Somebody Leaked the System Prompts of 32 AI Agents (Cursor, Claude Code)

Description

Somebody put the system prompt of every shipping AI coding tool into one GitHub repo — Cursor, Claude Code, Devin, v0, Lovable, thirty-plus more. It has 137,502 stars. And reading it tells you exactly which tool is lying. We grep across the leaked prompts of Cursor, Claude Code, Devin, and Lovable side-by-side. Devin's persona ("real code-wiz"). Lovable's hard-coded stack lock. Cursor's four-tool search router vs Claude Code's `system-reminder` block. The one rule that appears in every single prompt. And why this repo has survived 14 months without a single DMCA — same maintainer also runs ZeroLeaks, the paid service to *prevent* prompt extraction. TIMESTAMPS 0:00 - Cold open: LEAKED 0:25 - The repo: 137K stars, 32 folders 1:00 - Devin's pep talk 1:30 - Lovable's stack lock 2:03 - Cursor vs Claude Code 2:43 - The one shared line 3:06 - What the prompts reveal 3:36 - Why this repo survives 4:01 - Verdict: tool-use schema is the moat LINKS & RESOURCES - Repo: https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools - Cursor Agent Prompt 2.0: https://raw.githubusercontent.com/x1xhlol/system-prompts-and-models-of-ai-tools/HEAD/Cursor%20Prompts/Agent%20Prompt%202.0.txt - Anthropic Claude Code 2.0: https://raw.githubusercontent.com/x1xhlol/system-prompts-and-models-of-ai-tools/HEAD/Anthropic/Claude%20Code%202.0.txt - Devin AI system prompt: https://raw.githubusercontent.com/x1xhlol/system-prompts-and-models-of-ai-tools/HEAD/Devin%20AI/Prompt.txt - Lovable Agent Prompt: https://raw.githubusercontent.com/x1xhlol/system-prompts-and-models-of-ai-tools/HEAD/Lovable/Agent%20Prompt.txt - Repo README (latest update date): https://raw.githubusercontent.com/x1xhlol/system-prompts-and-models-of-ai-tools/HEAD/README.md - ZeroLeaks (same maintainer): https://zeroleaks.io TAGS #AiCodingTools #ClaudeCode #Cursor #SystemPrompt #DevinAi