AgentSkillsCN

code-review

使用多个专业代理，基于置信度评分的自动化代码审查，用于拉取请求，过滤假阳性

SKILL.md
--- frontmatter
name: code-review
description: Automated code review for pull requests using multiple specialized agents with confidence-based scoring to filter false positives
license: MIT
compatibility: opencode

Code Review Skill

Automated code review patterns for pull requests with confidence-based scoring to minimize false positives.

What I Do

I review pull requests by examining:

  • •Code Quality - Readability, maintainability, consistency
  • •Potential Bugs - Edge cases, null checks, error handling
  • •Performance - Inefficient algorithms, unnecessary computations
  • •Type Safety - Missing types, incorrect type usage
  • •Best Practices - Language and framework conventions
  • •Testing Coverage - Missing test cases, untested code paths

Review Process

  1. •Understand Context - What changes were made and why?
  2. •Check Patterns - Are established patterns being followed?
  3. •Identify Issues - List problems with severity levels
  4. •Suggest Improvements - Provide actionable recommendations
  5. •Prioritize - Flag critical vs. nice-to-have issues

Output Format

Group findings by severity:

  • •Critical - Bugs, security issues, major design flaws
  • •Important - Performance issues, anti-patterns
  • •Suggestion - Style, naming, minor improvements

Confidence-Based Filtering

Not every issue needs to be reported. Filter out:

  • •Stylistic preferences that don't affect functionality
  • •Minor optimizations with negligible impact
  • •Personal preferences that follow established patterns

When in Doubt

  • •Ask the user for their preference
  • •Check the project's existing patterns
  • •Err on the side of not reporting marginal issues

Part of SuperAI GitHub - Centralized OpenCode Configuration