Token导航 LogoToken导航

Agent Skill · 研究检索

pr-review-agent(公关审查 Agent)

pr-review-agent 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。

pr-review-agent 是研究检索类 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,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 53,927。

总安装 519周安装 21GitHub Stars 1下载 163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:pr-review-agent(公关审查 Agent)
来源仓库:https://github.com/vishal2457/open-orchestra
仓库路径:skills/pr-review-agent
安装命令:
npx skills add https://github.com/vishal2457/open-orchestra --skill pr-review-agent
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/vishal2457/open-orchestra --skill pr-review-agent

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于查找、检索和筛选相关信息。pr-review-agent 属于研究检索类 Skill,可作为该场景下的辅助能力补充。适合根据关键词或任务场景快速定位候选结果。可结合来源仓库和原始 README 核验具体用法。安装前建议确认权限范围和维护状态。注意是否会触发联网或文件读写。

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。

来源仓库内路径:skills/pr-review-agent

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/vishal2457/open-orchestra --skill pr-review-ag…;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

pr-review-agent 怎么安装?

在终端执行 npx skills add https://github.com/vishal2457/open-orchestra --skill pr-review-agent。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

pr-review-agent 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。

pr-review-agent 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/vishal2457/open-orchestra --skill pr-review-agent 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。

pr-review-agent 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。

pr-review-agent 用的人多吗?

公开数据是累计安装 519 次,近一周安装 21 次,下载 163 次,GitHub 1 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。

pr-review-agent 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

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。
  • 第三方扫描已通过,常见的高风险行为没有命中。

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。
  • 需要处理研究检索类任务、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。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
pr-review-agent

公关审查 Agent · 本页

5191执行命令未明确通过

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

PR Review Agent

Open Orchestra Settings

Before doing anything else, look for orchestra-settings.json in the current workspace or repository root.

  • If the file exists, treat it as the workflow source of truth.
  • Read workflow.sequence to understand the expected stage order.
  • Read workflow.agents.pr-review-agent to determine:

- dependsOn: agents that must be complete before this agent runs. - next: the next agent to hand off to. - invocation: whether the next agent should be invoked automatically (auto) or left for the user to invoke manually (manual).

  • If dependsOn lists agents that are not complete yet, stop and report that PR review is blocked by workflow configuration.
  • When this agent finishes, hand off only to the configured next agent.
  • If next is null, treat the workflow as complete unless the settings file defines a different continuation outside this agent.
  • If invocation is manual, do not auto-invoke the next agent; leave a clear handoff for the user.
  • If the file does not exist, use the built-in default flow: planning-agent -> implementation-agent -> pr-review-agent, with PR review as the terminal step.
  • Any future custom agent added by the user must be respected if it appears in orchestra-settings.json; do not assume the core three-agent workflow is exhaustive.

Purpose

Run a focused PR review that starts from changed code, validates only relevant requirements/spec sections, and reports blocking risks with minimal token use.

Runtime Configuration

  • Resolve the parent ticket reference from current conversation context first, then latest handoff payload if present.
  • If no parent ticket reference is available, ask the user for the ticket reference and stop.
  • Use the available issue tracker MCP directly for ticket operations.
  • If the required issue tracker MCP is unavailable, stop immediately and do not proceed with the task.
  • For every tracker comment/status update, include: Skill-Version: pr-review-agent@0.0.1.

When to Invoke

  • When a PR is ready for review and linked to an issue.

Required Inputs

  • Parent issue ID.
  • Parent issue tag pr-published.
  • Linked PR and PR diff / changed files.
  • Most recent prior handoff comment in <!-- OPEN-ORCHESTRA-HANDOFF --> format.

Outputs

  • PR comment:
  • Short review summary.
  • Findings that require changes (if any), tied to changed code.
  • Issue tracker comment:
  • Short review summary.
  • Required changes checklist (if any).
  • Status update:
  • In Progress when changes are required.
  • Done when changes are acceptable.
  • Optional invocation:
  • init-architect with scoped context when architecture-impacting changes are detected.
  • Parent issue tags:
  • pr-reviewed when review is clean.
  • open-pr-review-questions when review is blocked by missing context.
  • A handoff comment with a meaningful stage heading and wrapped JSON block:

Handing Off for PR Review

{
  "execution_trace": "Execution-Trace:\nActions:\n1. <action>\n2. <action>\nDecisions:\n- <review decision + reason>\nReferences:\n- <PR diff or requirement artifact>\nAssumptions:\n- <assumption>\nOpen-Questions: none|<question list>\nSkill-Version: pr-review-agent@0.0.1",
  "handoff_summary": {
    "from_skill": "pr-review-agent",
    "to_skill": "implementation-agent|none",
    "status": "ready|blocked|completed",
    "delta": ["<new findings or closure updates from this review pass>"],
    "key_decisions": [{"decision": "<decision>", "reason": "<reason>"}],
    "relevant_artifacts": [
      {
        "artifact": "<artifact name>",
        "hash": "sha256:<hash>",
        "last_modified": "<ISO-8601>",
        "summary": "<short relevance summary>"
      }
    ],
    "open_blockers": [{"blocker": "<text>", "owner": "<owner>", "next_action": "<next action>"}],
    "next_guidance": {
      "need_full": ["<artifact names implementation must fully read>"],
      "focus": ["<highest-priority fix/check items>"],
      "findings": [
        {
          "id": "<finding-id>",
          "severity": "high|medium|low",
          "file": "<path>",
          "problem": "<what is wrong>",
          "required_change": "<what to change>",
          "affected_subtasks": ["<subtask identifier>"]
        }
      ]
    }
  }
}
  • handoff_summary must be <= 600 tokens.

Context Gathering Order (Strict)

  1. Locate the most recent comment containing <!-- OPEN-ORCHESTRA-HANDOFF --> from the previous skill.
  2. Parse the JSON inside it; treat handoff_summary as primary context.
  3. Review relevant_artifacts and hashes first.
  4. Declare exactly which artifacts require full reads with need_full.
  5. Read full content only if hash changed or review cannot proceed without it.
  6. Do not read entire issue history or all prior execution traces by default.

Procedure (Diff-First)

  1. Resolve the parent ticket reference from context and verify the required issue tracker MCP is available.
  2. Validate parent issue has tag pr-published.
  3. Execute the strict context gathering order above.
  4. Fetch PR diff and changed files first. This is the primary review input.
  5. Build a minimal evidence set from the diff:
  • modified modules and behaviors.
  • affected API/data/auth/integration/runtime surfaces.
  • acceptance criteria touched by changed code.
  1. Based on step 5, declare need_full and fetch only relevant sections from:
  • acceptance criteria.
  • technical-details.
  • other artifacts explicitly referenced by prior handoff.
  1. Review only for correctness/regression risk in changed code and mismatch with the fetched relevant spec sections.
  2. In the same pass, detect architecture impact from the PR diff. Impact is true when one or more of these is present:
  • API contract changes.
  • Data model/migration/index/persistence contract changes.
  • Service/module boundary changes.
  • Integration/event flow changes.
  • Authn/authz/session/security boundary changes.
  • Runtime/infrastructure behavior changes that alter topology or critical flows.
  1. If architecture impact is true, prepare scoped payload for init-architect:
  • parent issue ID.
  • PR identifier/link.
  • changed files list.
  • concise diff summary.
  • matched architecture-impact cases.
  • suggested architecture doc sections to update.
  1. Do not over-engineer:
  • avoid style-only nits.
  • report only issues that can cause bugs, regressions, broken behavior, or criteria mismatch.
  1. Post a short PR comment with result and concise findings.
  2. Post a short issue tracker comment with the same outcome summary.
  3. If review is blocked by missing context:
  • Add open-pr-review-questions.
  • Post handoff JSON with status: blocked and explicit open_blockers.
  • Stop and wait for clarifications.
  1. If required code changes are found and review is not blocked:
  • Remove open-pr-review-questions if present.
  • Set issue status to In Progress.
  • Post handoff JSON with to_skill set to the configured next agent (default implementation-agent), status: ready, and next_guidance.findings populated.
  • If orchestra-settings.json allows automatic handoff, invoke the configured next agent with the same parent issue ID. Otherwise stop after leaving the handoff for the user.
  1. If no required changes remain:
  • Remove open-pr-review-questions if present.
  • Add tag pr-reviewed.
  • Set issue status to Done.
  • Post handoff JSON with to_skill: none and status: completed.
  1. If architecture impact is true, invoke init-architect with the scoped payload from step 9.

Guardrails

  • Use the PR diff as the first and primary source.
  • Do not read the full repository for this review.
  • Do not read entire issue history by default.
  • Use handoff summary first, then lazy-load only required artifact sections.
  • Prioritize correctness and functional risk over stylistic preferences.
  • Keep findings actionable and tied to specific changed files.
  • Do not run tracker operations until a parent ticket reference is resolved and the required tracker MCP is available.
  • Keep PR and tracker comments concise; do not paste raw command logs.

Handoff

Primary consumer: the next agent configured for pr-review-agent in orchestra-settings.json; default is none when review is clean and implementation-agent when fixes are required.

平台分布

Codex

33.86%
按下载量换算55

Claude

31.03%
按下载量换算51

Cursor

18.31%
按下载量换算30

Gemini CLI

8.69%
按下载量换算14

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

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/vishal2457/open-orchestra --skill pr-review-agent 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。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