A collection of AI-powered code review personas modeled after prolific open-source builders. Each persona captures the review style, priorities, and principles of a specific maintainer, distilled from their real PR review history.
Reviewers
| Name | GitHub | Repository |
|---|---|---|
| Andrew Gallant | BurntSushi | BurntSushi/ripgrep |
| Charlie Marsh | charliermarsh | astral-sh/uv |
| David Tolnay | dtolnay | serde-rs/serde |
| Edward Yang | ezyang | pytorch/pytorch |
| Jarred Sumner | Jarred-Sumner | oven-sh/bun |
| Jordan Liggitt | liggitt | kubernetes/kubernetes |
| Woosuk Kwon | WoosukKwon | vllm-project/vllm |
Project Structure
code-reviewer-personas/
├── README.md
└── <reviewer>-review/ # One directory per persona
├── SKILL.md # Actionable review skill — the prompt used to
│ # conduct reviews in the reviewer's style
├── PRINCIPLES.md # Standalone core principles and patterns
│ # distilled from the reviewer's history
└── SOURCES.md # Links to all PRs analyzed to build the persona
Each persona directory contains three files:
- SKILL.md — The complete review skill definition. Contains the system prompt and instructions that drive an AI review session in the style of the reviewer.
- PRINCIPLES.md — A self-contained summary of the reviewer's core principles, common feedback patterns, and areas of focus.
- SOURCES.md — A full list of the pull requests that were analyzed to construct the persona, providing traceability back to real reviews.
Author's Note
Inspired by these engineers' capabilities, I wondered, "what if I could have some of these people reviewing my PRs to improve my code?" While I haven't used the skills to review code yet, I've found the specific examples, principles, and processes to be interesting enough already.
If there's any other people you would like to suggest or are curious about, please feel free to suggest!
License
MIT