Token导航 LogoToken导航

Agent Skill · 研究检索

specs-review(规格审查)

specs-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。

specs-review 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,权限为只读;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 72,863。

总安装 186周安装 8GitHub Stars 2下载 65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:specs-review(规格审查)
来源仓库:https://github.com/b12consulting/skills
仓库路径:skills/specs-review
安装命令:
npx skills add https://github.com/b12consulting/skills --skill specs-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/b12consulting/skills --skill specs-review

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。

specs-review 用于查找、检索和筛选相关信息。适用于对技术规格或开发内容进行审查与筛选。通过 npx skills add 命令从指定仓库安装后调用。建议在使用前确认权限及是否允许联网查询外部资源。

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/specs-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/b12consulting/skills --skill specs-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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

specs-review 怎么安装?

在终端执行 npx skills add https://github.com/b12consulting/skills --skill specs-review。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

specs-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。

specs-review 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

specs-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。

specs-review 用的人多吗?

公开数据是累计安装 186 次,近一周安装 8 次,下载 65 次,GitHub 2 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。

specs-review 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。
  • 标注为只读权限,试用阶段不容易误改本地文件。

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。
  • 只想先试试效果、不希望它改动本地文件的人。

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
specs-review

规格审查 · 本页

1862只读未明确通过

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

Specs Review

Audit the health and consistency of a project's spec-driven documentation.

Skill Dependencies

This skill is part of a set of skills designed to work together:

  • spec-driven — Methodology reference (structure, formats, rules)
  • specs-setup — Initialize specs/ for a new project
  • specs-tickets — Create and execute tickets through their lifecycle
  • specs-review (this skill) — Audit specs health and consistency

If any of these skills are missing from the project, instruct the user to install them before proceeding:

npx skills add b12consulting/skills --skill <missing_skill>

Always load the spec-driven skill first for the full methodology reference.

Procedure

1. Structural Completeness

Verify all required files exist:

  • specs/README.md
  • specs/Vision.md
  • specs/PRD.md
  • specs/Goals.md
  • specs/Architecture/README.md
  • specs/Glossary.md
  • specs/Changelog.md
  • specs/decisions/ directory
  • specs/tickets/ directory

Report any missing files. Suggest running specs-setup if critical files are absent.

2. Document Quality

For each project-level document, check:

DocumentCheck
Vision.mdHas vision statement, problem statement, and target audience?
PRD.mdHas functional requirements, non-functional requirements, and scope? Are requirements numbered (FR-1, FR-2...)?
Goals.mdHas measurable success metrics? Are milestones defined with target dates?
Architecture/README.mdHas system overview, key components, and tech stack? Links to sub-documents if they exist?
Glossary.mdHas at least a few terms? Are terms used consistently in other documents?
Changelog.mdIs up to date with recent completed tickets?
README.mdDoes the active tickets table match actual ticket statuses? Is the "last updated" date recent?

Flag documents that are empty templates (never filled in) or have placeholder content.

3. Ticket Health

Scan all ticket folders in specs/tickets/ and check each one:

Frontmatter integrity:

  • Does every ticket have a README.md with valid frontmatter (id, title, status, created, updated)?
  • Are status values valid (research, specifying, open-questions, planned, in-progress, done, archived)?

Lifecycle consistency:

  • Does a ticket marked in-progress have a Tasks.md with at least some unchecked items?
  • Does a ticket marked done have all acceptance criteria in Spec.md checked off?
  • Does a ticket marked planned have both Spec.md and Plan.md?
  • Does a ticket marked archived have a documented reason in its README.md body?

Stale tickets:

  • Any ticket with status in-progress or research whose updated date is more than 2 weeks old? Flag as potentially stale.
  • Any ticket with status open-questions that has been waiting for decisions? Flag with the pending questions.

Orphaned files:

  • Any ticket folder missing a README.md?
  • Any Decisions.md with items still in the Open section for a ticket that has moved past specifying?

Report findings as a table:

#TitleStatusIssues
001...in-progressStale (last updated 3 weeks ago)
002...doneAcceptance criteria not all checked

4. Cross-Reference Consistency

Check that documents reference each other correctly:

  • README.md dashboard vs. reality: Do the "Active Tickets" and "Recently Completed" tables match actual ticket statuses?
  • ADR references: Are ADRs referenced in the tickets that triggered them? Are ADR-modified documents (Vision, PRD, Goals, Architecture) updated accordingly?
  • Glossary usage: Scan Vision, PRD, and Architecture for domain terms. Are they defined in the Glossary? Are there Glossary terms that appear unused?
  • Architecture sub-documents: If Architecture/README.md links to sub-documents, do those files exist?

5. Drift Detection

Compare specs against the codebase:

  • Does the tech stack in Architecture/README.md match the actual dependencies (check package.json, pyproject.toml, Cargo.toml, etc.)?
  • Do the key components described in Architecture match the actual project structure?
  • Are there functional requirements in PRD.md that don't correspond to any ticket (planned or completed)?

This step requires codebase analysis. Flag potential drift but note that confirmation with the user is needed — apparent drift may be intentional and just not yet documented.

6. Report

Summarize findings in three categories:

Critical — Blocks correct agent behavior:

  • Missing ground truth documents (Vision, PRD, Architecture)
  • Tickets with invalid or missing frontmatter
  • Conflicts between specs and code that could lead to wrong implementation

Warning — Should be addressed soon:

  • Stale tickets
  • Unresolved open questions
  • README dashboard out of sync
  • Empty or placeholder documents

Suggestion — Improvements for completeness:

  • Missing Glossary terms
  • Tickets without Journal.md that had complex processes
  • Architecture that could benefit from sub-documents

7. Propose Actions

For each finding, propose a concrete action:

  • "Create ADR to document the React-to-Vue migration that happened in ticket 005"
  • "Update specs/README.md — tickets 003 and 004 are marked done but still in Active Tickets"
  • "Ticket 007 has been in-progress for 3 weeks with no updates — ask the owner for status"
  • "Add 'workspace' and 'campaign' to the Glossary — used in PRD but not defined"

Ask the user which actions to take. Execute approved actions immediately.

平台分布

Codex

34.52%
按下载量换算22

Claude

30.48%
按下载量换算20

Cursor

19.47%
按下载量换算13

Gemini CLI

9.74%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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