From the team that runs a company with AI agents
PromptStack is a collection of battle-tested CLAUDE.md templates, agent persona definitions, and workflow skills extracted from a real AI-run company. 14 expert agents. 35 specialized skills. Zero fluff.
What is PromptStack
Most developers copy-paste generic prompts and wonder why their AI writes generic code. PromptStack gives you the same configuration files that run a real autonomous AI company.
Production-grade project instructions that turn your AI coding agent from a helpful assistant into an opinionated expert. Decision principles, safety guardrails, collaboration workflows.
14 agents modeled on real-world experts: Bezos for strategy, DHH for fullstack, Munger for critical thinking. Each with decision frameworks, communication styles, and domain knowledge.
35 specialized skills from deep research to security audits, from pricing strategy to SEO content. Plug-and-play capabilities that make your AI agent genuinely useful.
Pre-defined agent pipelines for common tasks: new product evaluation, feature development, product launch, pricing. The right agents in the right order.
Non-negotiable rules that prevent your AI from deleting repos, leaking credentials, force-pushing to main, or doing anything destructive. Battle-tested boundaries.
Plain Markdown files. No frameworks, no dependencies, no build step. Drop them into your project and your AI agent levels up immediately.
How it works
Add a CLAUDE.md to your project root. It tells your AI agent what the project is, how to make decisions, what tools to use, and what never to do. Your agent reads it on every conversation.
Each agent is a Markdown file with a specific persona, decision framework, and output format. Need a code review? Call the DHH agent. Pricing decision? Call the Patrick Campbell agent.
Skills are specialized capabilities: deep research, security audits, competitive analysis. Drop them in, and any agent can use them when the task demands it.
Free Preview
Each agent is modeled on a real-world expert's thinking style, decision frameworks, and communication patterns.
Day 1 mentality. Customer obsession. Flywheel thinking. PR/FAQ method for evaluating ideas. Two-way door decisions for speed. Regret Minimization Framework for big calls.
Convention over configuration. Majestic monolith. One-person framework philosophy. Clear over clever code. Ships fast, deletes more code than writes. Anti-overengineering.
Inversion thinking. Pre-mortem analysis. Psychology of misjudgment checklist. Circle of competence. Finds the fatal flaws before you ship them. The one agent that says "no."
Who this is for
You are a one-person company. Your AI agent should be a full team: CEO, CTO, designer, marketer, QA. PromptStack makes that real.
Stop spending hours crafting prompts. Use battle-tested agent configs that cover strategy, code, design, marketing, and finance.
Standardize how your team uses AI coding agents. Shared agent definitions mean consistent quality across the entire team.
Learn how to structure CLAUDE.md files, agent personas, and skill definitions by studying real production examples.
Start with the free CLAUDE.md guide. Or go straight to the full kit with all 14 agents and 35 skills.