my-git

AI Reviewer and Human Reviewer

English

AI Reviewers are suitable for finding repetitive problems, checking for omissions, and generating review checklists.

Human Reviewers are responsible for judging intent, boundaries, long-term maintenance costs, and whether a merge is permissible.

AI Reviewer Suitable for Checking

Whether the diff is too large
Whether there are irrelevant changes
Whether there are obvious bugs
Whether there are test gaps
Whether there are security risks
Whether there are style inconsistencies

Human Reviewer Responsible for Judging

Whether this requirement should be implemented this way
Whether semantic boundaries are correct
Whether an abstraction is worth introducing
Whether risks are acceptable
Whether merging is allowed

Reusable Prompt

Please act as an AI reviewer and review the current git diff.
List only specific risks, sorted by severity.
Each entry includes file location, problem, and suggestion.
Do not output vague praise or summaries.

Collaboration Principles

AI Reviewers can increase check coverage, but they cannot replace human owner judgment.

Recommended process:

The author self-reviews the diff first.
The AI Reviewer performs the first round of risk scanning.
The author addresses clear problems.
The Human Reviewer makes the final judgment.
Merge only after CI and necessary manual verification pass.

Cooperation with CODEOWNERS

The AI Reviewer can perform a general risk scan first, and CODEOWNERS can then assign the PR to the person who truly understands that directory.

Recommended combination:

AI reviewer -> author self-fix -> code owner review -> CI -> merge

This can reduce low-value checks for reviewers and save human attention for semantics, boundaries, long-term maintenance costs, and deployment risks.

Extended Reading

This site is open source. Improve this page.