Request a code review with AI
Give a separate reviewer a defined change range, the original requirements and enough implementation context to produce feedback the team can act on.
To request a code review with AI, tell the reviewer what the change was meant to achieve. A diff alone cannot reveal that a checkout change was supposed to preserve the behaviour of returning customers. The requesting-code-review skill prepares a focused handoff to a separate reviewer agent using a supplied template.
It includes the starting and ending commits, a concise description of the work and the plan or acceptance criteria. Findings can then be sorted by severity. A disputed suggestion can go back to the code and tests rather than becoming a decision by authority.
What it does
- Identify the exact commit range to review.
- Summarise the intended behaviour independently of the implementation.
- Send a scoped brief to a compatible reviewer agent.
- Distinguish critical findings from lower-priority improvements.
- Check contested feedback against relevant code and test evidence.
Install and use
- Install the skill and its reviewer template in a host that supports delegated review. Provide access to the Git repository, the two commit references and the requirements.
- Ask, “Review the invoice change between these commits. Check how duplicate submissions behave and whether the existing success path is preserved.”
- Read the findings with the implementation in view. Address important issues and verify fixes before treating the review as complete.
FAQ
Source
Made by Jesse Vincent and Prime Radiant. Original skill: requesting-code-review, published on GitHub under the MIT licence.
View on GitHub