Overview

What It Helps You Do

Use this skill to set up OpenSpec for your project through a guided interview that produces a complete openspec/config.yaml file configured for the QRSPI methodology.

Activate it with:

• /accelint-onboard-openspec
• Phrases like "set up openspec" or "generate openspec config"
• Related requests about QRSPI config, onboarding to OpenSpec, or config.yaml generation

It helps you:

• Generate project-specific OpenSpec configuration from scratch
• Import and restructure existing config files into the QRSPI schema
• Refresh stale configuration when your codebase has evolved
• Document your project's architectural facts and patterns for AI-generated artifacts

When to Use

Use this skill when:

• Users mention "openspec config", "config.yaml for openspec", or "QRSPI config"
• Starting a new project that will use OpenSpec for spec-driven development
• You have an existing config that needs to be restructured or updated
• Your codebase has changed significantly since the config was last generated

How It Works

The skill adapts its approach based on your project's current state:

Create Mode — Runs a full conversational interview covering project identity, tech stack, architecture, domain concepts, performance targets, code patterns, and per-artifact rules. Groups related questions into natural turns instead of asking everything at once.

Import Mode — When it finds existing config with unrecognized structure, offers three options: restructure onto QRSPI schema, append new sections alongside existing content, or dry run to preview without changes.

Refresh Mode — For existing QRSPI configs, runs abbreviated interview targeting only drift (new packages, tooling changes), unresolved TODOs, and external findings from completed changes.

After the interview, the skill spawns parallel discovery agents to fill gaps by scanning runtime versions, TypeScript config, test framework, directory organization, and CI/CD workflows. Inferred values show their source; unresolved fields get marked # TODO: fill in. You review a labeled preview before any files are written.

Good to Know

Good to know: This skill produces the project DNA layer (structural facts about what the project is). Its companion skill accelint-onboard-agents produces the behavior layer (how the agent acts and makes decisions). If you mention commit conventions, workflow steps, or tool preferences during the interview, the skill notes them for AGENTS.md instead of config.yaml.

Good to know: The skill infers missing fields from your codebase before asking. If inference fails, it marks the section # TODO: fill in rather than omitting it—an explicit TODO is actionable, while a missing section silently degrades every AI artifact.

Prerequisites

• Basic understanding of your project's tech stack and architecture
• Access to the project repository
• (Optional) ARCHITECTURE.md — if present, used to pre-fill deployment and infrastructure questions

What You Get

A complete openspec/config.yaml file with two sections:

context: — Objective facts injected into every AI artifact: stack details, architecture patterns, domain concepts, performance targets, code patterns, and anti-patterns.

rules: — Per-artifact quality standards: proposal scope requirements, design section structure, task vertical slicing preferences, and spec format conventions.

Examples

Example: Creating config from scratch

/accelint-onboard-openspec

Runs full interview, shows labeled preview with inference sources, writes openspec/config.yaml with complete QRSPI configuration.

Example: Refreshing stale config

/accelint-onboard-openspec

Detects existing QRSPI config, scans for drift (new packages, tooling changes), surfaces unresolved TODOs, shows only changed sections before updating.

Example: Importing existing config

/accelint-onboard-openspec

Detects unrecognized structure, presents restructure/append/dry-run options. Restructure mode maps existing content onto QRSPI schema and flags behavioral content for AGENTS.md.

Related

• accelint-onboard-agents — Generates AGENTS.md/CLAUDE.md (agent behavior layer)
• accelint-qrspi-propose — Uses config.yaml to plan OpenSpec changes
• accelint-architecture-doc — Generates ARCHITECTURE.md (system overview)