Proposal: Add Support for Agent Rules Standard with AGENTS.md for Cross-Agent Interoperability

[sgryphon]

Hey Roo Code community,

I'm proposing we add support for the Agent Rules standard, a minimalist community initiative for unifying natural language guidelines in AI coding agents via a single AGENTS.md file in the project root. This would complement Roo Code's existing Custom Modes (e.g., YAML prompts for agent behavior) without replacing them, allowing users to define rules like coding styles, security practices, or workflows once and share them across tools.

The problem: With the growing ecosystem of AI agents (Roo Code, Cline, Aider, GitHub Copilot, etc.), developers end up maintaining tool-specific configs -- .roomodes for Roo, CONVENTIONS.md for Aider, etc -- which leads to redundancy and friction in multi-tool workflows.

Key benefits:

• Interoperability: Reuse the same AGENTS.md across agents, simplifying teams that mix Roo with others.
• Efficiency: Less maintenance, easier onboarding, and consistent enforcement of project rules (e.g., "Prefer async/await, ensure 80% test coverage").
• Scalability: Inspired by standards like EditorConfig, it scales for large projects while keeping things simple (just Markdown/plain text, no required structure).

Tools already supporting or configurable for this: Codex CLI (native), OpenCode (native), Aider (via configuration), and Gemini CLI (via settings).

Note that this is meant to complement and be usable alongside roomodes, not replace it. It means that developers can try out Roo Code switching from an existing system (like ChatGPT Codex), and have a good "first try" experience because of the shared basic rules.

Rather than taking years to get a standard minimal configuration (like editorconfig), we have a chance to standardise this from the start.

For the full spec, examples, and rationale, check out https://agent-rules.org/.

I've detailed the implementation in this issue: #5966

What do you think? Would this be useful for your workflows? Upvote if you'd like to see it prioritized, or share thoughts/concerns below—feedback welcome! 🚀