Show HN: AI Coding Agents: Intent-Driven Development Guidelines (github.com)

A new open-source project, AI Intent Driven Development (IDD), publishes a structured guideline set for AI coding agents, assistants, and LLMs that formalizes how user requests become executable work. The core idea is the "Intent" — a self-contained document with WHY (motivation), WHAT (requirements, often expressed in Gherkin), and HOW (a taskized implementation plan). Delivered as a collection of markdown files and an installable script, the repo aims to make agent-driven development explicit and repeatable, reducing ambiguity and improving alignment before any code is written. Technically, the project bundles modular docs (AGENTS.md, INTENT_SPECIFICATION.md, PLANNING.md, DOMAIN_RESOURCE_ACTION_ARCHITECTURE.md, PHOENIX_DEVELOPMENT.md, CODE_GUIDELINES.md, etc.), a shell tool (ai-intent-driven-development.sh / aidd) with commands to scaffold language-agnostic or language-specific agent guidelines (examples: elixir, phoenix, CLAUDE.md), and testing via Bats-Core. It promotes a Domain-Resource-Action architecture where each Action is a single-responsibility module — designed to refactor Phoenix generators and portable across languages/frameworks. Usage notes: you must add the guidelines to an agent’s session (and prompt the agent to read all docs, ask clarifying questions, then produce an Intent) so the agent can execute tasks sequentially. The repo also includes contributor workflow and submodule initialization, making it practical for teams building repeatable, auditable LLM-driven development processes.