Skip to main content

Reviews that catch what a quick read misses.

Type what you want checked. The Code Reviewer preset shapes a structured prompt that asks any AI for a disciplined, line-aware review.

100% on-device 20-agent pipeline ~100 seconds
The problem

A vague request like "review my code" produces a vague answer. The model skims, praises the obvious, and overlooks the edge cases, security gaps, and concurrency bugs that actually break production. You spend more time re-prompting than reviewing.

What the Code Reviewer preset does

The Code Reviewer preset pre-tunes the pipeline to produce prompts that demand a methodical pass: correctness, security, performance, readability, and test coverage. Every generated prompt fixes a clear role, scope, and output format so the review is consistent and actionable.

From one sentence

Type a sentence. Get an expert prompt.

You type

Review my Python function that parses user-uploaded CSV files.

Meta Prompt Lab generates
Role: You are a senior software engineer conducting a thorough peer review of a Python function that parses user-uploaded CSV files.

Context: This function handles untrusted input in a production web service. Uploaded files may be large, malformed, or crafted to abuse the parser. Treat all input as hostile.

Task: Review the function I provide and assess it across five dimensions: correctness, security, performance, readability, and test coverage. For each issue, cite the relevant line or block, explain the concrete risk, and propose a specific fix with corrected code.

Constraints:
- Prioritise findings by severity (critical, major, minor).
- Flag any unbounded memory use, injection vectors, or unhandled encodings.
- Do not rewrite the whole function unless a structural flaw requires it.
- Assume Python 3.11 and the standard library unless I state otherwise.

Format: Return a severity-ordered list. Each entry: [Severity] Line ref — Issue — Risk — Suggested fix. Close with a short summary and three test cases I should add.
Inside the preset

Tuned for code reviewer work.

Severity ranking

Every generated prompt asks the AI to grade findings critical, major, or minor so you fix what matters first.

Security focus

Built to surface injection vectors, unsafe input handling, and dependency risks rather than cosmetic notes.

Line-level citations

Prompts insist on concrete line or block references, not hand-wavy commentary.

Concrete fixes

Each issue must come with corrected code or a precise remediation step.

Test suggestions

Prompts close by requesting the test cases that would have caught each bug.

How it works

Three steps, about 100 seconds.

1

Type one sentence

Describe the code or change you want reviewed in plain language.

2

20 agents refine

A pipeline of 20 on-device specialists shapes a rigorous review prompt in about 100 seconds.

3

Copy into your AI

Paste the finished prompt into ChatGPT, Claude, Gemini, or Copilot and get a structured review.

FAQ

Code Reviewer preset questions.

Does it review my code directly?+

No. Meta Prompt Lab generates the review prompt. You paste it, along with your code, into the AI assistant of your choice.

Which languages does the Code Reviewer preset support?+

Any language. The generated prompt adapts to the code you describe, from Python and Go to Swift and Rust.

Will the prompt stay private?+

Yes. The whole pipeline runs on-device using Apple Intelligence, so your sentence and the generated prompt never leave your iPhone.

Can I tune the review depth?+

Yes. Describe the depth you want in your sentence, for example a quick security-only pass, and the pipeline shapes the prompt accordingly.

How many prompts can I generate for free?+

Five runs a day on the free tier. Pro removes the limit at £4.99 a month, £29.99 a year, or £59.99 for lifetime access.

More presets

Explore related roles.

Build your code reviewer prompts on-device.

Free 5 runs a day. Pro from £4.99/mo, £29.99/yr, or £59.99 once. Nothing leaves your iPhone.

Works with ChatGPT, Claude & Gemini No account needed iOS 26