and a prompt library?
A prompt library gives you fish. Prompt Forge teaches the AI to fish. Libraries go stale as models update. Prompt Forge generates fresh prompts for any situation because it understands the principles, not just the patterns.
How do I install it?
Download the ZIP, extract the prompt-forge folder into your .claude/skills/ directory. Done. Takes about 2 minutes. The included guide walks through it step by step.
What if it doesn't work for me?
60-day money-back guarantee. Use the skill, build prompts, test them. If the quality doesn't impress you, email for a full refund. No forms, no hassle.
Machina
“”