Token导航 LogoToken导航

Agent Skill · 研究检索

retro(retro 分析)

retro 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。

retro 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,权限为需要联网;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 68,654。

总安装 225周安装 9GitHub Stars 35下载 73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:retro(retro 分析)
来源仓库:https://github.com/kazdenc/builder-skills
仓库路径:skills/retro
安装命令:
npx skills add https://github.com/kazdenc/builder-skills --skill retro
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/kazdenc/builder-skills --skill retro

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。

retro 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中快速定位候选结果。适用于关键词搜索、任务场景匹配或来源线索筛选等研究检索场景。通过 npx skills add 命令从 GitHub 仓库安装并使用。

安装前需确认权限范围、维护状态及是否触发联网或文件操作。建议结合原始 README 核验具体用法和功能边界。

来源仓库内路径:skills/retro

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/kazdenc/builder-skills --skill retro;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

    它需要需要联网,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

retro 怎么安装?

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

retro 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。

retro 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

retro 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。

retro 用的人多吗?

公开数据是累计安装 225 次,近一周安装 9 次,下载 73 次,GitHub 35 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。

retro 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

  • 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

  • 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
retro

retro 分析 · 本页

22535需要联网未明确通过

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

Run a retrospective that produces action items, not just observations. A retro without specific next steps is a venting session.

Principles — hold these throughout:

  • Blame-free. Focus on systems, processes, and circumstances — not people. "The deploy process let a bug through" not "Bob broke prod."
  • Actions over observations. Every theme must produce a concrete change. If it doesn't lead to action, note it and move on.
  • Max 3 action items. More than three won't get done. Prioritize ruthlessly.

Step 1: Set Context

Establish the boundaries before gathering input:

QuestionAnswer
What period/project?Name it specifically — "Sprint 14" or "Billing v2 launch"
What was the goal?What were we trying to accomplish?
What actually happened?State the outcome factually — shipped on time, missed by 2 weeks, hit metrics, didn't
Who was involved?List the team so contributions are visible
TimelineStart date, end date, key milestones

Write a 2-3 sentence summary: "We set out to [goal]. We [what happened]. The result was [outcome]."

If the user hasn't provided this context, ask for it. Don't fabricate project history.

Step 2: Gather Input — The 4Ls

Use the 4Ls framework to collect observations. Ask the user to contribute to each category:

CategoryPromptWhat goes here
LikedWhat went well? What should we keep doing?Practices, decisions, or moments that worked
LearnedWhat did we discover? What surprised us?New knowledge, skills, or insights gained
LackedWhat was missing? What held us back?Resources, information, tools, or processes we needed but didn't have
Longed ForWhat do we wish we had? What would make next time better?Improvements, tools, practices, or changes we want

Gathering tips:

  • Ask for specifics, not generalities. "Liked: good communication" is useless. "Liked: daily 10-min standups kept everyone unblocked" is actionable.
  • Aim for 3-5 items per category.
  • If working async, give people a template to fill in before the session.

Step 3: Identify Themes

Group related items across the 4Ls. Look for patterns:

  1. Cluster related items. If three people mention deployment issues, that's a theme.
  2. Name each theme clearly. "Deployment reliability" not "various issues."
  3. Find root causes. For the biggest themes, use the 5 Whys:
Problem: Feature launched 2 weeks late.
Why? → Integration testing took longer than expected.
Why? → We discovered API incompatibilities late.
Why? → No integration test environment existed.
Why? → We hadn't invested in test infrastructure for this service.
Why? → Test infra wasn't prioritized because the team was focused on feature work.

Root cause: No dedicated time for test infrastructure investment.

Stop when you reach something the team can actually change. Don't go deeper than useful.

  1. Rank themes by impact. Which ones, if fixed, would make the biggest difference next time?

Step 4: Generate Action Items

For each of the top 2-3 themes, create one action item. Every action must be:

RequirementExample
Specific"Set up staging environment with seed data" — not "improve testing"
AssignedOne owner, by name or role. "The team" owns nothing.
Time-bound"Complete by end of Sprint 15" — not "soon"
ACTION ITEM TEMPLATE:

Theme: [theme name]
Action: [specific thing to do]
Owner: [person or role]
Due: [date or sprint]
Success looks like: [how we know it's done]

Hard limit: 3 action items. If you have 5 good ones, pick the 3 with highest impact. Carry the rest to next retro if they're still relevant.

Step 5: Write the Retro Summary

Produce a single document the team can reference:

# Retro: [Project/Sprint Name]

**Period:** [start] — [end]
**Team:** [names/roles]
**Goal:** [what we set out to do]
**Outcome:** [what actually happened]

## Summary
[2-3 sentences: goal, what happened, result]

## 4Ls

### Liked
- [item]
- [item]

### Learned
- [item]
- [item]

### Lacked
- [item]
- [item]

### Longed For
- [item]
- [item]

## Themes
1. **[Theme name]** — [1-sentence description]. Root cause: [root cause].
2. **[Theme name]** — [1-sentence description]. Root cause: [root cause].
3. **[Theme name]** — [1-sentence description]. Root cause: [root cause].

## Action Items

| # | Action | Owner | Due | Done? |
|---|--------|-------|-----|-------|
| 1 | [specific action] | [owner] | [date] | [ ] |
| 2 | [specific action] | [owner] | [date] | [ ] |
| 3 | [specific action] | [owner] | [date] | [ ] |

## Notes
[Anything else worth recording that doesn't fit above]

Output

Deliver the retro summary using the template above. Customize it to the actual project — don't leave placeholder text. If information is missing, flag what you need from the user rather than filling in generic content.

平台分布

Codex

40.53%
按下载量换算30

Claude

28.08%
按下载量换算20

Cursor

17.7%
按下载量换算13

Gemini CLI

10.31%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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