Proposal: Add GitHub Copilot as a Supported Interface for Quint Code

[danstis]

Rationale

This proposal is a UX and ecosystem expansion for Quint Code.

Quint Code’s core value is providing structured, auditable reasoning for AI coding tools by installing framework-aware commands and configuration. Today, this is supported for Claude Code, Cursor, Gemini CLI, and Codex CLI.

GitHub Copilot is increasingly used as a primary AI interface for coding and design decisions inside repositories. Adding first-class support for GitHub Copilot would allow users to apply the Quint decision framework directly within Copilot, rather than re-implementing Quint concepts via ad-hoc prompts.

This keeps Quint authoritative while extending its reach to a widely adopted AI environment.

---

Problem Statement

Currently, Quint Code does not support GitHub Copilot as an execution interface for the framework. As a result:

• Quint commands (/q0-init, /q1-hypothesize, etc.) cannot be invoked from Copilot.
• Copilot has no canonical understanding of the Quint decision cycle, FPF stages, or expected outputs.
• Users must manually translate Quint’s methodology into prompts, losing consistency and auditability.

This creates friction and undermines Quint Code’s goal of turning AI-assisted development into a rigorous, repeatable reasoning process.

---

Proposed Solution

Add GitHub Copilot support as a first-class interface, similar to existing tool integrations.

Concretely, this could include:

1. Copilot Instruction / Command Files

   • Generate GitHub Copilot–compatible instruction files (e.g. .github/copilot-instructions.md, AGENTS.md, or equivalent)
   • Encode:
     • The Quint decision cycle (abduction → deduction → induction → audit → decide)
     • Command semantics equivalent to existing /q* commands
     • Output structure expectations aligned with .quint/ artifacts

2. quint-code init --copilot Flag

   • Add a CLI flag that scaffolds the required Copilot files
   • Parallels existing flags (--claude, --cursor, --gemini, --codex)
   • Optionally supports --all and --local where applicable

3. Framework Parity

   • Ensure Copilot usage preserves:
     • Evidence levels (L0 → L2)
     • Auditability and rationale capture
     • Epistemic debt tracking concepts

This would allow GitHub Copilot to act as an execution surface for Quint, not just a generic assistant.

[m0n0x41d]

Hello and thank you for submitting this issue.

I think adding Copilot might be a very nice feature for both Copilot and Quint Code.

However, I am very far from being an experienced Copilot user, let alone a non-Copilot user at all.

Please give me some time to try and read its docs before we pick this one into the roadmap.

Thank you

[m0n0x41d]

At first glance, it might be relatively easy to add Copilot support. MCP is supported. Instead of slash commands in Copilot, we might use pre-saved prompts..github/prompts/*.prompt.md

@danstis Let me know if you want to pick up the progress. If not, I will get back to you after the holidays.

[danstis]

If you are willing to work on this then you'll probably do a better job than me. However I would take a look if you were not willing to. I am happy to perform testing on new builds if desired, I have access to Copilot CLI.

[m0n0x41d]

Sure,this will work. I will implement Copilot support and will call you for testing

Thank you