Copilot Memory Plugin for Github Copilot CLI

rishblob — Mon, 16 Feb 2026 06:00:19 +0000

GitHub Copilot CLI Challenge Submission

This is a submission for the GitHub Copilot CLI Challenge

What I Built

Copilot Memory is a plugin system that extends GitHub Copilot CLI with persistent context awareness and project continuity capabilities. It addresses a fundamental challenge in AI-assisted development: maintaining coherent, long-term project context across sessions and team handoffs.

The system provides:

Smart Onboarding: Dual-mode initialization (--mode greenfield|brownfield) that adapts to whether you're starting fresh or integrating into existing codebases
Session Continuity: Handoff and resume workflows that preserve context, decisions, and progress across session boundaries and team member transitions
Phase-Based Development: Structured workflows for breaking down complex projects into manageable phases with goal-backward verification
Obsidian-Compatible Storage: Markdown vault structure with frontmatter metadata, making all captured context accessible in Obsidian or any text editor
Extensible Plugin Architecture: Clean packaging and distribution system with semantic versioning and GitHub Releases

Technical Highlights

The project demonstrates production-ready practices including:

53 passing tests validating core workflows (handoff, resume, phase management, vault operations)
Node.js ESM architecture using standard library features (no heavy dependencies)
12 specialized commands with detailed markdown documentation in plugin/commands/
Automated packaging via npm run plugin:pack script for distribution
Obsidian integration through compatible markdown structure (sessions, handoffs, phases, indexes)

Repository & Release

Repository: github.com/risnake/copilot-memory
Latest Release: v1.0.2

All commands

Initialization & Onboarding

Session Handoff Workflow

[Screenshot placeholder: Handoff document generation showing context preservation]

Phase Management

Key Points:

Installation via GitHub CLI plugin system
Onboarding adapts intelligently to project context (greenfield vs brownfield)
Handoff documents capture complete session state in markdown
Phase workflows support research, planning, and execution stages
All data stored as Obsidian-compatible markdown with frontmatter

My Experience with GitHub Copilot CLI

Building this extension revealed the true power of GitHub Copilot CLI as a platform, not just a tool. The experience transformed my understanding of what's possible when you combine AI assistance with thoughtful project structure.

What Worked Exceptionally Well

Context Awareness: Copilot CLI's ability to understand project structure made implementing the brownfield onboarding mode remarkably straightforward. It correctly identified existing patterns and suggested initialization strategies that respected the current architecture.

Iterative Refinement: The conversational nature of Copilot CLI excelled at refining complex workflows. When designing the handoff system, I could describe high-level goals and iteratively refine the implementation through natural dialogue rather than wrestling with documentation.

Testing Confidence: Copilot CLI helped generate comprehensive test scenarios I wouldn't have considered, particularly edge cases around session state corruption and concurrent handoff operations. This significantly improved the reliability of the final system.

Challenges & Solutions

Complex State Management: Managing persistent state across sessions required careful architectural planning. Copilot CLI helped design a clean separation between ephemeral session data and persistent project state, implemented through an Obsidian-compatible vault structure with distinct directories for sessions, handoffs, phases, and indexes.

Plugin Distribution: Packaging and distributing CLI plugins presented integration challenges. Working with Copilot CLI led to a clean packaging script (npm run plugin:pack) and manual release workflow through GitHub Releases.

Documentation Balance: Finding the right level of documentation detail was iterative. Too prescriptive and it hindered flexibility; too vague and users couldn't get started. Copilot CLI helped strike the right balance through example generation.

Impact on Development Velocity

This project benefited significantly from Copilot CLI. The ability to:

Rapidly prototype command structures and argument parsing
Generate comprehensive error handling patterns
Iterate on user experience flows through conversational refinement
Design test scenarios covering edge cases (53 tests across 12 suites)

...meant more time focusing on architecture and user value rather than implementation details.

Looking Forward

This plugin itself serves as infrastructure for future AI-assisted development. The memory and context preservation capabilities it provides will compound the effectiveness of Copilot CLI in long-running projects. It's meta-tooling: using AI to build tools that make AI more effective.

The experience has convinced me that extensible AI tooling isn't just about automation—it's about creating systems that learn and adapt to how teams actually work. That's the future of development tooling.

Minecraft MCP Server for testing model spatial reasoning capabilities

rishblob — Mon, 16 Feb 2026 05:34:44 +0000

This is a submission for the GitHub Copilot CLI Challenge

What I Built

I built minecraft-mcp-server: a local MCP server that connects AI agents to Minecraft through Mineflayer, designed specifically to test and stress 3D spatial reasoning in modern LLMs.

Instead of generic “chat with a game” behavior, this project exposes explicit, typed tools that force grounded spatial decisions in a real 3D world:

Creative mode tools for structure generation and world manipulation (setblock, fill, clone_area, fly_to, teleport_to, set_time, set_weather, etc.)
Survival mode tools for embodied task execution (go_to, dig_block, place_block, collect_block, craft_item, equip_item)
Config-driven local runtime (MC_MODE=creative|survival) so agents can run immediately without hand-configuring connection parameters

I also focused heavily on reliability for agent evaluation:

fly_to now uses a robust fallback chain (direct flight → arc flight → teleport fallback)
command tools return explicit confirmation metadata (executed, category, timedOut) so runs can be analyzed as true success/failure, not just “best effort”

To me, this is a practical testbed for the next generation of LLMs: not just language fluency, but spatial planning, coordinate reasoning, and grounded action feedback loops.

Demo

Project Repo: https://github.com/risnake/minecraft-mcp-server

Images

Reference images

#OpenAI GPT 5.3 Codex
#Anthropic Claude Opus 4.6 1. 2. 3.

Key takeaways

Opus 4.6 seems to have a much better spatial understanding compared to the latest GPT 5.3 Codex model.
Opus often added tiny details which it observed in the image which GPT often failed to do

My Experience with GitHub Copilot CLI

GitHub Copilot CLI felt like an orchestration layer for shipping an agentic system quickly: I used it to research APIs, generate implementation plans, dispatch parallel coding/research passes, and iterate on reliability bugs without losing momentum.

The biggest win was speed + structure: I could move from idea to working MCP server with mode-aware tooling, then harden it through targeted debugging (runtime interop issues, flight edge cases, command acknowledgment integrity) in a tight loop.

Most importantly, Copilot CLI helped turn a broad concept (“LLMs in Minecraft”) into a focused experiment platform for evaluating embodied 3D reasoning—which is exactly the frontier I care about.

Forem: rishblob

Copilot Memory Plugin for Github Copilot CLI

GitHub Copilot CLI Challenge Submission

What I Built

Technical Highlights

Repository & Release

All commands

Initialization & Onboarding

Session Handoff Workflow

Phase Management

My Experience with GitHub Copilot CLI

What Worked Exceptionally Well

Challenges & Solutions

Impact on Development Velocity

Looking Forward

Minecraft MCP Server for testing model spatial reasoning capabilities

What I Built

Demo

Images

Reference images

Key takeaways

My Experience with GitHub Copilot CLI