返回技能目录
人工审校软件开发

技能研究档案

spec-miner

来源声明

Reverse-engineering specialist that extracts specifications from existing codebases. Use when working with legacy or undocumented systems, inherited projects, or old codebases with no documentation. Invoke to map code dependencies, generate API documentation from source, identify undocumented business logic, figure out what code does, or create architecture documentation from implementation. Trigger phrases: reverse engineer, old codebase, no docs, no documentation, figure out how this works, inherited project, legacy analysis, code archaeology, undocumented features.

保留作者的英文原始描述。它表达设计意图,不代表已验证的实际效果。

Reverse-engineering specialist that extracts specifications from existing codebases. Use when working with legacy or undocumented systems, inherited projects, or old codebases with no documentation. Invoke to map code dependencies, generate API documentation from source, identify undocumented business logic, figure out what code does, or create architecture documentation from implementation. Trigger phrases: reverse engineer, old codebase, no docs, no documentation, figure out how this works, inherited project, legacy analysis, code archaeology, undocumented features.

developmentreferences

研究摘要

人工审校

Skill Atlas editorial review · editorial-preview-1.0.0

只读型逆向工程 Skill,探索无文档的代码库,并基于代码证据编写 EARS 格式规格说明。

审阅范围

对所收集的 SKILL.md 及辅助文本/代码进行确定性静态分析。机器辅助候选和发现等待人工编辑复核;不评估动态行为或效果。

可观察能力

  • 界定分析范围,并用 Glob、Grep 和 Read 梳理结构。
  • 沿代码追踪数据流和请求路径。
  • 用 EARS 格式记录观察到的需求并附代码位置。
  • 标注不确定之处,并将规格说明保存到 specs/{project_name}_reverse_spec.md。

限制与范围

  • 只能描述代码所体现的内容,意图和业务规则仍属推断。
  • 仅限 Read、Grep、Glob,无法运行系统来确认行为。
  • 未在真实遗留项目上验证输出质量。

适用读者

  • 接手遗留代码的工程师
  • 为老系统补文档的团队
  • 评估改造范围的规划者

技能包含哪些内容

文件清单来自记录提交中的采集源码。路径与大小属于结构观察,不代表质量评价。

个文件
5
Markdown 文件
5
个脚本
0
个引用
4
大小
12.1 KB
结构复杂度
低

文件清单

路径类型字节
skills/spec-miner/references/analysis-checklist.md引用2 KB
skills/spec-miner/references/analysis-process.md引用1.3 KB
skills/spec-miner/references/ears-format.md引用1.8 KB
skills/spec-miner/references/specification-template.md引用2.4 KB
skills/spec-miner/SKILL.md指令4.7 KB

SKILL.md 前置元数据

{
  "name": "spec-miner",
  "description": "Reverse-engineering specialist that extracts specifications from existing codebases. Use when working with legacy or undocumented systems, inherited projects, or old codebases with no documentation. Invoke to map code dependencies, generate API documentation from source, identify undocumented business logic, figure out what code does, or create architecture documentation from implementation. Trigger phrases: reverse engineer, old codebase, no docs, no documentation, figure out how this works, inherited project, legacy analysis, code archaeology, undocumented features.",
  "license": "MIT",
  "allowed-tools": "Read Grep Glob",
  "metadata": {
    "author": "https://github.com/Jeffallan",
    "company": "https://synergetic.solutions",
    "version": "1.1.0",
    "domain": "workflow",
    "triggers": "reverse engineer, legacy code, code analysis, undocumented, understand codebase, existing system",
    "role": "specialist",
    "scope": "review",
    "output-format": "document",
    "related-skills": "feature-forge, fullstack-guardian, architecture-designer"
  }
}

对运行顺序的一种解读

此工作流根据源码材料重建表面顺序。除非明确记录受控评估,否则它属于暂定解读,而非真实运行轨迹。

  1. 01
    输入

    界定分析范围

    确定边界:整个系统还是某个功能。

  2. 02
    操作

    探索与追踪

    用 Glob、Grep、Read 梳理结构,再追踪数据流和请求路径。

  3. 03
    判断

    检查覆盖度

    确认文件覆盖足够;若关键入口未读,继续探索后再写文档。

  4. 04
    输出

    输出规格说明

    保存规格说明,包含技术栈、结构、EARS 需求、不确定项和建议。

潜在风险面

静态发现用于提示安装前需要检查的位置,不能证明一项技能安全、恶意或有效。

有需要复核的项目

收集的文本中有 10 项确定性词法发现需要复核。发现可能仅位于文档中,不证明已执行或可被利用;仅含二进制的文件未扫描。

网络访问已观察到
文件写入已观察到
凭据使用未观察到
中

文本提及网络访问

6 次命中

此源文件行包含 URL 或网络客户端。应检查目标地址、发送的数据和信任边界;静态文本无法证明实际发生了网络活动。 该规则在收集文本中命中 6 处;此处列出 3 个代表位置。命中文本可能只是文档,分析器没有执行它。

编辑安全提示

Skill 仅限 Read、Grep、Glob,但保存的规格说明可能暴露敏感配置或业务逻辑,分享前请先检查。

上下文可能消耗在哪里

发现、激活与资源层具有不同含义。只有使用指定官方方法并记录模型的运行时,才显示精确数值。

精确 token 计数不可用

在管线使用明确模型执行官方精确计数之前,包括文件清单在内的所有 token 值均为 null。

未使用要求的官方方法记录 token 测量。本档案不会用字符数、文件大小或估计值替代。

受控评估

没有受控评估

该技能仅具备以上标明的证据。不声称任何任务成功率、token 效率、延迟或质量优势。

尚未测试

在源码可见的情况下使用

安装命令
git clone https://github.com/Jeffallan/claude-skills.git cd claude-skills # 将该 Skill 目录复制到以下任一位置: # ~/.claude/skills/spec-miner # .claude/skills/spec-miner
  1. 打开固定版本源码:https://github.com/Jeffallan/claude-skills/blob/67c1961b180dc92cf08983344627109e983e8f51/skills/spec-miner/SKILL.md。
  2. 审阅许可证、前置条件和所有捆绑资源。
  3. 将 Skill 安装到个人或项目级 Claude Code skills 目录。
  4. 重启或刷新 Claude Code,使该 Skill 可被发现。
验证安装

确认目标位置能够发现 spec-miner 的 SKILL.md,然后让 Claude Code 在一个小型测试任务中使用它。

前置条件

  • 只能描述代码所体现的内容,意图和业务规则仍属推断。
  • 仅限 Read、Grep、Glob,无法运行系统来确认行为。
  • 未在真实遗留项目上验证输出质量。

示例提示词

代表性任务

Use the spec-miner skill on the billing module in this repository. Map entry points and data flows, write observed requirements in EARS format with file locations, and list open questions separately.

使用前检查

Read the spec-miner skill files at https://github.com/Jeffallan/claude-skills/blob/67c1961b180dc92cf08983344627109e983e8f51/skills/spec-miner/SKILL.md. Explain its workflow, required tools, permissions and the main risks before I decide whether to install it.

附上限制条件的判断

优势

  • 要求区分观察到的事实与推断。
  • 只读工具白名单限制了副作用。

注意事项

  • 大型代码库可能超出探索所能覆盖的范围。
  • 会在仓库的 specs 目录写入规格文件。
研究结论

为无文档系统补充文档的谨慎低风险方式,前提是其推断经过人工复核。

从结构上保持可追溯

档案是对公开工件的限时阅读,而非背书。请通过来源链接检查确切的上游材料。

Jeffallanclaude-skillsskills/spec-miner
来源类型
社区
提交
67c1961b180dc92cf08983344627109e983e8f51
采集时间
2026年11月20日
最近核查
2026年11月20日
上游许可证
MIT
仓库星标
12,464
内容依据

仅对提供的文件进行确定性词法解析、YAML frontmatter 校验、文件清单、来源描述的工作流程提取和启发式暂定分类。由机器生成;无 LLM、执行、动态测试或人工编辑批准。