菜单
Reusable Meta_Kim file inventory classification flow. It helps separate durable sources, generated evidence, runtime mirrors, temporary state, and risky unknowns before cleanup or commit.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/KimYx0207/Meta_Kim --skill same-set-reusable-flow-for-project-file-inventor命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | same-set-reusable-flow-for-project-file-inventor |
| version | 0.1.0 |
| author | Meta_Kim |
| user-invocable | false |
| candidateType | skill |
| sourceGapId | gap-create-skill-same-set-reusable-flow--c89068df05 |
| approvalEvidence | warden-approved-file-inventory-skill-2026-06-05 |
| trigger | project file inventory|file classification|repo cleanup|项目文件整理|文件用途整理|清理项目 |
| tools | ["shell","filesystem"] |
| description | Reusable Meta_Kim file inventory classification flow. It helps separate durable sources, generated evidence, runtime mirrors, temporary state, and risky unknowns before cleanup or commit. |
Use this skill when a Meta_Kim workspace has too many files and the user needs to know which are useful, generated, risky, or safe to ignore.
Produce a source-backed inventory, not a cleanup guess. The output should help the user decide what to keep, commit, ignore, or inspect again.
Start with read-only evidence:
git status --short --branchgit ls-filesgit ls-files --others --exclude-standardgit check-ignore -v <path> for suspicious ignored pathsrg -n "<filename-or-command>" package.json docs scripts tests canonical config for tracked candidatesnpm run script inventory from package.jsonDo not classify a tracked file as removable only because it has few references. Check package scripts, tests, docs, sync scripts, setup scripts, and contracts first.
canonical/, config/contracts/, config/capability-index/, durable scripts, package scripts, and tests that define behavior..claude/, .agents/, .cursor/, openclaw/, .codex/ mirrors generated from canonical sources or sync scripts..meta-kim/state/default/, graph outputs, reports, SQLite run state, smoke artifacts, and temporary proof files.task_plan.md, findings.md, progress.md; keep during active work even if ignored.Return a compact table with:
End with a short "do not delete yet" list for any file whose role is uncertain.
This skill binds the execution-tools-and-commands, capability-discovery-and-retrieval, planning-continuity, runtime-native-surfaces, and verification-eval-and-release abstract capability families. It may inspect filesystem, Git, scripts, config, docs, and runtime projections, but it does not mutate the workspace by itself.
task_plan.md, findings.md, and progress.md are preserved as update-only continuity state.inspect again rather than deleted or ignored.Block cleanup or write actions when evidence is missing, Git state is unclear, the file is active planning state, or the user has not explicitly approved mutation.
Return to Fetch when file purpose or source-of-truth status is unclear. Return to Thinking when the inventory exposes a capability gap or a cleanup path with competing trade-offs. Return to Verification when the classification depends on generated evidence that needs a fresh command.
Use read-only commands such as git status --short --branch, git ls-files, git ls-files --others --exclude-standard, git check-ignore -v <path>, targeted rg, package script inspection, and the most specific validator for the affected file class. Run npm run meta:prompt:validate after editing this skill.
Preserve Skills, WebSearch/browser/research, filesystem, shell, apply_patch, MCP, memory, Graphify, graph, hooks, commands, rules, agents, subagents, approval, sandbox, runtime tools, package scripts, setup, sync, install, uninstall, status, doctor, validators, runtime projections, and active planning files.