Build a small, reusable review checklist from the change type and your own repository rules. The generator runs in this browser: it does not send code or checklist text to a service. Its prompts are a starting point; review the diff, run project checks, and keep required human review.
Create a review checklist
Describe the change briefly, choose its closest type, and add exact repository rules the reviewer must check. Do not paste source code, credentials, or private data.
What should every generated checklist ask?
- Compare the diff with the task's stated outcome and acceptance criteria.
- Check the applicable repository rules and verify that each stated invariant still holds.
- Look for a concrete case that should trigger a finding, a safe counterexample, and an unrelated change that should not trigger one.
- For each finding, cite the changed location, the rule or expected behavior, and a specific consequence. Omit speculative comments.
- Run the relevant project tests and checks, inspect the final diff, and keep human review and repository protections in place.
How is this generator grounded?
OpenAI recommends concise, scoped Codex review rules that describe a consequential invariant and a safe path. Its guide suggests trying an intended violation, a safe counterexample, and an unrelated change. A September 24, 2026 Codex issue describes a user's attempt to derive a reusable checklist from existing repository rules; it is one qualitative report, not evidence of frequency or general model behavior. The change-type prompts here are practical suggestions, not official Codex requirements or a guarantee that review will catch defects.
Frequently asked questions
Does this tool upload my code or rules?
No. The checklist is assembled by JavaScript in this page. The fields do not make a network request; keep sensitive information out of them anyway.
Does generating a checklist make an AI review reliable?
No. Treat it as a prompt aid. Verify each finding against the code and requirements, run the relevant checks, and keep your normal human review and branch protections.
What evidence supports this tool?
OpenAI's documentation explains how to write and evaluate scoped review rules. The linked Codex issue is an individual report asking for a concrete reusable checklist. It is qualitative evidence of a user task, not a representative survey, benchmark, or proof of prevalence.
Sources · checked November 3, 2026
Related: Codex repository review rules and how to verify AI code review findings.
Methodology and limitations
The browser-only checklist combines general diff and verification prompts with change-type prompts and optional rules supplied by the reader. It sends no form data to a service and makes no claim that a checklist or AI review catches all defects. The official Codex review guide recommends concise, scoped repository rules and checking an intended violation, a safe counterexample, and an unrelated change. One public Codex issue is cited only as qualitative evidence of a request for reusable checklist help.
Affiliate placeholder {{AFF_TOOLNAME}} — referral link not configured.