// the find
x1xhlol/system-prompts-and-models-of-ai-tools
FULL Augment Code, Claude Code, Cluely, CodeBuddy, Comet, Cursor, Devin AI, Junie, Kiro, Leap.new, Lovable, Manus, NotionAI, Orchids.app, Perplexity, Poke, Qoder, Replit, Same.dev, Trae, Traycer AI, VSCode Agent, Warp.dev, Windsurf, Xcode, Z.ai Code, Dia & v0. (And other Open Sourced) System Prompts, Internal Tools & AI Models
A dump of leaked and scraped system prompts, tool schemas, and model configs from dozens of commercial AI coding tools (Cursor, Claude Code, Devin, Windsurf, v0, Replit, etc). It's a reference collection for anyone reverse-engineering how these products structure their agent instructions and tool definitions, not a library or a piece of software.
Breadth is genuinely useful — seeing Cursor's, Devin's, and Windsurf's agent prompts side by side makes patterns in tool-calling conventions and guardrail phrasing obvious in a way no single vendor's docs would. Tool JSON schemas are included alongside prompts for several tools, which is the more actionable artifact if you're building your own agent. Content is organized by vendor/tool with consistent file naming, so it's easy to grep for a specific product.
No provenance or versioning: files are just dropped in with no source, date scraped, or note on how they were obtained, so you can't tell what's current vs. stale or verify authenticity. There's no code here at all — it's a text/markdown dump with zero programmatic structure (no index, no metadata file, no way to diff prompt versions over time). The README leads with crypto donation addresses and a paid 'security service' pitch rather than any explanation of what's in the repo or how it was assembled, which undercuts trust in the content. Legally and ethically murky — much of this is scraped/exfiltrated proprietary material from companies that explicitly try to keep their system prompts private, so treat anything you copy from here accordingly.