Skip to content

DirectoryPrompts

Claude Opus 5.5

Security Code Review – Security code review prompt for ChatGPT and Claude

Freewith a Recaply account
  • Free with an account
  • Copy filled in or download
  • Updates included
votes views

Claude Opus 5.5

This prompt reviews your code for security holes. Paste what it does and who can reach it, your stack and the code; you get every vulnerability rated by severity, how it is exploited and a fix.

What you get

  • A short risk summary that names the most serious problem
  • A table of every vulnerability with its severity, location, CWE number and who can exploit it
  • How each attack works and a fix written for your stack
  • Hardening ideas, kept apart from the real findings

How to use it

  1. 1
    Say who can reach the code

    State whether it is public, behind a login or staff only, and what each user may see. The severity of every finding depends on it.

  2. 2
    Fix Critical and High first

    Work down the findings table, and look up each Needs check item in your middleware and config.

  3. 3
    Retest each fix

    Reproduce each attack in your own test setup, apply the fix, and confirm the same request now fails.

Better results

  • Spell out the access rules, such as "each account sees only its own invoices". Without them, the AI cannot tell a missing ownership check from intended sharing.
  • Paste the login middleware and session config too, so Needs check items turn into real findings or drop off the list.
  • Treat this as a first pass. For code that handles payments, logins or personal data, add an automated scanner and a review by whoever owns security.

Questions

Can ChatGPT find vulnerabilities in code?

Yes, ChatGPT finds many common vulnerabilities in code you paste, such as SQL injection, missing ownership checks, path traversal and cross-site scripting. It misses problems in code it cannot see and can raise false alarms, so confirm each finding with a test before you rely on it.

How do I use Claude Code for a security review?

Run Claude Code in your repository and give it this prompt, naming the routes or folder to review instead of pasting code. It can then read the middleware and config itself. Claude Code also has a built-in /security-review command for the changes on your current branch.

Is an AI security review enough on its own?

No. An AI review is a fast first pass that catches common flaws, but it can miss issues that span files or depend on runtime settings. Pair it with an automated scanner in your build and a human review for code that handles money, logins or personal data.

Updated 6 Oct 2026

Reviews

No reviews yet
Reviews open soonThey will come only from signed-in people who used this prompt.

Similar prompts

More coding prompts

Image prompts

Video prompts

Trending prompts

Newest prompts

Browse all prompts