What it does

This repository collects and publishes system prompts, internal tools, and AI models extracted from popular AI coding assistants and productivity tools. The list is extensive: Cursor, Claude Code, Devin, Replit, Windsurf, VSCode Agent, Lovable, Perplexity, and roughly fifteen others. Think of it as a field guide to how these products instruct their underlying LLMs behind the scenes.

The interesting bit

System prompts are usually treated as proprietary IP—the “secret sauce” that shapes tone, tool use, and safety guardrails. By surfacing them side-by-side, the repo turns competitive intelligence into a public good (or public spectacle, depending on your legal team). The author also runs a side business, ZeroLeaks, that helps startups prevent exactly this kind of extraction.

Key highlights

• Covers 25+ products including newer entrants like Junie, Kiro, Leap.new, and Z.ai Code
• Includes not just prompts but “internal tools & AI models” where obtainable
• Active Discord community (LeaksLab) for ongoing contributions
• Sponsored by Latitude, an open-source monitoring platform for AI agents
• Accepts crypto, Patreon, and Ko-fi—funding model is unusually transparent

Caveats

• README is almost entirely marketing and sponsorship info; actual prompt contents are not shown in the preview
• “Latest Update” date is listed as 10/05/2026, which is in the future—likely a typo or placeholder
• Legal status of redistributing proprietary prompts is unclear; no DMCA or licensing discussion visible

Verdict

Worth a bookmark if you build AI products and want to benchmark how competitors frame tool use, context windows, and user safety. Skip it if you’re looking for clean, analysis-ready research—the value is in the raw leaks, not the presentation.