A curated collection of best practices, patterns, and workflows for AI/LLM-powered software development — aggregating knowledge across prompting, agents, tooling, and Git integration into one place.
AI tools (GitHub Copilot, Cursor, Claude, ChatGPT, etc.) are becoming core parts of the development workflow. But scattered tips and fragmented docs make it hard to build a consistent, effective practice. This repo serves as a single aggregation point — bringing together proven techniques from across the AI-assisted development ecosystem so you can learn, reference, and apply them quickly.
| Directory | Description |
|---|---|
agents/ |
Agent instructions & AGENTS.md — how to write project-level rules that make AI Agents behave predictably and follow your team's conventions |
prompt/ |
Prompt engineering best practices — the 5 key elements of a good task prompt, Plan vs Agent mode, session management |
git/ |
Git + LLM integration — auto-generating commit messages, configuring Copilot for Git workflows |
mcp/ |
MCP (Model Context Protocol) — extending agent capabilities with external tools |
- Read
agents/Repos-Agents-instructions.mdto learn how to set upAGENTS.mdfor your project - Read
prompt/best-practice.mdfor writing effective prompts and managing AI sessions - Apply the patterns to your own repos and iterate
PRs and issues welcome. Keep entries concise and actionable.