arch-init

Initializes a multi-agent project harness with git, AGENTS.md, stack, and quality gate.

Updated May 8, 2026
One-click install
npx skills add https://github.com/juanca202/sdd-devkit --skill arch-init-juanca202
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: arch-init
Source: https://github.com/juanca202/sdd-devkit/tree/main/skills/arch-init
Command: npx skills add https://github.com/juanca202/sdd-devkit --skill arch-init-juanca202

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and assets (resource) components.

What problem does it solve? Starting a project for AI-agent-assisted development requires consistent scaffolding: a git repository, agent instruction files, a defined technology stack, documented architecture decisions, and a minimum quality gate. Doing this manually is error-prone and inconsistent, especially across single-repo and multi-repo solutions. ## Core Features & Use Cases - Project bootstrap from any starting point: Detects whether a project has no code, scaffolded code, existing implementation, or is specs-only, and completes whatever harness pieces are missing. - Harness file creation: Generates AGENTS.md, .agents/MEMORY.md, .sdd-devkit/settings.json, ADR and standards indexes, and a root README description from fixed templates, with idempotent re-runs. - Stack setup and quality gate: Researches or suggests a technology stack, scaffolds it, and configures a minimum test gate (unit tests plus relevant layers like E2E or API testing). - Multi-repo support: Creates a specifications repository that aggregates other repos as git submodules, each with its own AGENTS.md and architecture indexes. - Use Case: You have an empty folder for a new product. Invoke the skill and it asks what you want to build, suggests a stack, scaffolds it, creates all agent instruction files, configures a test framework, and documents the initial decisions as ADRs. ## Quick Start Ask the agent to initialize the harness for this project with arch-init and answer its questions about topology, stack, and testing layers.

Frequently Asked Questions about arch-init

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up AGENTS.md for a new project?▼

Run arch-init, which copies a fixed AGENTS.md template with context sources, general rules, and a technology stack section. The stack section is filled in at the end of the flow after the stack is detected or installed.

How to bootstrap a multi-repo project for AI agents?▼

arch-init asks whether the solution spans one or more repositories. For multiple repos it creates a specifications repository as the main root and adds each additional repository as a git submodule, giving each its own AGENTS.md and README.

Does arch-init work on a project that already has code?▼

Yes. It classifies the project as having an existing implementation, skips stack installation, and delegates to arch-discover to mine architecture decisions from the code. Existing harness files are checked against templates and only completed if missing.

Can I re-run arch-init on an already initialized project?▼

Yes, it is idempotent. Files already conforming to their templates are left untouched, missing pieces are created, and files with non-conforming formats are delegated to plugin-migrate for normalization.

What test frameworks does the quality gate setup support?▼

It configures unit tests by default using the stack-appropriate framework such as Vitest, pytest, JUnit 5, go test, PHPUnit, or xUnit. Additional layers like E2E with Playwright or API testing are suggested based on project type.

When should I not use arch-init?▼

Do not use it to define user stories, plan tasks, or implement features; those belong to work-define, work-plan, and work-implement. It only bootstraps the harness and normally runs once per project.