It feels increasingly backwards that PR reviews are where we catch basic bugs, style issues, and convention violations.
With AI writing a large chunk of code now, the volume of small, avoidable mistakes has gone up. Yet we still wait until PRs to point out things that could have been flagged instantly, pre-commit, on the author’s machine.
PR review should be about design, correctness, and tradeoffs — not “missing await”, inconsistent patterns, or obvious slop.
Is there a good reason pre-commit review isn’t the norm, or are we just accepting an inefficient workflow because it’s familiar?
Linters are great for style and syntax, but logic errors still rely heavily on human review.
Do you think a linter for code logic - for example: risky condition changes, missing error handling, inconsistent early returns, or unsafe null/undefined assumptions - would be useful in practice if it ran at lint-stage or commit time?
I have found that PR based code reviews are two noisy and non configurable. Also unpredictable as to when they will run or if all is ok. Also it is never clear who’s problem it is (more like no one is intrested in dealing with it or it is just ignored totally)
So I created this tool that runs on commit. It is easily configurable and bypassable.
Would love to hear what community thoughts are on this idea.
You can also add custom prompts, choose context option and severity.