production-check
Skill1 starsChanged 2 days agoMITView on GitHub
Checks a Next.js App Router project, and a NestJS API beside it, before real users arrive. A read-only script finds what a scan can see (committed env files, public secrets, loose versions, lockfiles, error and not-found pages, robots, sitemap, CORS, security headers, rate limits, validation, schema sync, leftover logs), then the steps walk through what needs judgment. Use before a first launch, after a large AI-written change, or when asked to fix a vibe-coded app or make it production ready.
In this repository10 other files it gives its agents.
- add-admin-dashboardSkill
- app-screenshotsSkill
- design-rulesSkill
- docs-checkSkill
- i18n-keys-checkSkill
- nextjs-claude-mdSkill
- nextjs-hooksSkill
- rtl-checkSkill
- template-rebrandSkill
- vps-deploySkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 11 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Read-only until step 4. Report before you change anything.
- Very specific: line 15
Check: the last line reads N fail, N warn, N info, N ok, and the apps: line names every Next.js app and NestJS API in the repository.
- Distinctive: line 24
Check: every FAIL and WARN has a verdict and a one-line reason with a file path.
- Distinctive: line 28
Write pass, fail or n/a for each, with the file or command that proves it.
- Distinctive: line 32
Check: name the guard or the line for each one.
- Distinctive: line 34
Check: one request with another user's id is refused (test or curl).
- Distinctive: line 38
Check: a burst of quick attempts ends in 429.
- Distinctive: line 40
Check: an unsigned request is refused.
- Distinctive: line 48
Check: the heaviest image on the home page and its size.
- Distinctive: line 50
Check: view the source of one page.
- Distinctive: line 52
Check: go through the main flow once with the keyboard only.
- Very specific: line 59
Check: the line is gone or has its by-design reason, and the builds pass (next build, and the API's build).
- Distinctive: line 65
Check: every item marked unsure ends in a question for the user.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Production check
Read-only until step 4. Report before you change anything.
1. Run the scan
node ${CLAUDE_SKILL_DIR}/scripts/production-check.mjs . --quiet
Check: the last line reads N fail, N warn, N info, N ok, and the apps: line names every
Next.js app and NestJS API in the repository. If one is missing, run the script from its folder.
2. Give every FAIL and WARN a verdict
Open each file a line names and sort the line into: true problem, by design, or false positive. The script cannot know intent: a demo password that is public on purpose, a controller with no guard that serves a public catalogue, a startup log line.
Check: every FAIL and WARN has a verdict and a one-line reason with a file path.
3. Go through what a script cannot see
Write pass, fail or n/a for each, with the file or command that proves it.
- Auth on the server: every route handler, server action and API route that reads or writes user data checks the session and the role itself. A hidden button is not a check. Check: name the guard or the line for each one.
- Ownership: a user reads and changes only their own rows. Look for queries by id alone. Check: one request with another user's id is refused (test or curl).
- Validation: every body, query and form goes through a schema on the server (DTO with class-validator, or zod). Check: an unknown field and a wrong type are both refused.
- Rate limits: sign-in, sign-up, password reset, anything that sends email or calls a paid API. Check: a burst of quick attempts ends in 429.
- Webhooks: payment webhooks verify the provider's signature on the raw body. Check: an unsigned request is refused.
- Errors and logs: responses carry no stack traces or SQL; logs carry no tokens, passwords or reset links, including when NODE_ENV is unset. Check: read every logging call that prints user data.
- Database: schema changes go through migrations or a reviewed command; backups run; one restore was tried. Check: name the backup and the date of the last restore test.
- Headers on the live site: the proxy or CDN may set them where the code does not.
Check:
curl -sI https://your-domain | grep -iE 'strict-transport|content-security|x-frame|x-content-type|referrer' - Images: next/image or width, height and lazy loading by hand; uploads are resized. Check: the heaviest image on the home page and its size.
- SEO: every public page has a title, a description and a canonical on the real domain. Check: view the source of one page.
- Accessibility:
langon<html>in the served HTML, labels on inputs, visible focus. Check: go through the main flow once with the keyboard only. - Staging and monitoring: a staging copy with its own database and keys, an uptime check on the health route, and a person who gets the alert. Check: name all three.
4. Fix one problem per change
Fix only what the user agreed to, smallest change first, then run the scan again.
Check: the line is gone or has its by-design reason, and the builds pass (next build, and the
API's build).
5. Report
A table: item, verdict, file, fix or next step, with the scan's summary line before and after. Check: every item marked unsure ends in a question for the user.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

