One prompt for your coding agent. You get a report with a score, a maturity stage and open items: What's missing before you go live? What can you fix yourself — and where does support make sense?
You need a coding agent with access to your project — Claude Code, Cursor, Windsurf, GitHub Copilot or Codex. Without full project access, the report is correspondingly limited. The check runs in your agent — whether your code goes to its model provider is decided by your agent settings. Nothing reaches us unless you want it to.
Simplified example with 64 equally weighted checks: 16 met, 19 partial, 26 missing, 3 not verifiable. Your report covers all 90 checks with evidence, effort estimates and an assessment of who should close each open item.
Met counts 1 point, partial 0.5, missing 0. Critical categories count double.
Score = weighted points earned ÷ weighted checkable points × 100. Anything not verifiable is left out.
The math is in the report. You can recalculate it.
The check places your project in one of five maturity stages and lists the open items for the next step.
Most checkable requirements are still open.
Some are met; significant gaps remain.
At least half of the weighted points reached; open risks remain.
Most checkable requirements are met.
The checkable requirements are met. Open evidence is listed alongside.
A blocker caps the maturity stage at Proof of Concept, even with a high score — for example, proven unprotected access to other people's data. The check finds indications in the code; it does not replace a full security audit.
The check is a prompt, not a tool. No account with us, no installation, nothing to upload. You work with the coding agent you already use — its costs stay yours.
In the coding agent you already work with. The agent leaves your code unchanged and saves the report as a file in the project. Without write access, it prints the report in the chat.
You can answer twelve short context questions, but you don't have to. Anything you leave open, the agent only fills in when it can justify it from the project — and marks it as an assumption.
A PRODUCTION_READINESS.md in your project: score, maturity stage, risks, estimated service costs — and a to-do list you can hand straight back to your agent. Plus a PRODUCTION_READINESS.html to view it: from there you can share your result on LinkedIn or send us the report with one click.
The categories are weighted for your project: sensitive or unclear personal data make data protection count double — even for internal tools. The weights that apply are stated in the report.
Credentials in the code or the browser bundle, missing database access rules (RLS), publicly reachable file storage, unvalidated input.
Timeouts, idempotency, backups with a tested restore, clean error behaviour.
Checks whether tests cover the core logic: payments, permissions, data processing — not just helper functions.
Reproducible build, separate environments, rollback in minutes, kill switch.
Structured logs without personal data, error tracking, alerts that actually reach a human.
Suitable database indexes, pagination, queues for long-running tasks, reused database connections.
Where personal data lives, how it gets deleted, how long it is kept, which contracts exist with service providers.
Manipulated inputs, limited tool permissions, capped agent runs, repeatable quality tests, cost brakes.
Can a new developer find their way around quickly? Is there duplicated logic that can drift apart?
Licences, legal notice, NIS2 applicability, EU AI Act.
For every gap you get an assessment: do it yourself, have it reviewed or bring in specialists. Which one fits also depends on your experience.
Mechanical, well understood, low risk. The report estimates the effort for every single item.
You can build these — but have them reviewed so they also hold up under errors and unexpected input.
Needs experience and someone who takes responsibility. Each item comes with one line on why experience or clearly assigned responsibility is needed.
↳And then the honest part: You can tackle the 🛠️ items yourself — each one comes with an effort estimate. The 🤝 items are where we come in — or send us the report and we'll tell you which of them matter for your project.
The complete prompt, no sign-up. Answer the context questions in English, and your report comes in English.
## Context profile Answer what you know after the arrows. Leave anything blank. 1. What does this software do, in one sentence? → 2. Who uses it? (internal team · business customers · the public · API) → 3. Where is it today? (prototype · pilot with real users · production) → 4. Personal data it handles: (none · basic · sensitive · not sure) → 5. Sector: (regulated or critical · not regulated) → 6. AI components: (none · LLM features · agents that take actions) → 7. Expected scale in 12 months: (under 100 · thousands · 100k+ · unknown) → 8. Availability needed: (best effort · business hours · 24/7) → 9. Who runs it after launch? (solo · small team, no ops · on-call team) → 10. Hosting constraints: (any cloud · EU-only · on-premise required) → 11. Money: does it handle payments, invoices, anything with a price? → 12. The one thing that must never happen → ## Instructions for the agent You are a senior production-readiness auditor. Audit the codebase in this workspace the way BReact's engineers do before they take a system live. ### Procedure — not optional Work through Steps 1 to 9 in order. Do not begin writing the report before Step 8 is finished. Print the GATE block first. ### Ground rules - Read-only. Analyze everything, modify nothing. - Evidence or nothing. Every mark cites a real file. - ❓ is a real answer: excluded from the score, carried into "Questions only you can answer". - Honest, never mean. A clean bill of health is a valid result. - Don't sell. Classify each gap by who can realistically close it. - Shareable output: no secrets, no personal data, no hostnames. […]
🔒Nothing reaches us. The check runs in your agent, with the provider of your choice — whether your code goes to its model is decided by your agent, not by us. We see none of it unless you send us the report at the end. The prompt instructs your agent to leave out credentials, personal data and internal addresses. Still, review the report before you pass it on.
We read your report and give you a second assessment: what's urgent, what fits your project, what's still open. Free, no strings attached, no sales call.
The quickest way is from the report: open PRODUCTION_READINESS.html and click “Contact BReact” — your full report will already be in the contact form.