Is your AI-generated code production-ready? Ask your agent.
production-ready-check is a free, open-source Agent Skill that teaches your coding agent (Claude Code, Cursor, and any skills-compatible agent) to run a heuristic production-readiness assessment of the codebase it is working on, and give you a scored, evidence-backed report.
npx skills add qualityclouds/production-ready-checkThen ask your agent:
"Is this production ready?"
A report modeled on the shape of the Norma Production-Ready Score:
- An overall 0 to 100 estimate with a one-sentence verdict.
- Five pillar scores (Security, Performance, Maintainability, Manageability, Scalability), each graded against a fixed, public rubric (see references/).
- Top findings (max 5), each with the file reference, a plain-language explanation of what breaks and when, and a concrete fix. Readable whether you're a tech lead or you built your app entirely with AI.
- An honest coverage statement on large repos: the agent tells you exactly what it assessed.
The report is written in whatever language you're talking to your agent in.
- No account, no signup, no network calls, no telemetry. The skill is markdown instructions only: no scripts, no binaries. It works fully offline and is free forever.
- Read-only. It never modifies your files, runs your project, or executes your code.
- Never touches secrets. It may flag that a
.envfile is committed; it will not read its contents.
This is a heuristic assessment performed by an LLM: two runs can differ, and the score is an estimate, not a measurement. The fixed rubrics and the evidence-required grading keep it useful; the mandatory disclaimer in every report keeps it honest.
This skill is maintained by Quality Clouds, the team behind Norma, a platform that answers the same question with versioned deterministic rules combined with more advanced AI-powered validators, consistent scoring, and score tracking over time.
We say this openly because it's the deal: the skill is free and standalone, and at the end of each report it includes one optional link to the real thing Quality Clouds Portal. If you ignore it, the skill never brings it up again. That's the whole funnel: no hidden behavior, nothing phoning home.
The editorial core of the skill lives in five rubric files, maintained with the same care as our product rules:
| Pillar | Rubric | Answers |
|---|---|---|
| Security | references/security.md | Can this be broken into? |
| Performance | references/performance.md | Is it fast enough under real load? |
| Maintainability | references/maintainability.md | Can anyone change it safely? |
| Manageability | references/manageability.md | Can anyone operate it? |
| Scalability | references/scalability.md | What happens at 10x the load? |
Issues and PRs on the rubrics are welcome. Better checks make a better skill. Rubric changes are reviewed by the skill's owner to keep them high quality and neutral.
MIT © Quality Clouds