Token导航 LogoToken导航TokenDH.com
研究检索敏感数据github未标认证来源可访问许可证需确认审计通过

contextual-review情境审查

Agent Skill

contextual-review 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

285

周安装

12

GitHub Stars

1,706

下载量

1
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:contextual-review(情境审查)
来源仓库:https://github.com/flowglad/flowglad
仓库路径:skills/contextual-review
安装命令:
npx skills add https://github.com/flowglad/flowglad --skill contextual-review
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/flowglad/flowglad --skill contextual-review

简介

contextual-review 对代码变更、实现计划和架构决策进行质量与安全审查。

  • 适合在 Codex、Claude、Cursor、Gemini CLI 中作为自动化评审助手提供反馈。
  • 依据项目基线指南判断类型安全、事务完整性等标准,标注常见盲点风险。
  • 输出为 actionable 建议列表,而非绝对结论,需人工复核后再实施修改。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

Contextual Review

Perform comprehensive reviews of code changes, implementation plans, and architecture decisions. Analyzes for quality, correctness, security, and adherence to project standards.

First: Read the Base Guidelines

Before reviewing any code, read base-review.md. It establishes:

  • Reviewer philosophy (respect existing patterns, burden of proof on changes)
  • Core quality standards (type safety, transaction integrity, data access, caching)
  • Severity calibration with codebase-specific examples
  • Common blind spots that reviewers unfamiliar with this codebase miss

The base guidelines apply to ALL reviews. Area-specific guides add targeted checklists.

When to Use

  • Reviewing pull request changes
  • Examining workspace diffs before creating a PR
  • Getting feedback on code changes
  • Identifying potential issues before merging
  • Reviewing gameplans and implementation plans before execution
  • Validating data model and API design decisions

Area-Specific Guidelines

Based on what files changed, consult the appropriate reference:

Changed FilesReference
platform/docs/docs-review.md - Documentation review guidelines
platform/flowglad-next/src/db/schema/, openapi.json, api-contract/api-review.md - Data model and API review
packages/packages-review.md - SDK package review
playground/playground-review.md - Example project review
platform/flowglad-next/platform-review.md - Main platform review

For reviewing implementation plans before code is written:

Review TypeReference
Gameplans / Implementation Plansgameplan-review.md - Pre-implementation plan review

Read the relevant reference file(s) based on the diff to get area-specific checklists and guidelines.

Review Process

0. Checkout PR

Run gh pr checkout <PR> to get the PR code locally. If it fails, continue with the review.

1. Gather Context

First, understand the scope of changes:

# Get the diff statistics to understand what files changed
GetWorkspaceDiff with stat: true

# Then examine individual file changes
GetWorkspaceDiff with file: 'path/to/file'

2. Review Categories

Analyze changes across these dimensions:

CategoryFocus Areas
CorrectnessLogic errors, edge cases, null handling, off-by-one errors
SecurityInput validation, injection risks, auth/authz, secrets exposure
PerformanceN+1 queries, unnecessary loops, missing indexes, memory leaks
MaintainabilityCode clarity, naming, DRY violations, complexity
TestingTest coverage, edge cases tested, test quality
TypesType safety, proper typing, avoiding any

3. Project-Specific Checks

For this codebase, also verify:

  • Bun: Using bun instead of npm or yarn
  • Drizzle ORM: Schema changes use migrations:generate, never manual migrations
  • Testing Guidelines:

- No mocking unless for network calls - No .spyOn or dynamic imports - No any types in tests - Each it block should have specific assertions, not toBeDefined - One scenario per it with exhaustive assertions

  • Security: Check OWASP top 10 vulnerabilities (XSS, injection, etc.)

4. Provide Feedback

Use the DiffComment tool to leave targeted feedback:

DiffComment({
  comments: [
    {
      file: "path/to/file.ts",
      lineNumber: 42,
      body: "Potential SQL injection vulnerability. Consider using parameterized queries."
    }
  ]
})

Review Checklist

Code Quality

  • Clear, descriptive variable and function names
  • Functions are focused and not too long
  • No dead code or commented-out code
  • Error handling is appropriate
  • Edge cases are handled

Security

  • No hardcoded secrets or credentials
  • Input is validated and sanitized
  • No SQL injection vectors
  • No XSS vulnerabilities
  • Authentication/authorization is correct
  • Sensitive data is not logged

Performance

  • No unnecessary database queries
  • Appropriate use of indexes
  • No obvious memory leaks
  • Pagination for large datasets
  • Caching where appropriate

Testing

  • New code has tests
  • Tests cover happy path and error cases
  • Tests are meaningful, not just for coverage
  • No flaky test patterns

TypeScript

  • Proper types used (no any without justification)
  • Type narrowing is correct
  • Generic types are appropriate
  • Null/undefined handled properly

Output Format

Provide a structured review with:

  1. Summary: Brief overview of what the PR does
  2. Findings: Categorized issues (Critical, High, Medium, Low, Suggestions)
  3. Positive Notes: Good patterns or improvements noticed
  4. Recommendation: Approve, Request Changes, or Comment

Severity Levels

  • Critical: Security vulnerabilities, data loss risks, breaking changes
  • High: Bugs, significant performance issues, missing error handling
  • Medium: Code quality issues, missing tests, unclear logic
  • Low: Style issues, minor improvements, nitpicks
  • Suggestion: Optional improvements, alternative approaches

Example Review

## Summary
This PR adds user authentication using JWT tokens with refresh token support.

## Findings

### Critical
- **src/auth/token.ts:45**: JWT secret is hardcoded. Move to environment variable.

### High
- **src/auth/login.ts:23**: Missing rate limiting on login endpoint.

### Medium
- **src/auth/validate.ts:12**: Token expiration check should use `<=` not `<` to handle exact expiration time.

### Suggestions
- Consider adding request ID to auth logs for debugging.

## Positive Notes
- Good separation of concerns between token generation and validation
- Comprehensive error types for different auth failures

## Recommendation
**Request Changes** - Address the critical security issue before merging.

Workflow

  1. Attempt to checkout the PR with gh pr checkout <PR> (continue if it fails)
  2. Get diff statistics with GetWorkspaceDiff(stat: true)
  3. Review changed files systematically
  4. Use DiffComment for inline feedback
  5. Provide overall summary and recommendation
  6. Offer to help fix any critical issues

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.32%
按下载量换算0

Claude

27.72%
按下载量换算0

Cursor

19.01%
按下载量换算0

Gemini CLI

8.58%
按下载量换算0

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills