@indie_maker_ffox: Sharing my experience using matt's skill set to keep codex running projects stably for a long time. The same process has been fully run for 2 projects, tested and reliable. 1. First, set up matt skills well and choose an issue tracker. I always use GitHub: automatically create is…

X AI KOLs Timeline Tools

Summary

Sharing experience using matt's skill set with Codex to run projects stably, including setting up GitHub, requirement alignment, prototyping, task breakdown, and implementation steps, and recommending the use of TanStarter template.

Sharing my experience using matt's skill set to keep codex running projects stably for a long time. The same process has been fully run for 2 projects, tested and reliable. 1. First, set up matt skills well and choose an issue tracker. I always use GitHub: automatically create issues, configure labels, and establish dependencies, which makes subsequent progress much smoother. 2. Then use wayfinder / grill-with-docs to align requirements: what to do, how to do it, acceptance criteria, and end conditions should all be clearly explained. Note that at this step, the AI will ask follow-up questions until the details are thoroughly covered. 3. Then, do prototyping for the interface. It will directly write code to build the interface skeleton, and you can iterate and modify if not satisfied. Many issues where 'the effect is not right' often stem from this step not being done or done inadequately. 4. Next, to-spec + to-tickets: generate a development plan, break down large tasks into executable small tickets, and establish dependencies. Completing one automatically unlocks the next, making progress clear and pace easy to control. 5. Finally, let codex create a new goal, and sequentially call implement according to the plan for implementation. Implement will include TDD + code review to ensure quality, not relying on luck. Tip: For each new issue, let codex create a new task 'implement issue xx' to generate an independent session in the sidebar. The benefit is that each issue has a separate context, preventing interference; moreover, if desired, sub-tasks can be advanced in parallel. Note: It should be noted that I develop based on the TanStarter template. Friends with experience should know that AI excels at learning from existing project code. If the previous code is a 'shit mountain' (messy code), then what's written will be like 'carving flowers on shit'. If the previous code architecture is excellently designed, then the written code will be fast, good, easy to test, and task execution efficient. It's recommended to develop based on high-quality goals, such as TanStarter (http://tanstarter.dev) or MkSaaS (http://mksaas.com).
Original Article
View Cached Full Text

Cached at: 08/23/26, 03:32 AM

Sharing my experience using Matt’s skill set to run Codex stably on long-term projects

I’ve fully completed two projects with the same workflow, and it’s proven reliable

  1. First, set up Matt’s skills and choose an issue tracker. I always use GitHub: automatically creating issues, configuring labels, and pulling in dependencies makes later progress much smoother

  2. Use wayfinder / grill-with-docs to align on requirements: clarify what to do, how to do it, acceptance criteria, and termination conditions. A tip: the AI will ask follow-up questions to ensure you’ve thoroughly discussed all details

  3. Next, create a prototype for the interface. It will directly write code to build the UI skeleton, and you can iterate if unsatisfied. Many “the effect isn’t right” issues often stem from skipping or insufficiently completing this step

  4. Then use to-spec + to-tickets: generate a development plan, break down large tasks into executable tickets, and establish dependencies. Completing one ticket automatically unlocks the next, making progress clear and pace controllable

  5. Finally, have Codex create a goal and sequentially call implement to execute according to the plan. Implement includes TDD + code review, ensuring quality without relying on luck

Tip: For each new issue processed, have Codex create an “implement issue xx” task to generate a separate session in the sidebar. The benefit is that each issue has its own context, avoiding cross-talk; moreover, subtasks can even be advanced in parallel if desired

Note: It’s important to mention that I develop based on the TanStarter template. Experienced friends should know that AI excels at learning from a project’s existing codebase. If the prior code is messy, the output will be like gilding the lily; if the existing architecture is well-designed, the code will be fast, high-quality, easy to test, and efficient to execute. I recommend developing on a high-quality foundation, such as TanStarter (http://tanstarter.dev) or MkSaaS (http://mksaas.com)


TanStarter - Ship Faster with TanStack, Cost Less with Cloudflare

Source: https://tanstarter.dev/ Join Discord (3000+ members) (https://mksaas.link/discord)The complete TanStack Start boilerplate for building profitable SaaS, packed with AI, auth, payments, database, storage, email, newsletter, blog, dashboard, SEO and more, fully deployed on Cloudflare Workers

  • Private GitHub access
  • Lifetime updates
  • Unlimited projects

Time saved

Auth, roles and session handling18+ hrs

Stripe / Creem / Waffo payments and webhooks26+ hrs

D1 database schema and migrations14+ hrs

R2 file storage and user file flows14+ hrs

Transactional email and newsletter14+ hrs

AI text, chat, image and TTS demos26+ hrs

Landing, pricing, blog and legal pages30+ hrs

SEO, sitemap, OG images and i18n18+ hrs

Dashboard, admin users and API keys20+ hrs

Cloudflare Workers deployment setup20+ hrs

Setup work you can avoid200+ hrs

What you get

A production SaaS foundation, not just another landing template

TanStarter packages the repeated parts of every SaaS so your first week goes into product logic, customer workflows and launch polish instead of vendor setup.

Private GitHub repositoriesLifetime updatesUnlimited projects

TESTIMONIALS

What our customers are saying

Showcases

What our customers are building

Launch your SaaS with TanStarter and get featured here for free!

MkImage ### MkImage Make any images possible — generate, edit, and transform images with AI models like GPT Image, and Gemini in seconds. (https://mkimage.ai/)HuntifyAI ### HuntifyAI Discover the best AI tools, models, and alternatives, hunted daily. Browse AI for writing, image, video, and more on HuntifyAI. (https://huntifyai.com/)Nbility AI Gateway ### Nbility AI Gateway Nbility provides teams with unified access to mainstream models like OpenAI, Claude, Gemini, centrally handling authentication, quotas, billing, logs, and multi-device synchronization. (https://nbility.dev/)CosyVoice ### CosyVoice Discover CosyVoice, a multilingual zero-shot TTS model, featuring natural pronunciation and high-quality speech synthesis with advanced features. (https://cosyvoice.org/)LetsFindDomain ### LetsFindDomain Check domain availability across 50+ extensions instantly. AI name suggestions, price lookup, WHOIS & competitive research — free. (https://letsfinddomain.com/)ImgEnhancer ### ImgEnhancer Enhance image quality online with AI. Upload a photo to sharpen details, unblur images, upscale resolution, and download clearer results in seconds. (https://imgenhancer.co/)

Newsletter

Join the community

Subscribe to our newsletter for the latest news and updates

Indie Fox (@indie_maker_fox): Recently, I’ve looked into Matt’s skill set and experienced the development workflow using this skill set. The main workflow is as follows:

1、grill-me / grill-with-docs: Before starting, interrogate you first; some details need to be thought through and confirmed before writing code. There will be many questions, so generally just follow the suggestions.

2、to-spec and to-tickets: For large tasks, it’s recommended to run these two skills to generate spec

Similar Articles

@aronhouyu: https://x.com/aronhouyu/status/2063561548145275255

X AI KOLs Following

Introduces an open-source repository called awesome-codex-skills, which contains thousands of preset skills for Codex (as well as Claude Code, Gemini CLI, etc.), covering development, data, collaboration, and other scenarios. It also provides installation and usage guides to help users reuse workflows.

@php_martin: Codex Parallel Development Practical 2026|Full Video Edition. You think AI coding tools have reached their peak? Codex directly lets me simultaneously develop 6 projects in parallel: from iOS App, landing page, investor deck, to Remotion release video + social media automation...

X AI KOLs Timeline

Martin shares a practical workflow using Codex to simultaneously develop 6 projects in parallel, covering serial instruction parallel execution, custom Skill/Plugin reuse, and collaboration with Claude, significantly improving multi-task development efficiency.

@LukeDing14: Many people on X still don't know how to use Codex to complete projects. They find it strange that their Codex cannot write high-quality code or organize projects more efficiently. A month ago, the popular learn-harness-enginnering resource already covered many tips. Source link here ht…

X AI KOLs Timeline

LukeDing14 points out that many people still don't use Codex correctly for projects, and recommends the learn-harness-engineering resource along with his own harness file, sharing tips for efficient coding and project organization.

@AI_jacksaku: https://x.com/AI_jacksaku/status/2068702722807771294

X AI KOLs Timeline

This article compiles 10 high-value Skills essential for Codex, covering planning, code quality, frontend implementation, automated testing, CI/CD, security, and external tool integration, and provides installation methods and recommended combinations to help users improve development efficiency.