If you use AI coding tools, you are probably suffering from AI Rule Fatigue.
Right now, the ecosystem is deeply fragmented. You write a perfect .cursorrules file for Cursor, but then you want to test Claude Code and have to completely rewrite your slash commands. Then your team member uses Windsurf (Cascade) or Cline, and your configurations break all over again.
We shouldn't have to rewrite our workflow logic for every new AI agent that hits the market.
That is exactly why I built EGM Universal Agent.
The Elephant-Goldfish Model (EGM) is a framework originally conceived for Spec-Driven Development (SDD) with autonomous AI coding agents. It balances long-term structural architecture (the Elephant memory) with fast, context-isolated execution iterations (the Goldfish memory).
The Universal Agent Edition takes this methodology and turns it into a single, canonical, cross-compatible standard.
Instead of maintaining separate prompt systems, this repository acts as a cross-agent runtime adapter.
👉 Check out the project here: yathamsriram243-blip/egm-universal-agent
Here is what EGM Universal Agent brings to your autonomous coding environment:
.cursorrules, .clinerules, .windsurfrules, and more.
Whether you prefer heavy IDE integration or native terminal agents, EGM Universal Agent maps across the board:
| Native IDEs & Extensions | Terminal & CLI Agents | Open-Source Agents |
|---|---|---|
| Cursor | Claude Code | OpenCode |
| Windsurf (Cascade) | Gemini CLI | Continue |
| VS Code Copilot | OpenAI Codex | Cline /Roo Code /OpenHands |
Ready to try it out? You can clone the standard environment and configurations straight into your current project workspace:
git clone https://github.com.git
cd egm-universal-agent
pip install -r requirements-dev.txt
Inside the repository, you'll find the .agents/ and plugins/ architectures ready to be loaded or linked right into your IDE settings.
This is an open-source initiative licensed under the Apache-2.0 License. As the AI agent workspace shifts rapidly, keeping our rules and workflows standardized is a community effort.
What AI coding agent are you currently building with, and what is your biggest headache keeping your custom prompt rules consistent? Let's discuss in the comments below!