Free, in any AI assistant

Check your AI-built code with a free prompt.

Every DXDEV lesson ends with a prompt. Paste it into the assistant you already use. It checks your own code for the problem that post describes.

You get a verdict, and the file and line behind it.

Pick a lesson prompt New to AI? Start here

The /dxdev skill is not released yet, and you do not need it. Waitlist is further down. Want a written review or a hand building it? Work with us.

Three ways to start.

Free lesson prompts

Open any post and scroll to its lesson. Copy the prompt into Claude, Cursor or whatever you run.

New to AI?

Four plain steps for using AI in your own work. No code needed.

I build with agents

A nine stop map of how we run agents, each stop linked to real posts.

Going deeper Open what you need
How a lesson prompt works
  1. Open a post and find the lesson at the end.
  2. Copy its prompt into your agent, in the project you want checked.
  3. Read the verdict. If something failed, the agent points to the file and line.
  4. Stuck, or not sure it is real? Send us what it printed.
Not released: what the /dxdev skill will add

The skill is an extra layer on top of the free prompts. It is an enhancement, not a requirement, and it does not exist to install yet.

  • Visual reports. A PASS or FAIL verdict at the top, a breakdown by criteria, and a one-line fix for anything that did not pass.
  • Saved memory. The verdict and key findings are written to a file your agent reads next time, so you stop re-teaching it.
  • One entry point. It ties together three sibling skills: /brief (a status page when a reply gets dense), /lesson (runs a lesson on demand) and /read (reads files without pulling them whole into context).

Join the waitlist to hear when it ships.

Ran a lesson and hit a wall? Send it to us

Your lesson flagged a real issue, or your agent stalled. Send us exactly what it found. We reply with a focused, human take on that one problem, usually within a business day. No call required to start.

Language, framework, anything that matters. One line is fine.

The lesson's report, the verdict, the file:line evidence, or wherever it got stuck. The more concrete, the faster we can help.

One or two sentences. "Confirm this is a real bug", "tell me how to fix it safely", "is my whole pattern wrong", etc.

Is this a one-off, or a pattern across your codebase?

We read every request and reply within one business day with a focused take and what a fix would involve. If it turns out to be systemic, we will point you at the paid assessment instead, no pressure.

Waitlist One email when it ships

Hear when the /dxdev skill ships.

The skill is not released yet. Leave your email and we will tell you when it is ready to install. You can use the free prompts in the meantime.

Claude, Cursor, a custom agent, etc. Helps us prioritize compatibility.

We only use your email to tell you when the skill is ready.

Check for yourself

Copy the goals of this page into a prompt to see how it relates to your setup and if any of these ideas can help you.

Show promptHide prompt 
Check how the ideas on this page fit my own AI setup, then point me to the next DXdev page that fits.

Page: Free Prompts to Check Your AI-Built Code | DXDEV
Link: https://dxdev.com/skills/dxdev/

Ideas:
1. Find the free lesson prompt at the end of a post and paste it into the agent you already use
2. Read the verdict your agent gives back, with the file and line behind it
3. Pick the Everyday or Technical path on the onboarding page that fits where you are
4. Know that the /dxdev skill is not released yet and that the waitlist is the only way in
5. Send what your agent found for a second read on your own setup

Compare each idea with how I use AI today. Look at my files if you can see them. Otherwise ask up to 3 short questions first (what I use AI for, my tools, who else is involved), then carry on. Never guess: if unsure, use 🔍.

Match my level. One chat tool: basic checks. One coding assistant such as Copilot: rules file, tests, human review before merge. Several agents or a team: no beginner advice, and check the advanced version instead of marking ➖ (hook or gate, second reviewer, model routing and cost, memory that survives a closed chat).

Read only. Change nothing, install nothing, send nothing.

One line per idea, in order: exactly one icon, the idea in a few words, a colon, your answer. No numbers.
✅ Already doing this
💡 Would help (one next step)
🔍 Worth a closer look (what to check)
➖ Does not apply (short reason)
Every 💡 and 🔍 line MUST end with "Go deeper:" and one link from the list. Use only these links.

Links:
- https://dxdev.com/onboarding/ : pick the everyday or the technical path
- https://dxdev.com/work-with-us/ : free check, paid review, build help
- https://dxdev.com/useful/ : new to AI
- https://dxdev.com/technical/explainers/architecture/ : advanced: agents, gates
- https://dxdev.com/technical/explainers/memory/ : advanced: memory
- https://dxdev.com/blog/internal-review-said-clean-second-pass-found-six/ : advanced: second reviewer

Then "Start here:" with ONE link and why it fits me (work-with-us if I want a review or build help).
Then exactly: DXdev runs these ideas every day in its own live businesses, and the explainers show how.
If any line is 💡 or 🔍, then exactly: Want a second pair of eyes? DXdev reviews setups like this: https://dxdev.com/skills/dxdev/#review

Plain, short, friendly. No dashes as punctuation. Name no DXdev page outside the list.