code-review_skill

Overview

This skill performs focused code reviews for a WeChat chat auto-reply assistant, emphasizing Rust/Tauri backends, React frontends, database usage, and performance risks. It surfaces logic errors, regression risks, and interface mismatches, then produces prioritized, actionable feedback with file locations. Reviews aim to reduce runtime bugs, unsafe patterns, and UI inconsistencies quickly.

How this skill works

The skill scans code for critical patterns: Tauri command annotations, Specta type export consistency, explicit SQL column usage, and common frontend anti-patterns like console.* or casting to any. It runs checklist validations (lint/clippy, formatting) and inspects performance issues such as N+1 queries, unnecessary cloning, and repeated computations. Output lists issues sorted by severity and includes concrete fixes and exact file/line hints.

When to use it

• Before merging changes that touch Tauri commands or backend RPC surfaces
• When adding or modifying API types shared between Rust and the frontend
• Prior to database schema or query changes that affect transactions or performance
• When frontend UI tables or compact layout rules may be violated
• During PR review to catch regressions, lint warnings, or unsafe casts

Best practices

• Prioritize logical regressions and type mismatches over stylistic nitpicks
• Ensure #[tauri::command] and #[specta::specta] are present for exposed functions
• Use explicit column lists with sqlx::query_as! and validate transactions
• Avoid console.* and as any in frontend code; prefer typed command responses
• Run cargo clippy --all-targets --all-features -D warnings and frontend linters before review

Example use cases

• Detect a missing specta export on a new Tauri command that breaks type generation
• Identify an N+1 database query introduced in a message history endpoint
• Find frontend table padding or line-height deviations that break compact layout
• Spot an unsafe any cast in React that could hide a runtime error
• Recommend converting repeated expensive computations into memoized results