A Codex skill package for preparing AI-coding-ready project context, starter documentation, domain guardrails, and agent instruction files before implementation starts.
🌐 Languages:
|
AI Project Starter prepares project context, PRDs, architecture notes, guardrails, and agent instruction files before an AI coding agent starts implementation. Start here: Usage guide shows the supported modes. |
AI ile kod yazmaya başlamadan önce projenin brief'ini, PRD'sini, mimarisini, risklerini ve ajan talimatlarını toparlamak için hazırladığım başlangıç seti. Koddan önce zemini netleştirmek isteyen projeler için. Buradan başla: Türkçe başlangıç için README.tr.md; modlar ve kullanım akışı için USAGE. |
- 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.
| 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. |
| 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. |
| Generator smoke test | npm run check validates repository structure and dry-runs the starter materializer without writing files. |
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.
| 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 |
| 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 |
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.
PROMPT_ONLYDOCS_ONLYPLAN_ONLYCREATE_FILESAUDIT_CONTEXTREPAIR_CONTEXTENTERPRISE_STARTERVIBE_GUARDRAILSRESEARCH_BACKED
.
|-- .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
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.
npm run check
git diff --check
gitleaks dir . --no-banner --redact
git status --short --branchnpm run check runs the repository validator and a dry-run smoke test for .agents/skills/ai-project-starter/scripts/create_starter.py.
Do not open public issues for vulnerabilities, leaked credentials, private prompts, or accidental disclosure. See SECURITY.md.
MIT. See LICENSE.