AI Project Starter

A Codex skill package for preparing AI-coding-ready project context, starter documentation, domain guardrails, and agent instruction files before implementation starts.

🌐 Languages: 🇩🇪 · 🇪🇸 · 🇬🇧 · 🇧🇷 · 🇹🇷 · 🇫🇷

Status: initial public-ready skill package
License: MIT
Project type: Markdown-based Codex skill package for context engineering and project starter documentation
Note: Independent community project. Not affiliated with, endorsed by, or sponsored by OpenAI.

AI Project Starter helps turn a raw project idea into a complete documentation foundation before an AI coding agent starts writing code. It is built for Codex, Cursor, Claude Code, Windsurf/Devin, Continue, Copilot, Aider, and similar tools.

🧭 Enterprise Evaluator Path

If you need to prove...	Start with	Evidence you get
The starter will not overwrite existing work by default	Install guide and Usage guide	Modes, copy path, and safe materialization expectations.
The generated context has security and quality gates	Security model	Secret handling, approval boundaries, testing expectations, and public-safe rules.
The skill can support a real enterprise starter	Examples	SaaS, cybersecurity, API, web, mobile, desktop, data, internal-tool, and library/CLI patterns.
The public repository is safe to share	Public repo checklist	Leak-prevention checklist before commit, tag, release, or reuse.

✅ Trust Signals

Signal	Standard
Context before code	The skill prepares product, architecture, security, testing, operations, and task context before implementation starts.
Mode-gated output	`PROMPT_ONLY`, `DOCS_ONLY`, `PLAN_ONLY`, `CREATE_FILES`, `AUDIT_CONTEXT`, and `REPAIR_CONTEXT` keep the workflow explicit.
Public-safe package	Examples and templates use placeholders and avoid credentials, private prompts, customer data, and local operator paths.
Agent interoperability	Output targets Codex, Claude Code, Cursor, Continue, Copilot, Devin/Windsurf, Aider, and similar tools without locking the project to one editor.

🎯 Why This Exists

AI coding agents fail when the project has no durable context. Vague "build this" prompts often create duplicate modules, weak security, unclear architecture, and features outside the real goal.

This repository packages a reusable Codex skill that creates or repairs:

project briefs and PRDs,
technical specs and architecture docs,
implementation plans and task lists,
security, testing, and quality gates,
domain starter packs for SaaS, cybersecurity, AI products, APIs, web, mobile, desktop, data products, internal tools, and library/CLI work,
agent instruction files such as AGENTS.md, CLAUDE.md, Cursor rules, Continue rules, Copilot instructions, and Codex context files.

⚡ Start Fast

I want to...	Use this
Start a new AI-coded project safely	Skill entrypoint
Understand modes and examples	Usage guide
Install/copy the skill	Install guide
Review file structure	Skill structure
Check public repo safety	Public repo checklist
Prepare GitHub metadata	GitHub settings
Improve discoverability	SEO and discoverability
Review security model	Security model

🧩 What You Get

Capability	What it gives you
Project context foundation	`README.md`, `PROJECT_BRIEF.md`, `PRODUCT_REQUIREMENTS.md`, `TECHNICAL_SPEC.md`, `ARCHITECTURE.md`, `IMPLEMENTATION_PLAN.md`, `TASKS.md`
Agent-ready instructions	`AGENTS.md`, `CLAUDE.md`, Cursor, Continue, Copilot, Devin/Windsurf, Aider, and Codex context files
Enterprise starter	security, testing, observability, deployment, operations, risk register, ADRs
Domain packs	SaaS, cybersecurity, AI product, API, web, mobile, desktop, data, internal tool, library/CLI docs
Vibe-coding guardrails	anti-slop, non-goals, milestone control, no duplicate systems, verification-before-done
Context audit/repair	detect weak, stale, duplicate, conflicting, or missing project context docs
Template materializer	`scripts/create_starter.py` creates safe starter docs without overwriting by default

🚀 Quick Start

Use the skill from this repository:

Use the ai-project-starter skill to prepare project starter docs for: [your project idea]

Create files for an enterprise SaaS cybersecurity product:

Use the ai-project-starter skill in CREATE_FILES + ENTERPRISE_STARTER mode.
Project idea: A SaaS cybersecurity platform for vulnerability triage, tenant-based teams, RBAC, audit logs, AI-assisted remediation planning, and secure reporting.

Ask for guardrails only:

Use VIBE_GUARDRAILS mode. Prepare only the agent context and quality gates needed to keep this project from drifting during vibe coding.

🧰 Supported Modes

PROMPT_ONLY
DOCS_ONLY
PLAN_ONLY
CREATE_FILES
AUDIT_CONTEXT
REPAIR_CONTEXT
ENTERPRISE_STARTER
VIBE_GUARDRAILS
RESEARCH_BACKED

📁 Repository Structure

.
|-- .agents/skills/ai-project-starter/
|   |-- SKILL.md
|   |-- agents/openai.yaml
|   |-- references/
|   |-- scripts/create_starter.py
|   `-- templates/
|-- docs/
|   |-- INSTALL.md
|   |-- USAGE.md
|   |-- EXAMPLES.md
|   |-- SKILL_STRUCTURE.md
|   |-- PUBLIC_REPO_CHECKLIST.md
|   |-- SEO.md
|   |-- GITHUB_SETTINGS.md
|   |-- SECURITY_MODEL.md
|   `-- ROADMAP.md
|-- README.md
|-- README.tr.md
|-- SECURITY.md
|-- CONTRIBUTING.md
|-- CHANGELOG.md
`-- RELEASE_NOTES.md

🛡️ Public Safety Rules

This repository must not contain:

API keys, tokens, credentials, cookies, private keys, certificates, or private URLs,
customer data or internal company data,
private system prompts,
proprietary project implementation details,
unredacted logs, screenshots, local paths, or personal notes.

Before publishing, use docs/PUBLIC_REPO_CHECKLIST.md.

🔐 Security

Do not open public issues for vulnerabilities, leaked credentials, private prompts, or accidental disclosure. See SECURITY.md.

⚖️ License

MIT. See LICENSE.