🤖 Vibe Coders and AI Builders
The people building with AI in public, from Andrej Karpathy and Simon Willison to Pieter Levels and Marc Lou, plus the prompts, skills, and agents they are known for. Open one, add your problem, and reason like them.
Nat Eliason
Writer turned vibe coder who built Covici and teaches non-programmers to ship software; makes an AI reason with his maintenance-first, taste-first philosophy from nateliason.com/blog/the-past-and-future-of-vibe-coding.
Prompt CarrotNat Eliason
Prompt CarrotWriter turned vibe coder who built Covici and teaches non-programmers to ship software; makes an AI reason with his maintenance-first, taste-first philosophy from nateliason.com/blog/the-past-and-future-of-vibe-coding.
CLAUDE.md Writer: Boris Cherny's Agent Memory Rules
Persistent memory for the 137k-star agentic coder at github.com/anthropics/claude-code: drafts or prunes your CLAUDE.md with the guide's include and exclude rules and the test "would removing this cause mistakes?"
Prompt CarrotShip a Paid Remote MCP Server (mcp-boilerplate)
Cloudflare Workers boilerplate with Google or GitHub OAuth and three Stripe payment models, 1,000 stars: this skill walks an AI through shipping a paid remote MCP server on it. github.com/iannuttall/mcp-boilerplate
Prompt Carrot
Sahil Lavingia
Minimalist entrepreneurship as a working mindset: GOD mode (Growth, Optimization, Destruction), solo founder conviction, no meetings, and automate-everything reasoning applied to whatever you paste in; drawn from sahillavingia.com
Prompt CarrotSahil Lavingia
Prompt CarrotMinimalist entrepreneurship as a working mindset: GOD mode (Growth, Optimization, Destruction), solo founder conviction, no meetings, and automate-everything reasoning applied to whatever you paste in; drawn from sahillavingia.com
Cursor Bug Report and Feature Request Writer
The 33k-star public intake repo for Cursor takes bug reports and feature requests, not source code; this skill drafts issues to that bar: environment, minimal repro, expected versus actual. Source: github.com/cursor/cursor.
Prompt CarrotThe Viral 5-7 Sources Cursor Debugging Prompt
Stuck on a bug? The one-sentence incantation from x.com/tedx_ai/status/1891514191179309457 (5-7 possible sources, distill to 1-2, logs before fixes) expanded into a hypothesis-and-logs debugging loop.
Prompt CarrotTaskmaster Next-Task Loop for Coding Agents
Agents drift on long builds; this loop applies Taskmaster's cure: pick the one next unblocked task, implement, mark done, expand anything oversized, repeat with a commit between tasks; from tessl.io/blog/claude-task-master.
Prompt Carrot
Yohei Nakajima
The mind behind the 140-line script that ignited the 2023 agent boom: a VC at Untapped Capital who ships tiny AI experiments in public and thinks agents should build themselves. Grounded in github.com/yoheinakajima.
Prompt CarrotYohei Nakajima
Prompt CarrotThe mind behind the 140-line script that ignited the 2023 agent boom: a VC at Untapped Capital who ships tiny AI experiments in public and thinks agents should build themselves. Grounded in github.com/yoheinakajima.
Claude's Character Method: Design Your AI's Personality
Character training gave Claude curiosity instead of bare guardrails; this skill runs the same trait, scenario, and ranking loop to write your own assistant's personality, per anthropic.com/research/claude-character.
Prompt Carrot
Mitchell Hashimoto
HashiCorp co-founder rigor for agentic coding: plan into spec.md, scope sessions, scaffold fill-in-the-blank stubs, run anti-slop cleanups, never ship unreviewed code (mitchellh.com/writing/non-trivial-vibing).
Prompt CarrotMitchell Hashimoto
Prompt CarrotHashiCorp co-founder rigor for agentic coding: plan into spec.md, scope sessions, scaffold fill-in-the-blank stubs, run anti-slop cleanups, never ship unreviewed code (mitchellh.com/writing/non-trivial-vibing).
Write in My Voice: the Spiral Method
Voice cloning the way Spiral does it (writewithspiral.com): paste finished samples, get a stylometric fingerprint of your sentence length, word choice, and structure, then drafts that sound like you instead of a bot.
Prompt Carrot
Ian Nuttall
Indie hacker judgment from the builder of KeepSave, ilo, and a 1,000-star MCP starter kit: makes an AI vet your idea for distribution first, smallest paid version, and an agent-facing angle. Grounded in ian.is.
Prompt CarrotIan Nuttall
Prompt CarrotIndie hacker judgment from the builder of KeepSave, ilo, and a 1,000-star MCP starter kit: makes an AI vet your idea for distribution first, smallest paid version, and an agent-facing angle. Grounded in ian.is.
Replit Agent Style App Builder Brief
One plain-English sentence in, a deployable app out: modeled on Replit Agent (replit.com/ai), this brief makes any AI scaffold your idea with stack, packages, database, deploy steps, and a bug-fix pass a non-coder can follow.
Prompt CarrotAudit AI Coding Costs the ccusage Way
Local logs from Claude Code and other agent CLIs hide your real AI spend; this skill runs ccusage daily, session, and blocks reports and turns them into cost-cutting actions. Method: github.com/ryoppippi/ccusage.
Prompt Carrot
Amjad Masad
Replit's founder stopped writing code and became an agent manager: this system prompt reasons with his closed-loop, ship-from-plain-English playbook, grounded in replit.com/ai and his SaaStr AI 2026 interview on saastr.com.
Prompt CarrotAmjad Masad
Prompt CarrotReplit's founder stopped writing code and became an agent manager: this system prompt reasons with his closed-loop, ship-from-plain-English playbook, grounded in replit.com/ai and his SaaStr AI 2026 interview on saastr.com.
The LangChain ReAct Agent Prompt
Question, Thought, Action, Observation: the zero-shot ReAct agent loop that ships inside LangChain, adapted so any chat AI works your task as a visible, debuggable reasoning trace (github.com/langchain-ai/langchain).
Prompt CarrotRevid.ai Prompt-to-Short-Video Production Pipeline
Prompts, scripts, or links become finished TikTok and Shorts packages: a skill recreating revid.ai's pipeline (hook, voiceover script, scene visuals, captions, publish kit) as one paste-ready brief (revid.ai).
Prompt Carrot
Greg Isenberg
Community-first operator thinking from the Late Checkout CEO who built IdeaBrowser and The Vibe Marketer: the AI reasons in demand signals, vibe coded prototypes, and distribution-first launches, per gregisenberg.com.
Prompt CarrotGreg Isenberg
Prompt CarrotCommunity-first operator thinking from the Late Checkout CEO who built IdeaBrowser and The Vibe Marketer: the AI reasons in demand signals, vibe coded prototypes, and distribution-first launches, per gregisenberg.com.
Shubham Saboo
Reasons like the builder of the 118k-star awesome-llm-apps collection and Unwind AI: ship single-file agents that run in 3 commands, provider-agnostic, zero gatekeeping. Grounded in github.com/Shubhamsaboo.
Prompt CarrotShubham Saboo
Prompt CarrotReasons like the builder of the 118k-star awesome-llm-apps collection and Unwind AI: ship single-file agents that run in 3 commands, provider-agnostic, zero gatekeeping. Grounded in github.com/Shubhamsaboo.
Midday Open Source SaaS Build Playbook
Invoicing, time tracking, a Magic Inbox and an AI assistant in one 14.6k-star open source app: this skill plans your SaaS the Midday way, from Bun monorepo stack to AGPL plus commercial licensing. github.com/midday-ai/midday
Prompt Carrot
Nicolas Zullo
Plan first, code second: the operating mindset behind the 4.7k-star vibe coding guide and a WW2 dogfight game shipped with 500 prompts and zero human-written code, distilled into a persona (github.com/EnzeD/vibe-coding).
Prompt CarrotNicolas Zullo
Prompt CarrotPlan first, code second: the operating mindset behind the 4.7k-star vibe coding guide and a WW2 dogfight game shipped with 500 prompts and zero human-written code, distilled into a persona (github.com/EnzeD/vibe-coding).
Jeremy Howard
Code-first practicality, ULMFiT transfer learning, and the llms.txt proposal power a system prompt that reasons like the fast.ai founder and applies his practical craft to whatever you paste (jeremy.fast.ai).
Prompt CarrotJeremy Howard
Prompt CarrotCode-first practicality, ULMFiT transfer learning, and the llms.txt proposal power a system prompt that reasons like the fast.ai founder and applies his practical craft to whatever you paste (jeremy.fast.ai).
Vercel AI SDK Feature Blueprint
The 25.5k-star TypeScript toolkit's method as a planning skill: pick generateText, streamText, or generateObject, wire tools, keep providers swappable, stream UI with useChat. From github.com/vercel/ai and ai-sdk.dev.
Prompt Carrot
Hamel Husain
Twenty-plus years of ML engineering distilled into an evals-first reviewer: error analysis, LLM-as-judge, and trace inspection applied to your AI product the way the hamel.dev essays and 4,500-student Maven course teach.
Prompt CarrotHamel Husain
Prompt CarrotTwenty-plus years of ML engineering distilled into an evals-first reviewer: error analysis, LLM-as-judge, and trace inspection applied to your AI product the way the hamel.dev essays and 4,500-student Maven course teach.
Kent Beck
TDD's creator now supervises a coding genie: strict red-green cycles, hunting agent cheating, and splitting structural from behavioral commits, all judged by tidy code that works. Grounded in newsletter.kentbeck.com.
Prompt CarrotKent Beck
Prompt CarrotTDD's creator now supervises a coding genie: strict red-green cycles, hunting agent cheating, and splitting structural from behavioral commits, all judged by tidy code that works. Grounded in newsletter.kentbeck.com.