Review your own pull request the way you personally tend to review code before asking a teammate.
Preview what a specific teammate would likely flag on a pull request before requesting their review.
Build a written profile of a reviewer's habits, tone and common concerns from their past comments.
| mainframecomputer/skills | agentigris/agentfund | alange/tinyclaude | |
|---|---|---|---|
| Stars | 5 | 5 | 5 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | easy | hard | moderate |
| Complexity | 2/5 | 5/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Install with npx skills add, then needs GitHub access to read past review comments.
This repository holds agent skills used by a team called Mainframe, meant to be installed into Claude Code with a single command. The main feature is a skill called review as, which reviews a pull request the way a specific person on your team actually would, instead of giving a generic code review. It works by privately reading through a person's past year of GitHub review comments and pull request comments to learn their style, then using that learned style to review a new pull request. You can run it as yourself, to catch the things you personally tend to flag, or run it as a teammate, to see what they might say before you ask them to review it for real. The README includes a full example profile built this way, showing the kinds of things that reviewer tends to push back on, such as whether a piece of code is actually needed, whether error handling is hiding a real problem, whether a change fits the existing patterns in the codebase, and whether comments look like they were generated by an AI rather than written by a person. It also captures tone: short messages, phrased as questions, informal language, and comments that almost always end in approval even when they raise real concerns. The profile format itself is a markdown file with some metadata at the top, such as which account it covers, how many comments were examined, and what date range the evidence spans. It also notes limitations honestly, for example if there is not enough history to know what a hard rejection from that person would look like. This is useful for teams that want pull request review to sound like a specific colleague's real habits rather than a boilerplate checklist, and for individuals who want a second look styled on their own past reviewing patterns before sending a pull request out.
A Claude Code skill that learns a specific teammate's real code review style from their GitHub history and reviews pull requests in that voice.
Mainly TypeScript. The stack also includes TypeScript, Claude Code, GitHub.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.