Theia Ide System Prompts

---
title: "Theia AI: Open System Prompts by Design"
source: "https://eclipsesource.com/blogs/2025/04/29/system-prompt-theia-ai-leaked/"
category: "system-prompts"
tags: [theia, open-source, transparency, MIT-license, eclipse]
---

# Theia AI: Open System Prompts by Design

The internet is full of posts and discussions claiming to reveal leaked system prompts from popular proprietary AI coding tools. While not all of these prompts have been officially confirmed — and some may be incomplete, outdated, or fabricated — they highlight an important point: Prompt engineering is the central gear in AI-native IDEs — and it's almost always hidden.

## The Great Prompt Leaks of 2023-2025

Here's a snapshot of what the community has uncovered or discussed:

- The system prompt of GitHub Copilot Chat was allegedly exposed via prompt injection, revealing its internal behavior rules and response guidelines.
- Cursor, a Claude-based AI IDE, reportedly exposed its full system prompt to users during a 2025 update.
- Replit Ghostwriter has been the subject of several community-shared prompt dumps.
- A prompt attributed to Windsurf, Codeium's new tool, circulated online showing a highly dramatized alignment strategy. Codeium later clarified this was experimental, not used in production.
- Similar claims have emerged around Vercin v0 and Devin AI.

Whether all of these leaks are genuine or not, they reinforce a broader issue: in many AI-powered tools, users and developers are left in the dark about what drives the assistant's behavior.

## Open by Design: The Theia IDE Difference

With the AI-powered Theia IDE, all prompts are:

- Stored as individual, readable files
- Licensed under the MIT License
- Editable live in the IDE — even by end users

Users and developers alike can fully inspect, modify, and experiment with AI behavior. This contrasts with proprietary tools where prompts are the "secret sauce."

## Why MIT?

The MIT license removes legal friction and encourages innovation: prompt tweaks for industry-specific use cases, experimental workflows for unique team setups, and community-driven enhancements.