Conversation
Configures the automated code review to focus on real bugs and correctness issues, with explicit severity levels and a false-positive exclusion list tailored to this Rust/Web3 codebase.
Gives Claude real Rust code intelligence (type checking, diagnostics, go-to-definition) during PR reviews.
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a new REVIEW.md file, which establishes comprehensive code review guidelines for the repository. The guidelines cover review priorities, specific checks, items to ignore, severity classifications, and formatting for review comments. A review comment suggests an improvement to the newly added REVIEW.md by recommending that the defined severity levels (Bug, Nit, Question) be updated to align with standard GitHub API severity levels (Critical, High, Medium, Low) for better consistency and interpretation.
REVIEW.md
Outdated
| ## Severity | ||
|
|
||
| - **Bug** — will cause incorrect behavior; must fix before merge | ||
| - **Nit** — minor improvement, not blocking | ||
| - **Question** — unclear intent, needs author clarification | ||
|
|
||
| If unsure whether something is a real issue, mark it as a **Question** rather | ||
| than asserting a bug. |
There was a problem hiding this comment.
The severity levels (Bug, Nit, Question) in this document are inconsistent with the standard GitHub API severity levels (critical, high, medium, low). This mismatch can lead to miscategorization or misinterpretation of review findings by the system.
| ## Severity | |
| - **Bug** — will cause incorrect behavior; must fix before merge | |
| - **Nit** — minor improvement, not blocking | |
| - **Question** — unclear intent, needs author clarification | |
| If unsure whether something is a real issue, mark it as a **Question** rather | |
| than asserting a bug. | |
| ## Severity\n\n- **Critical** — will cause incorrect behavior; must fix before merge\n- **High** — significant issue that should be addressed soon\n- **Medium** — improvement that should be considered for future work\n- **Low** — minor improvement or stylistic suggestion\n\nIf unsure whether something is a real issue, mark it as **Low** or **Medium** rather\nthan asserting a critical or high severity bug. |
| drops valid quotes | ||
| - New `async` code does not block the Tokio runtime (no blocking I/O, no | ||
| `std::thread::sleep`, no heavy computation without `spawn_blocking`) | ||
| - Database queries include appropriate indexes and won't degrade at scale |
There was a problem hiding this comment.
Given that these operations require actual review of the DB query and planning, maybe. it should first request the DB plan from before and after so it can perform an actually useful review
| Review in this order. Stop at the first category that has findings — do not | ||
| bury critical bugs under a wall of style nits. |
There was a problem hiding this comment.
I get the idea about stopping at the first issues, but it might lead to a bunch of back and forth
| - File path and line number | ||
| - What is wrong (one sentence) | ||
| - Why it matters (one sentence) | ||
| - Suggested fix (code snippet if non-obvious) |
There was a problem hiding this comment.
Sometimes the fix is kind of retarded. I'm not sure how to improve it but maybe the suggestion should be optional
| - Suggested fix (code snippet if non-obvious) | ||
|
|
||
| Do not summarize the PR. Do not list what looks correct. Only report findings. | ||
| If there are no issues, say so in one sentence. |
There was a problem hiding this comment.
Maybe using an example:
* If you do not find issues, simply comment: LGTM 🤖
| https://github.com/anthropics/claude-plugins-official.git | ||
| plugins: | | ||
| code-review@claude-code-plugins | ||
| rust-analyzer-lsp@claude-plugins-official |
There was a problem hiding this comment.
Did you double check that this is runnable given the possible lack of the right rust version in the runner?
2 participants
Summary
Adds Claude Code GitHub integration with two workflows, review guidelines, and Rust tooling.
Workflows
claude.yml— Responds to@claudementions in PR/issue comments. Claude analyzes the context and executes the request.claude-code-review.yml— Automatic code review when a PR is marked ready for review. Uses the officialcode-reviewplugin with multi-agent verification. Does not run on drafts, pushes, or reopens. Includesrust-analyzer-lspplugin for real Rust code intelligence (type checking, diagnostics, go-to-definition) during reviews.Review guidelines (
REVIEW.md)Controls what Claude flags during reviews:
Security
Test plan
@claudemention triggers in a test PR comment