Token导航 LogoToken导航

Agent Skill · 待分类

gh-process-review(GH 流程审查)

gh-process-review 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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。

gh-process-review 是待分类类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为执行命令;许可证信息还不明确,第三方扫描有提醒项,按安装量在待分类类的 11,209 个 Skill 里排第 6,005。

总安装 494周安装 20GitHub Stars 10下载 155核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:gh-process-review(GH 流程审查)
来源仓库:https://github.com/keboola/ai-kit
仓库路径:skills/gh-process-review
安装命令:
npx skills add https://github.com/keboola/ai-kit --skill gh-process-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/keboola/ai-kit --skill gh-process-review

这个 Skill 是做什么的

这个 Skill 的分类还没最终确定,用法以来源仓库的说明为准。

gh-process-review 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息。适合在 Codex、Claude、Cursor、Gemini CLI 中围绕仓库状态、代码变更或协作事项进行整理。通过 npx skills add 命令从指定 GitHub 仓库安装并使用。

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。可结合来源仓库和原始 README 继续核验具体用法。

来源仓库内路径:skills/gh-process-review

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/keboola/ai-kit --skill gh-process-review;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要执行命令,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

gh-process-review 怎么安装?

在终端执行 npx skills add https://github.com/keboola/ai-kit --skill gh-process-review。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

gh-process-review 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

gh-process-review 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/keboola/ai-kit --skill gh-process-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。

gh-process-review 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

gh-process-review 用的人多吗?

公开数据是累计安装 494 次,近一周安装 20 次,下载 155 次,GitHub 10 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

gh-process-review 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理待分类类任务、又不想每次都手写提示词的人。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

待分类类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
infsh-cli

infsh CLI · Codex、Claude

4,374,576382权限未标注未明确未收录
ai-video-generation

AI 视频生成 · Codex、Claude

4,313,376382调用外部服务未明确有提醒
lark-openapi-explorer

云雀 OpenAPI 浏览器 · Codex、Claude

2,064,1209,011需要联网未明确有提醒
lark-workflow-standup-report

Lark 工作流程站立报告 · Codex、Claude

2,041,6328,978需要联网未明确通过
use-my-browser

使用我的浏览器 · Codex、Claude

1,848,49140操作浏览器未明确未通过
gh-process-review

GH 流程审查 · 本页

49410执行命令未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

GitHub Review Processing

Process PR review threads efficiently by storing them locally and addressing them one by one.

Working Directory Context

CRITICAL: All commands MUST be run from the user's project root directory, NOT from the skill directory.

  • The user will be in THEIR project directory when invoking this skill
  • All script calls use $SKILL_DIR/scripts/review.sh — the single entry point
  • The script auto-detects the reviews file from the current branch's PR
  • DO NOT cd into the skill directory - the script handles path resolution internally

SKILL_DIR = directory containing this SKILL.md (automatically resolved by Claude)

CRITICAL Rules

  1. STAY IN PROJECT ROOT - Never cd into the skill directory. Call $SKILL_DIR/scripts/review.sh from the project root.
  2. ONE COMMIT PER FIX - Each review thread fix MUST be committed separately. Never batch multiple fixes into one commit.
  3. START IN PLANNING MODE - Always enter planning mode first (unless context explicitly says otherwise like "skip planning" or "no planning").
  4. NEVER RESOLVE THREADS - Do NOT use the GitHub API to resolve review threads. Only the reviewer may resolve their own comments. Use mark for local tracking only.

Workflow

Phase 1: Setup

Fetch reviews for current PR and list unresolved threads:

"$SKILL_DIR/scripts/review.sh" fetch

This auto-detects the PR from the current branch, fetches all review threads, and prints unresolved threads.

Continue mode (/gh-process-review continue)

Skip fetching, list unresolved threads from existing reviews file:

"$SKILL_DIR/scripts/review.sh" list

Phase 2: Planning (default)

After fetching, enter planning mode using EnterPlanMode tool.

CRITICAL: Context is cleared when the plan is approved. The plan must be self-contained. $SKILL_DIR will no longer be set after context clears, so resolve it to the absolute path in the plan. Do NOT use shell variables — write out the full path in every command. Beyond the thread list and proposed fixes, always include these sections in the plan (replacing /absolute/path/to with the actual resolved $SKILL_DIR/scripts/review.sh path):

## Skill Script
Commands (use the literal path, not a variable):
- `/absolute/path/to/scripts/review.sh get <ID>` — get thread details
- `/absolute/path/to/scripts/review.sh reply <ID> <commit-hash>` — reply with commit
- `/absolute/path/to/scripts/review.sh reply <ID> -m "message"` — reply with message only
- `/absolute/path/to/scripts/review.sh reply <ID> -m "message" <commit-hash>` — reply with both
- `/absolute/path/to/scripts/review.sh mark <ID>` — mark locally resolved

## Rules
- NEVER resolve threads via GitHub API — only the reviewer may resolve their own comments
- ONE commit per fix — never batch multiple fixes
- Each fix: implement → commit → push → reply → mark → next thread
- STAY in project root — never cd into skill directory

In the plan:

  1. List all unresolved threads with thread ID, file:line, and brief summary
  2. For each thread, outline the proposed fix approach
  3. Identify any dependencies between fixes
  4. Include implementation order

Skip planning only if user explicitly requests it (e.g., "skip planning", "no planning", "just fix it").

Phase 3: Implementation

For each unresolved thread (one at a time):

  1. Get thread details: "$SKILL_DIR/scripts/review.sh" get PRRT_...
  2. Read the file and understand the feedback
  3. Implement the fix
  4. COMMIT THIS FIX IMMEDIATELY - Do not continue to next thread without committing: git add <changed-files> git commit -m "$(cat <<'EOF' AI-XXXX Address review: <brief description> Addresses comment by @<reviewer> on <file>:<line> Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> EOF)"
  5. Push the commit
  6. Reply with commit: "$SKILL_DIR/scripts/review.sh" reply PRRT_...
  7. Mark resolved: "$SKILL_DIR/scripts/review.sh" mark PRRT_...
  8. Only then proceed to next thread

Phase 4: Reflection

After all threads are addressed, reflect on the reviewer's feedback:

  1. Could this feedback have been prevented? If the reviewer caught issues that better instructions (in CLAUDE.md, MEMORY.md, or other project memory files) would have avoided, suggest specific additions to those files so the same mistakes don't recur.
  2. Did you get confused while addressing any feedback? If any review comment was ambiguous, required multiple attempts, or led you down a wrong path, suggest updating memory files with clarifying guidance for future iterations.
  3. Present the suggested memory updates to the user for approval before writing them.

CLI Reference

Single entry point: $SKILL_DIR/scripts/review.sh

Global flags

  • --file <path> — Override auto-detected reviews file path

Subcommands

fetch
"$SKILL_DIR/scripts/review.sh" fetch                                    # Current branch's PR
"$SKILL_DIR/scripts/review.sh" fetch 123                                # Specific PR number
"$SKILL_DIR/scripts/review.sh" fetch https://github.com/owner/repo/pull/123  # PR URL

Fetches review threads via GraphQL, saves to .scratch/reviews/, and auto-lists unresolved threads.

list
"$SKILL_DIR/scripts/review.sh" list                # summary (default)
"$SKILL_DIR/scripts/review.sh" list full           # full JSON details
"$SKILL_DIR/scripts/review.sh" list ids            # just thread IDs
get
"$SKILL_DIR/scripts/review.sh" get PRRT_kwDOAbcd1234
"$SKILL_DIR/scripts/review.sh" get PRRT_kwDOAbcd1234 PRRT_kwDOEfgh5678

Returns: Single thread as JSON object, or JSON array when multiple IDs given. Fields: path, line, comments[], isResolved, isOutdated. Warns on missing IDs.

reply

Requires at least one of -m or a commit hash.

"$SKILL_DIR/scripts/review.sh" reply PRRT_kwDOAbcd1234 abc1234                         # Commit hash → "Fixed in abc1234"
"$SKILL_DIR/scripts/review.sh" reply PRRT_kwDOAbcd1234 -m "No changes needed"          # Message only
"$SKILL_DIR/scripts/review.sh" reply PRRT_kwDOAbcd1234 -m "Refactored per suggestion" abc1234  # Both → "Fixed in abc1234\n\nRefactored per suggestion"
mark
"$SKILL_DIR/scripts/review.sh" mark PRRT_kwDOAbcd1234
"$SKILL_DIR/scripts/review.sh" mark PRRT_kwDOAbcd1234 "Fixed in commit abc123"

JSON Structure

  • pr: PR metadata (number, title, branch names, url)
  • threads[]: id, path, line, isResolved, isOutdated, local_resolved, local_notes, comments[]
  • reviews: Review summaries

Key Thread Fields

  • id: GraphQL ID (starts with PRRT_)
  • path: File path
  • line: Line number
  • comments[0].body: Main feedback
  • comments[0].author: Reviewer

平台分布

Codex

38.06%
按下载量换算59

Claude

28.65%
按下载量换算44

Cursor

20.2%
按下载量换算31

Gemini CLI

9.65%
按下载量换算15

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/keboola/ai-kit --skill gh-process-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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills