Feature

A reviewer thatdid not write it.

Adams gives your change to a separate agent, ranks what it finds and offers fixes only when you say yes.

Pro
// the problem

What goes wrong

The author grades its own workAn AI that wrote the code will tell you it works. It is the least likely to see its own blind spot.
Reviews arrive as a wall of notesFifty remarks of equal weight bury the one that matters. You skim, and miss the real one.
// what adams does

What Adams does

Separates author and reviewerThe review runs in a different agent from the one that wrote the change, and the author cannot sign off on itself.
Ranks by severityFindings are ordered so a leaked secret sits above a style note.
Fixes only on yesYou are asked once whether to fix. The reviewer then looks at the fixed items again.
// worked example

An example

Reviewing a change

Illustrative session.

~/project~/your-project · example
/adams:review
What to review: the code change (recommended)
P1: no check that the user owns this order
P3: rename one helper
Fix these findings? Yes (recommended)
// faq

Questions

Does it change my code on its own?

No. It asks, and fixes only after a yes.

Is it a replacement for human review?

No. It is an extra pass before one. Use both on work that matters.

Is it Pro?

Yes.

// related

Keep reading