Token导航 LogoToken导航TokenDH.com
研究检索执行命令github未标认证来源可访问许可证需确认审计提醒

github-reviewGitHub 审查

Agent Skill

用于围绕 GitHub 仓库、Issue、Pull Request、分支、提交和代码协作流程提供辅助能力。它适合让 Agent 查询项目状态、整理变更、辅助创建或检查协作事项,并把仓库中的信息转成可执行的下一步。使用时需要区分只读查询和写入操作;涉及创建 PR、修改 Issue、推送分支或访问私有仓库时,应确认 token 权限、目标仓库范围和用户授权。

总安装

196

周安装

8

GitHub Stars

80

下载量

63
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:github-review(GitHub 审查)
来源仓库:https://github.com/holon-run/holon
仓库路径:skills/github-review
安装命令:
npx skills add https://github.com/holon-run/holon --skill github-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/holon-run/holon --skill github-review

简介

审查开放拉取请求的正确性、回归风险与安全漏洞,输出结构化评价。

  • 可直接发布评论摘要,支持内联建议但不实施代码修复。
  • 需通过 npx skills add 从 GitHub 安装,依赖 gh CLI 与 API 读取权限。
  • 要求 GITHUB_TOKEN 能读取 PR 数据并发表评论,仅限只读分析。
  • github-review 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

GitHub Review Skill

Summary

Use this skill when you need to review a pull request, identify the highest-signal findings, and publish one structured GitHub review.

When To Use

  • Reviewing an open pull request for correctness, regressions, or safety issues
  • Publishing a review summary plus optional inline comments
  • Working directly from raw GitHub CLI and API data

Do Not Use

  • Implementing fixes on the PR branch
  • Opening a new PR from an issue
  • Project-wide prioritization or PM analysis

Prerequisites

  • gh CLI authentication is required.
  • GITHUB_TOKEN/GH_TOKEN needs permissions to read PR data and publish reviews/comments.

Runtime Paths

  • GITHUB_OUTPUT_DIR: output artifacts directory (caller-provided preferred; otherwise temp dir).
  • GITHUB_CONTEXT_DIR: context directory (default ${GITHUB_OUTPUT_DIR}/github-context).

Inputs (Manifest-First)

Preferred input when already available:

  • ${GITHUB_CONTEXT_DIR}/manifest.json

Optional inputs:

  • Any context artifact listed as status=present in manifest.json.

If no manifest is provided, collect PR context directly with gh:

gh pr view <pr_number> --repo <owner/repo> --json number,title,body,state,url,baseRefName,headRefName,headRefOid,author,createdAt,updatedAt,mergeable,reviews,changedFiles,additions,deletions
gh pr view <pr_number> --repo <owner/repo> --json files
gh pr diff <pr_number> --repo <owner/repo>
gh api repos/<owner>/<repo>/issues/<pr_number>/comments --paginate
gh api graphql -f query='
  query($owner:String!, $repo:String!, $number:Int!) {
    repository(owner:$owner, name:$repo) {
      pullRequest(number:$number) {
        reviewThreads(first:100) {
          nodes {
            isResolved
            comments(first:100) {
              nodes { id body path line author { login } }
            }
          }
        }
      }
    }
  }' -F owner=<owner> -F repo=<repo> -F number=<pr_number>

This skill must not assume fixed context filenames. Use manifest.artifacts[] (id, path, status, description) to determine available context.

Workflow

1. Collect context

  • If ${GITHUB_CONTEXT_DIR}/manifest.json exists, use it.
  • Otherwise, collect PR metadata, files, diff, comments, and review-thread context directly with gh.

2. Perform review

Generate:

  • ${GITHUB_OUTPUT_DIR}/review.md
  • ${GITHUB_OUTPUT_DIR}/review.json
  • ${GITHUB_OUTPUT_DIR}/summary.md
  • Optional ${GITHUB_OUTPUT_DIR}/manifest.json (execution metadata)

3. Publish review

Use gh api with a JSON payload file:

gh api repos/<owner>/<repo>/pulls/<pr_number>/reviews -X POST --input <review-payload.json>

Review Standards

Scope and priority

Review focus order:

  1. Correctness bugs
  2. Security/safety issues
  3. Performance/scalability risks
  4. API compatibility and error handling
  5. High-impact maintainability issues

Incremental-first

  • Prioritize newly introduced changes (new commits and new diff hunks).
  • Expand scope only when needed to validate correctness or safety.

Historical deduplication

  • Check existing review threads/comments before raising findings.
  • Do not repeat already-raised issues unless there is new evidence or changed impact.
  • If re-raising, explain the delta briefly.

Keep signal high

  • Avoid low-value style nitpicks unless they affect behavior/maintainability.
  • Keep feedback concise, specific, and actionable.
  • Prefer fewer high-impact findings over exhaustive noise.

Output Contract

review.md

Human-readable review summary containing:

  • conclusion-first summary
  • key findings ordered by severity
  • actionable recommendations

review.json

Structured inline findings:

[
  {
    "path": "path/to/file.go",
    "line": 42,
    "severity": "error|warn|nit",
    "message": "Issue description",
    "suggestion": "Optional concrete fix"
  }
]

Severity semantics:

  • error: must-fix before merge
  • warn: should-fix
  • nit: optional improvement

summary.md

Short execution summary:

  • reviewed ref/head
  • context coverage summary from manifest
  • number of findings and publish outcome
  • explicit degradation/failure reason when context is insufficient

Degradation Rules

  • If core review artifacts are missing (for example pr_metadata, diff and files both unavailable), do not fabricate certainty.
  • Either:

- produce summary-only review with explicit limitations and no inline comments, or - fail with clear reason in summary.md.

Publishing Guardrails

  • Publish at most one review per execution round.
  • A successful primary publish is terminal; do not run alternate publish paths.
  • Before fallback publish, check whether an equivalent Holon review already exists for the same head SHA and skip if present.

Configuration

  • DRY_RUN=true: preview only.
  • MAX_INLINE=N: cap inline comments.
  • POST_EMPTY=true: allow posting empty review.

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

36.49%
按下载量换算23

Claude

29.26%
按下载量换算18

Cursor

20.35%
按下载量换算13

Gemini CLI

9.86%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/holon-run/holon --skill github-review 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。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