Code Review Prompt Templates
Get actionable feedback: bugs, risks, and missing tests
1) Review a PR diff (bugs + tests + risks)
Act as a senior engineer doing a code review.
Input:
- PR goal: [GOAL]
- How tested: [COMMANDS]
- Diff (or changed files):
[PASTE DIFF]
Output:
- Findings ordered by severity
- Any missing tests
- Potential regressions
- Suggested follow-ups
When reviewing pull requests, check that cited file paths and line references match the actual code and the latest diff. Identify the conditions under which the issue appears and what it does when triggered. Clearly separate checks that actually run (with their results) from suggested tests or unverified claims. Mark any missing context as unknown rather than guessing. For readers who need more detail, see the full AI code review workflow. Remember: a single review finding alone is not proof of correct implementation.
2) Security-focused review
Review this code for security issues.
Focus on:
- auth/authorization mistakes
- injection risks
- secrets handling
- data validation
- unsafe defaults
Code or diff:
[PASTE]
3) Performance review
Review this change for performance risks.
Focus on:
- algorithmic complexity
- N+1 queries / repeated calls
- unnecessary re-renders (frontend)
- caching opportunities
Code or diff:
[PASTE]
4) Review for maintainability
Review this module for maintainability.
Focus on:
- clarity and naming
- separation of concerns
- error handling and observability
- public API design
Code:
[PASTE]
5) Test gap review
Given this feature change, list the test cases that should exist.
Feature:
[DESCRIBE]
Changed code/diff:
[PASTE]
Return:
- A prioritized list of tests to add
- Why each test matters
FAQ
What input produces the best reviews?
A diff or changed files, plus context about expected behavior and how it was tested.
How do I get less generic feedback?
Ask for file/line references, severity ranking, and concrete follow-up steps.
What should a review always include?
Behavioral risks, security concerns, and missing tests.