Token导航 LogoToken导航

Agent Skill · 研究检索

spec-reviewer(规格审查员)

spec-reviewer 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。

spec-reviewer 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,权限为接触敏感数据;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 63,891。

总安装 309周安装 13GitHub Stars 210下载 108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:spec-reviewer(规格审查员)
来源仓库:https://github.com/codemie-ai/codemie-code
仓库路径:skills/spec-reviewer
安装命令:
npx skills add https://github.com/codemie-ai/codemie-code --skill spec-reviewer
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/codemie-ai/codemie-code --skill spec-reviewer

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。

spec-reviewer 用于审查技术规格文档,确保其满足 Jira 需求、项目设计指南和架构原则。适用于在解决方案架构阶段,对技术规格进行合规性和完整性检查的场景。通过二进制判定(APPROVED 或 NEEDS WORK)提供关键反馈,不输出次要评论或代码片段。

安装前需确认权限范围、维护状态及是否涉及联网、命令执行或文件读写操作。建议结合来源仓库和原始 README 进一步验证具体使用方法和约束条件。

来源仓库内路径:skills/spec-reviewer

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/codemie-ai/codemie-code --skill spec-reviewer;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

    它需要接触敏感数据,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

spec-reviewer 怎么安装?

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

spec-reviewer 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。

spec-reviewer 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

spec-reviewer 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。

spec-reviewer 用的人多吗?

公开数据是累计安装 309 次,近一周安装 13 次,下载 108 次,GitHub 210 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。

spec-reviewer 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

  • 需要接触敏感数据,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。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 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
spec-reviewer

规格审查员 · 本页

309210接触敏感数据未明确有提醒

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

Spec Reviewer: Technical Specification Review

Purpose

This skill reviews technical specifications produced by the solution-architect agent to ensure they:

  • Address all requirements from the Jira ticket
  • Follow project design guidelines from .codemie/guides/
  • Comply with architectural principles and patterns
  • Are focused, clear, and implementation-ready

The skill provides a binary verdict (APPROVED or NEEDS WORK) with critical feedback only—no minor comments, no code snippets.

When to Use This Skill

Use this skill when:

  • Solution-architect agent has produced a technical specification
  • Before starting implementation of a complex feature
  • User asks to "review spec", "validate specification", "check design doc"
  • Need to verify specification against Jira ticket requirements
  • Want to ensure spec follows project design principles

Review Workflow

Phase 1: Input Gathering

Step 1: Obtain Specification

Get the technical specification to review:

  • From user message: User provides spec content directly
  • From file path: User provides path to spec file (use Read tool)
  • From previous context: Spec was generated earlier in conversation

Step 2: Identify Jira Ticket

Extract Jira ticket ID from specification or ask user:

  • Look for EPMCDME-XXXXX pattern in spec
  • If not found, ask user for ticket ID
  • Use brianna skill to fetch ticket description and summary
Use Skill tool with skill="brianna" and args:
"Get ticket details for EPMCDME-XXXXX. I need only the description and summary fields."

Phase 2: Criteria Loading

Step 3: Load Relevant Project Guides

Based on spec content, load applicable guides from .codemie/guides/:

Spec MentionsLoad Guide (P0)Also Load (P1)
Architecture, layers, components.codemie/guides/architecture/architecture.md-
API, endpoints, REST.codemie/guides/api/ (if exists).codemie/guides/architecture/architecture.md
Agent, plugin, registry.codemie/guides/architecture/architecture.md.codemie/guides/integration/external-integrations.md
Security, auth, credentials.codemie/guides/security/security-practices.md.codemie/guides/development/development-practices.md
Testing, mocking, coverage.codemie/guides/testing/testing-patterns.md-
Error handling, logging.codemie/guides/development/development-practices.md.codemie/guides/standards/code-quality.md
Provider, LLM, integration.codemie/guides/integration/external-integrations.md.codemie/guides/architecture/architecture.md
Git, workflow, CI/CD.codemie/guides/standards/git-workflow.md-

Step 4: Identify Design Principles

Extract key design principles from loaded guides:

  • Layered architecture rules (CLI → Registry → Plugin → Core → Utils)
  • Plugin isolation principles
  • Error handling patterns
  • Security requirements
  • Testing strategies
  • Dependency rules

Phase 3: Critical Review

Step 5: Verify Against Jira Ticket

Compare specification to Jira ticket requirements:

CRITICAL Issues (Must report):

  • ❌ Missing acceptance criteria not addressed in spec
  • ❌ Misalignment with ticket goals or scenarios
  • ❌ Spec solves different problem than ticket describes
  • ❌ Key user-facing functionality omitted

NOT Critical (Skip):

  • Minor wording differences
  • Implementation details beyond ticket scope
  • Additional nice-to-have features

Step 6: Verify Against Design Principles

Check spec compliance with project design guidelines:

Architecture Violations (CRITICAL)

From .codemie/guides/architecture/architecture.md:

Must Report:

  • ❌ Skipping architectural layers (e.g., CLI directly calls Plugin)
  • ❌ Core layer depending on Plugin layer (dependency inversion violation)
  • ❌ Plugin-to-Plugin direct dependencies
  • ❌ Business logic in CLI layer
  • ❌ Missing registry registration for new plugins

Example Feedback Format:

**Architecture Violation**: Spec proposes CLI command directly instantiating ClaudePlugin.
**Principle**: CLI → Registry → Plugin flow (5-layer architecture)
**Reference**: .codemie/guides/architecture/architecture.md:246-273 (Communication Rules)
**Impact**: Breaks plugin isolation, makes testing difficult, violates Open/Closed principle
Security Violations (CRITICAL)

From .codemie/guides/security/security-practices.md:

Must Report:

  • ❌ Hardcoded credentials or API keys in spec
  • ❌ Missing input validation for user-provided data
  • ❌ Logging sensitive data without sanitization
  • ❌ File path operations without security checks
  • ❌ Missing CredentialStore usage for credential storage

Example Feedback Format:

**Security Violation**: Spec shows API key stored in configuration file.
**Principle**: No hardcoded credentials, use CredentialStore
**Reference**: .codemie/guides/security/security-practices.md (Credential Storage section)
**Impact**: Credentials exposed in version control, security risk
Error Handling Violations (CRITICAL)

From .codemie/guides/development/development-practices.md:

Must Report:

  • ❌ Using generic Error instead of specific error classes
  • ❌ Missing error context for debugging
  • ❌ Swallowing errors without logging
  • ❌ No error propagation strategy defined

Example Feedback Format:

**Error Handling Violation**: Spec uses generic Error for agent not found.
**Principle**: Use specific error classes from src/utils/errors.ts
**Reference**: .codemie/guides/development/development-practices.md (Error Handling section)
**Impact**: Poor user experience, difficult debugging, no structured error handling
Testing Violations (CRITICAL)

From .codemie/guides/testing/testing-patterns.md:

Must Report:

  • ❌ No testing strategy defined for complex features
  • ❌ Mixing unit and integration test concerns
  • ❌ Missing test isolation strategy
  • ❌ Incorrect mocking approach (static imports without dynamic loading)

Example Feedback Format:

**Testing Violation**: Spec proposes static imports for modules that need mocking.
**Principle**: Use dynamic imports after beforeEach for mockable modules
**Reference**: .codemie/guides/testing/testing-patterns.md (Dynamic Imports section)
**Impact**: Tests cannot properly mock dependencies, brittle test suite
Integration Violations (CRITICAL)

From .codemie/guides/integration/external-integrations.md:

Must Report:

  • ❌ Direct integration without provider abstraction
  • ❌ Missing error handling for external service failures
  • ❌ No retry or timeout strategy for external calls
  • ❌ Hardcoded external service URLs

Step 7: Verify Focus and Clarity

Check specification quality:

CRITICAL Issues (Must report):

  • ❌ Spec is vague or ambiguous about key implementation details
  • ❌ Multiple disconnected features bundled together
  • ❌ Missing critical interfaces or contracts
  • ❌ Unclear component responsibilities
  • ❌ No clear success criteria or validation approach

NOT Critical (Skip):

  • Minor typos or grammatical issues
  • Formatting inconsistencies
  • Missing diagrams (unless critical for understanding)
  • Overly verbose explanations

Phase 4: Verdict and Feedback

Step 8: Provide Review Verdict

Format review results as follows:

If NO Critical Issues Found:
## Specification Review: APPROVED ✅

**Jira Ticket**: EPMCDME-XXXXX
**Specification**: [Title or path]

### Verdict
This specification is **APPROVED** for implementation.

### Review Summary
- ✅ Addresses all Jira ticket acceptance criteria
- ✅ Follows 5-layer architecture principles
- ✅ Complies with security guidelines
- ✅ Proper error handling strategy defined
- ✅ Clear component responsibilities and interfaces
- ✅ [Additional positive findings]

### Next Steps
Proceed with implementation following the specification. Use tech-lead skill to guide implementation.
If Critical Issues Found:
## Specification Review: NEEDS WORK ⚠️

**Jira Ticket**: EPMCDME-XXXXX
**Specification**: [Title or path]

### Verdict
This specification **REQUIRES ADDITIONAL WORK** before implementation.

### Critical Issues

#### 1. [Issue Category] - [Brief Title]
**Violation**: [What principle/requirement is violated]
**Principle**: [Which design principle from guides]
**Reference**: [Guide path and section]
**Impact**: [Why this matters, consequences of not fixing]

#### 2. [Issue Category] - [Brief Title]
**Violation**: [What principle/requirement is violated]
**Principle**: [Which design principle from guides]
**Reference**: [Guide path and section]
**Impact**: [Why this matters, consequences of not fixing]

[Continue for all critical issues]

### Jira Ticket Alignment

[If applicable]
- ❌ Acceptance criterion "[text]" not addressed
- ❌ Scenario "[text]" not covered
- ❌ [Other alignment issues]

### Recommendations

[High-level guidance - NO code snippets]
1. [Action to address issue category 1]
2. [Action to address issue category 2]
3. [Action to address issue category 3]

### Next Steps
Address critical issues above, then resubmit specification for review.

Critical Review Criteria

✅ What to Report (CRITICAL Only)

CategoryReport If
ArchitectureViolates 5-layer architecture, breaks dependency rules, skips layers
SecurityHardcoded credentials, missing validation, unsafe operations, logging sensitive data
Error HandlingUsing generic errors, missing context, swallowing exceptions
TestingNo strategy for complex features, incorrect mocking approach
Jira AlignmentMissing acceptance criteria, wrong problem being solved
ClarityVague key details, unclear responsibilities, no success criteria
IntegrationDirect coupling to external services, no error handling

❌ What NOT to Report (Minor Issues)

CategorySkip If
StyleFormatting, minor typos, grammar issues
OptimizationPerformance suggestions not affecting correctness
ExtrasMissing nice-to-have features beyond ticket scope
PreferencesAlternative approaches that are equally valid
DocumentationMinor documentation improvements

Key Principles

Do's

✅ Focus on CRITICAL issues only (design principle violations, missing requirements) ✅ Reference specific guides and sections ✅ Explain WHY issue is critical (impact) ✅ Fetch Jira ticket to verify alignment ✅ Load applicable guides before review ✅ Provide clear verdict (APPROVED or NEEDS WORK) ✅ Give focused feedback without code snippets ✅ Be constructive and specific

Don'ts

❌ Don't report minor style or formatting issues ❌ Don't provide code snippets or implementation fixes ❌ Don't suggest "nice to have" improvements ❌ Don't be overly pedantic about minor details ❌ Don't assume—verify against actual guides ❌ Don't approve specs with critical violations ❌ Don't provide vague feedback like "improve clarity"

Example Reviews

Example 1: APPROVED Specification

User: "Review this spec for EPMCDME-10500"
[Spec: New REST endpoint following existing patterns]

Spec Reviewer:
1. Fetches EPMCDME-10500 via brianna
2. Loads .codemie/guides/architecture/architecture.md
3. Reviews spec:
   - Follows CLI → Registry → Plugin architecture ✅
   - Uses existing error classes ✅
   - Addresses all acceptance criteria ✅
   - Clear interfaces defined ✅
4. Verdict: APPROVED ✅
5. Recommends: Proceed with implementation

Example 2: NEEDS WORK - Architecture Violation

User: "Review this spec for EPMCDME-10600"
[Spec: New agent with CLI directly calling plugin code]

Spec Reviewer:
1. Fetches EPMCDME-10600 via brianna
2. Loads .codemie/guides/architecture/architecture.md
3. Identifies CRITICAL issue:
   - Spec shows CLI command directly instantiating agent plugin
   - Violates 5-layer architecture (CLI → Registry → Plugin)
   - Reference: architecture.md:246-273
4. Verdict: NEEDS WORK ⚠️
5. Feedback: "CLI must call AgentRegistry.getAgent(), not instantiate plugin directly"

Example 3: NEEDS WORK - Security Violation

User: "Review this spec for EPMCDME-10700"
[Spec: Provider integration with API key in config file]

Spec Reviewer:
1. Fetches EPMCDME-10700 via brianna
2. Loads .codemie/guides/security/security-practices.md
3. Identifies CRITICAL issue:
   - API key stored in configuration file
   - Violates credential storage principle
   - Reference: security-practices.md
4. Verdict: NEEDS WORK ⚠️
5. Feedback: "Use CredentialStore.getInstance() for secure credential storage"

Example 4: NEEDS WORK - Missing Jira Requirements

User: "Review this spec for EPMCDME-10800"
[Spec: Agent feature but missing key acceptance criterion]

Spec Reviewer:
1. Fetches EPMCDME-10800 via brianna
2. Ticket has acceptance criterion: "Support batch mode processing"
3. Spec only covers streaming mode
4. Identifies CRITICAL gap:
   - Acceptance criterion not addressed
   - Spec incomplete for ticket requirements
5. Verdict: NEEDS WORK ⚠️
6. Feedback: "Spec must address batch mode processing (acceptance criterion 3)"

Integration with Other Skills

Solution-Architect Skill

  • Input source: Specs produced by solution-architect
  • Workflow: solution-architect creates spec → spec-reviewer validates → implement or revise
  • Feedback loop: If NEEDS WORK, solution-architect revises based on feedback

Brianna Skill

  • Purpose: Fetch Jira ticket for alignment verification
  • Usage: Request description + summary fields only
  • Handle missing: If ticket not found, cannot verify alignment (note in review)

Tech-Lead Skill

  • Handoff: After APPROVED verdict, tech-lead can guide implementation
  • Workflow: spec-reviewer approves → tech-lead implements
  • Dependency: tech-lead should not start without approved spec for complex features

Error Handling

Specification Not Provided

Error: No specification provided for review.

Please provide:
- Specification content (paste directly)
- File path to specification document
- Reference to spec in conversation history

Jira Ticket Not Found

Warning: Unable to fetch Jira ticket EPMCDME-XXXXX.

Proceeding with guide compliance review only. Cannot verify alignment with ticket requirements.

Consider:
- Verifying ticket ID format
- Checking ticket exists and is accessible
- Reviewing ticket requirements manually

Guides Not Available

Error: Required guide not found: [path]

Cannot complete review without design guidelines.

Please ensure .codemie/guides/ directory is available with:
- architecture/architecture.md
- security/security-practices.md
- development/development-practices.md
- [Other applicable guides]

Success Criteria

A successful spec review results in:

  • ✅ Specification content obtained and understood
  • ✅ Jira ticket fetched and reviewed
  • ✅ Relevant guides loaded and consulted
  • ✅ Critical issues identified (if any)
  • ✅ Clear verdict provided (APPROVED or NEEDS WORK)
  • ✅ Focused feedback with guide references (if NEEDS WORK)
  • ✅ User has actionable next steps

Additional Resources

Reference Files

For detailed review criteria:

  • references/review-checklist.md - Comprehensive checklist for each review category
  • references/violation-examples.md - Examples of critical violations by category

Integration Points

This skill coordinates with:

  • CLAUDE.md: Uses guide references and task classifier
  • .codemie/guides/: Loads all applicable guides for compliance verification
  • brianna skill: Fetches Jira ticket information for alignment check
  • solution-architect skill: Reviews specs produced by this skill
  • tech-lead skill: Hands off to tech-lead after APPROVED verdict

平台分布

Codex

34%
按下载量换算37

Claude

32.89%
按下载量换算36

Cursor

20.59%
按下载量换算22

Gemini CLI

8.35%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

可疑

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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