By LearnPrompt
Polishes a functional Skill into a publishable, installable, and verifiable public asset through five craft stages: material inspection, ecosystem benchmarking, multi-dimensional measurement, incremental refinement, and post-release observation.
把你的Skill拿到班门前,让祖师爷重新打磨一遍。
把一个"能用的Skill",打磨成"能被理解、能被安装、能被传播、能被验证、能持续进化"的公共资产。
你写了一个Skill,自己用着挺好。然后呢?
普通的做法是"帮我润色一下"。鲁班的做法是把它当作品收进工坊:先挑战它值不值得雕,再看同行凭什么立足,三把尺量出短板,一刀一刀刨,每刀都要过验证门,最后发版立规矩。
| 动作 | 干什么 | 一句狠话 |
|---|---|---|
| 验料 | 先挑战这个Skill的前提是否成立 | 朽木不可雕也,不值得就直说 |
| 访行 | 联网找同行,看清生态位 | 闭门造车出不了好工具 |
| 过尺 | 结构、实测、活体三把尺量分 | 绿色的CI会撒谎,要拉真实产物对账 |
| 慢刨 | 冻结基线,改动过验证门才保留 | 量不过就回刀,绝不为显得干了活而多刨 |
| 回炉 | 发布后留对标观察清单,下轮从反馈进 | 交活不是终点 |

npx skills add LearnPrompt/luban-skill -g
Claude Code 用户也可以走 plugin marketplace(自动更新):
/plugin marketplace add LearnPrompt/luban-skill
/plugin install luban
装完对Agent说:
让鲁班看看我这个skill:[你的Skill目录 / GitHub仓库链接 / SKILL.md内容]
鲁班会先完成验料、访行、定位、过尺,给你三个打磨方向并推荐一个——在你选方向之前,它不会动你一行字。
鲁班的第一单活:把 ai-news-radar(约1k星)从 v0.6 打磨到 v0.7.0,一个对话内完成,4个PR全部合并:
backtest_scoring.py,并立下"动评分必须附 ≥14 天回放"的项目规矩全程记录(含事故与教训):skills/luban/examples/ai-news-radar-case.md——每个数字都挂着可点击的PR链接。
| 直接让Agent"帮我改好看点" | 鲁班 | |
|---|---|---|
| 起点 | 直接改文案 | 先挑战前提:这Skill值得存在吗? |
| 依据 | 模型的审美 | 同行对标(带URL)+ 三把尺评分(带证据) |
| 改法 | 一次全改,无法归因 | 冻结基线,单提交单面,过验证门才保留 |
| 验证 | "看起来更好了" | 真实数据回放,给出翻转数字 |
| 结束 | 改完就完 | 验证工具入库、立规矩、留回炉清单 |
git reset --hard 之类的暴力回滚。luban-skill/
├── skills/luban/
│ ├── SKILL.md # 工作流本体:五个动作、九步流程、班规与验收单
│ └── examples/
│ └── ai-news-radar-case.md # 实战案例:真实仓库、真实数字、全程可查证
├── assets/ # demo GIF 与可复现的录制脚本(vhs tape)
├── .claude-plugin/ # Claude Code plugin marketplace 清单
├── README.md
└── LICENSE
装完用这句验收:
让鲁班看看这个skill:https://github.com/anthropics/skills
合格的表现:它先输出验料挑战和带URL的同行对标,给三个方向并停手等你选——而不是直接开始改写。
鲁班的五个动作不是凭空造的,它站在这些工作的肩膀上:
MIT — 随便用,随便改,随便让它帮你打磨。
验料 · 访行 · 过尺 · 慢刨 · 回炉
学手艺,不偷皮。
LearnPrompt 出品 · 同门手艺
鲁班·Skill打磨 · 庖丁·博主蒸馏 · 蔡伦·对话造纸 · 阿福·LLM Todo · AI雷达·零API资讯 · 淘金小镇·ClawHub日榜 · Irasutoya·正文配图 · Humanize PPT·简报编排 · CC Harness·六件套
公众号「卡尔的AI沃茨」 · X @aiwarts · learnprompt.pro
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub learnprompt/luban-skill --plugin luban24小时AI更新雷达。雷达Skill(ai-radar)让Agent零API读取GitHub Pages公开JSON,回答"今天AI圈有什么";伯乐Skill(ai-news-radar)帮你维护信源、抓取脚本、Pages部署。数据来自LearnPrompt/ai-news-radar,GitHub Actions每日刷新。
Six portable harness skills distilled from a CC-style coding agent: dream-memory, memory-extractor, verification-gate, swarm-coordinator, structured-context-compressor, kairos-lite. The parts that separate a fun demo from a stable toolchain.
面向 Agent 的 PPT 简报编排器:把资料变成 AST 大纲+逐页素材决定,交给下游 PPT Skill(guizang/frontend-slides)100%原生渲染,再用 QA 循环盯住渲染结果。Humanize 自己不渲染。PPT is an audience-state changer, not an information container.
淘金小镇 | Skillrush Town — 每天从 ClawHub 下载榜 Top100 里淘出新进榜、增速榜和潜力 Skill。每日快照 + 历史对比 + 日报 + 公告板页面,网页是公告板,Skill 是淘金方法。
建立新技能、修改和改進現有技能、衡量技能效能。用於從零開始建立技能、編輯或優化現有技能、執行評估測試、基準測試效能分析、或優化技能描述以提升觸發準確度
Use when reviewing a Claude Code skill, auditing skill quality before publish, or asking "is this skill any good?". Produces structured reports with file:line citations, severity-rated findings cards, and an optional second-opinion pass. Lens variants: full (default), safety, discoverability, architecture, parseability, tests, quick.
Self-evolving skill engine for Claude Code. Creates, scores, repairs, and hardens skills autonomously through recursive improvement cycles.
Local-only skill quality and security evaluator for Claude Code/OpenClaw. Scores six dimensions, runs local validators, and stores reports and snapshots under .skill-compass/.
Create and manage Claude Code skills, plugins, subagents, and hooks. Use when building new skills, validating existing skills, testing skills empirically, creating plugins, converting projects to plugins, creating hooks, or managing plugin automation. Includes /skills-toolkit:skill-composer, /skills-toolkit:skill-refiner, /skills-toolkit:skill-tester, /skills-toolkit:plugin-creator, /skills-toolkit:subagent-creator, /skills-toolkit:hook-creator, and /skills-toolkit:ask-user-question skills.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.