Back to browse
GitHub Repository

Use subagents to meet your requirements

2 stars

Skillwave – agent orchestrator with async comms and goal-alignment loop

by guyzana·Mar 30, 2026·2 points·0 comments

AI Analysis

●●SolidBig BrainShip It

Stores agent state in Markdown files for full Git history and auditability.

Strengths
  • Markdown-based state storage enables full Git versioning of agent decisions.
  • Async INBOX.md pattern prevents agent blocking while waiting for user input.
  • Dynamic role creation allows specialized subagents for different task requirements.
Weaknesses
  • Tightly coupled to GitHub Copilot limits portability to other LLM providers.
  • Scaffold requires significant configuration before becoming a production-ready solution.
Category
Target Audience

Developers building autonomous AI agents with GitHub Copilot

Similar To

CrewAI · AutoGen · LangGraph

Post Description

Skillwave is a seed repository for scaffolding end-to-end projects with an autonomous agent workflow.

You give it a GOAL.md file with roughly 50–100 user stories, and it decomposes the work into tasks, dynamically creates distinct roles, assigns tasks and executes via subagents, and keeps going until everything all requirements met or all tasks are blocked on user input.

A few design choices that may be interesting/novel:

- INBOX.md for async user communication, so questions don't halt the whole system

- GOAL.md as the persistent source of truth across sessions

- Git commit per completed task(s), linked back to the backlog

- A periodic goal-alignment loop that re-checks the backlog every 5 completed tasks

- All state stored in Markdown and Git rather than a separate database

I tested it on several projects and got better results than I expected, so I cleaned it up and open-sourced it.

Similar Projects

Developer Tools●●Solid

MBC – AI agent orchestration for Laravel

Claude agent orchestration for Laravel, but multi-agent automation exists in CrewAI, Autogen.

Ship ItSolve My Problem
undergracelabs
213mo ago