Agent Skills › flutter/agent-plugins › api-review

api-review

GitHub

根据规范 API 设计指南审查代码，评估接口清晰度、安全性及易用性，并生成包含评分、关键问题和改进示例的结构化报告。

.agents/agents/reidbaker-agent/skills/api-review/SKILL.md flutter/agent-plugins

触发场景

用户要求审查 API 代码检查代码是否符合 API 设计原则

安装

npx skills add flutter/agent-plugins --skill api-review -g -y

更多选项

非标准路径

npx skills add https://github.com/flutter/agent-plugins/tree/main/.agents/agents/reidbaker-agent/skills/api-review -g -y

不安装直接使用

npx skills use flutter/agent-plugins@api-review

指定 Agent (Claude Code)

npx skills add flutter/agent-plugins --skill api-review -a claude-code -g -y

安装 repo 全部 skill

npx skills add flutter/agent-plugins --all -g -y

预览 repo 内 skill

npx skills add flutter/agent-plugins --list

SKILL.md

Frontmatter

{
    "name": "api-review",
    "description": "Reviews the specified code against the canonical API Design guidelines. Use this skill when the user asks for an API review or to check code against API design principles."
}

API review skill

This skill reviews code against the canonical API Design guidelines.

Instructions

Load Guidelines: Read the API design guidelines from references/canonical_api_design.md to ensure they are fully available in the context.
Identify Target: Identify the code to review.
- If the user specified files (e.g., "review main.dart"), use those.
- If the user has an open file in their context, assume that is the target.
- If neither, ask the user to specify the target files.
Analyze: For each target file, perform a deep analysis against the "Foundations of Canonical API Design Principles" (loaded in step 1), specifically looking for:
- Contract-First: Is the interface clear and decoupled from implementation?
- KISS/YAGNI: Are there unnecessary parameters or over-generalized features?
- Ergonomics: Are names intent-revealing? Do they follow the Principle of Least Astonishment?
- CQS: Are commands and queries separated?
- Safety: Are types used strictly (Enums vs Strings)? Is validation visible?
- Explicit Configuration: Are dependencies explicitly injected rather than implicitly resolved via global state, registries, or environment variables?
Report: Generate a structured report:
- Score: Give a letter grade (A-F) based on alignment.
- Critical Issues: Violations that must be fixed (e.g., severe strictness or safety issues).
- Suggestions: Ergonomic improvements (renaming, rearranging).
- Code Examples: Provide before vs after code blocks for the suggested improvements.
- Save the report as a markdown artifact in the conversation artifacts directory (e.g., <appDataDir>/brain/<conversation-id>/api_review_results.md).