Quickstart Factory

An AI-powered tool for managing the full lifecycle of Red Hat AI Quickstarts — from idea to published quickstart. Just say "hello" and your AI assistant takes it from there.

What it does

Quickstart Factory connects your AI coding assistant (Claude, Cursor, Gemini, or Codex) to the AI Quickstart backlog on GitHub, giving you a conversational interface to:

• Browse and search the backlog — filter by author, assignee, label, or keyword
• Create issues — add new quickstart suggestions with metadata
• Groom the backlog — readiness scoring, prioritization, coverage analysis
• Write blog posts — generate drafts for completed quickstarts
• Identify gaps — find missing industries, technologies, or use cases

How it works

Open the project in your AI client and say "hello". The assistant automatically syncs skills, fetches the live backlog from GitHub, and presents a dashboard:

From there, just ask what you need in natural language — query issues, drill into details, discover linked implementation repos, and generate blog drafts:

See the full walkthrough with screenshots for the complete workflow.

Getting started

Prerequisites

• Python 3
• gh CLI installed and authenticated (gh auth status)
• An AI coding client: Claude Code, Cursor, Gemini Code Assist, or Codex

Setup

1. Clone the repository:

git clone https://github.com/ccamacho/quickstart-factory.git
cd quickstart-factory

2. Open in your AI client and say "hello":

claude                       # Claude Code
cursor quickstart-factory/   # Cursor IDE

The assistant will automatically sync skills on first run — no manual setup needed.

Skills

Skill	What it does
gh-backlog-reader	Read, filter, and summarize GitHub issues. Supports --issue (single issue with comments and linked repos), --author, --assignee, --label, --search, --state, --detail.
gh-issue-creator	Create new [Quickstart suggestion] issues with duplicate detection. Supports --name, --description, --technology, --industry, --dry-run.
pipeline-grooming	AI-assisted backlog grooming with readiness scoring, industry/technology coverage analysis, and prioritization recommendations.
blog-writer	Generate blog post drafts for completed quickstarts using standard templates and messaging guidelines.
quickstart-identifier	Gap analysis comparing the current backlog against industry trends and technology adoption targets.

CLI usage

You can also run the scripts directly:

# Backlog summary
python3 core/skills/gh-backlog-reader/scripts/read_backlog.py --summary

# Single issue with comments and linked repos
python3 core/skills/gh-backlog-reader/scripts/read_backlog.py --issue 39

# Issues by a specific author
python3 core/skills/gh-backlog-reader/scripts/read_backlog.py --author ccamacho

# Search for topics
python3 core/skills/gh-backlog-reader/scripts/read_backlog.py --search "Financial Services"

# Filter by label
python3 core/skills/gh-backlog-reader/scripts/read_backlog.py --label Acknowledged

# Create an issue (dry-run first)
python3 core/skills/gh-issue-creator/scripts/create_issues.py \
  --name "GraphRAG" \
  --technology "Llama Stack" \
  --dry-run

Resources

• Backlog: https://github.com/rh-ai-quickstart/ai-quickstart-contrib/issues
• Quickstart template: https://github.com/rh-ai-quickstart/ai-quickstart-template
• Quickstart catalog: https://docs.redhat.com/en/learn/ai-quickstarts
• GitHub org: https://github.com/rh-ai-quickstart

Project structure

AGENTS.md               ← AI agent guidelines (read by all clients)
CLAUDE.md               ← Symlink → AGENTS.md (for Claude Code)
GEMINI.md               ← Symlink → AGENTS.md (for Gemini)
core/
  skills/
    gh-backlog-reader/  ← Read GitHub backlog
    gh-issue-creator/   ← Create GitHub issues
    pipeline-grooming/  ← Backlog grooming
    blog-writer/        ← Blog post generation
    quickstart-identifier/ ← Gap analysis
  scripts/
    sync-clients.sh     ← Sync skills to AI clients
data/                   ← Generated output (gitignored)
  blog-drafts/          ← Blog post drafts
  reports/              ← Grooming reports

Multi-client support

The same skills work across Claude, Cursor, Gemini, and Codex. Each client reads AGENTS.md through its preferred file:

Client	Reads from
Claude Code	CLAUDE.md → AGENTS.md
Gemini	GEMINI.md → AGENTS.md
Codex	AGENTS.md directly
Cursor	.cursor/rules/agents.md → AGENTS.md

Skills are synced as symlinks to each client's directory via bash core/scripts/sync-clients.sh.

License

Licensed under Apache License 2.0. See LICENSE.