AI-Native Development OS
AI-Native Development OS
A workflow layer that gives your AI coding agent explicit context, skills, tasks, and review criteria — instead of starting from scratch every session.
npm install -g @donihadimas/aios aios init Setup → aios adopt Context → aios next Tasks → aios kit Review $ aios next
context: .aios/config.json exists
docs: docs/product/prd.md ready
tasks: docs/tasks/ has 3 open tasks
→ Recommended: T2 — Implement auth module
$ cat .aios/config.json
{
"mode": "full",
"docsRoot": "docs",
"skillDelivery": "native"
}
$ ls docs/tasks/
T1-implement-schema.md
T2-implement-auth.md
T3-add-tests.md How AIOS Works
A repeatable loop from idea to reviewed implementation.
Setup
Install AIOS, create or adopt a project.
aios init my-project Product Vision
Agent interviews you and fills the vision doc.
product-discovery PRD
Generate a PRD with testable acceptance criteria.
prd-generator Architecture
Design architecture and record decisions.
architecture-design Tasks
Break work into small, verifiable tasks.
task-breakdown Implement
Build one task at a time with tests.
task-implementation Review
Review the diff against acceptance criteria.
code-review Next
Ask what to do next, then repeat from step 5.
aios next Choose Your Path
Start a new project, adopt an existing one, or just ask what to do next.
Guided setup. Answers questions and creates your project step by step.
Create a fresh project with full AIOS structure and workflow kit.
Add AIOS structure to a project you already have. Never overwrites.
Read-only. Prints the recommended next step based on your project.
.aios/config.jsonProject Config
Single source of truth for mode, docsRoot, projectShape, and integration flags. Controls how your agent reads context.
skills/Agent Skills
Reusable Markdown workflows for product discovery, PRD generation, architecture, task breakdown, code review, and more.
templates/Document Templates
Starter structures for PRDs, ADRs, architecture docs, migration plans, and release notes.
workflows/Workflow Sequences
Ordered steps that chain skills together: discovery → PRD → architecture → tasks → implementation → review.
docs/tasks/Verifiable Tasks
Small work units with acceptance criteria, affected files, dependencies, and done conditions before moving on.
docs/adr/Architecture Decisions
Record trade-offs, constraints, and decisions as ADRs so agents and humans share the same rationale.
context/Context Maps
Entry points that tell agents which documents to read before planning, implementing, or reviewing.
commands/CLI Commands
Helper commands that create, copy, validate, and recommend next steps. Human-owned decisions stay human.