AI 执行,人类决策。666 个技能分为蜂群(并行选优)与管线(顺序执行),共 34 组。
从你的任务出发,直达正确的 Group 和命令。
蜂群:描述任务,系统从多个专家技能中选出最合适的执行。
/design-ui-swarm设计系统、图表绘制、海报生成、UI/UX 评审、原型、PPT
触发词: design, ui, ux, diagram, ppt, slides, prototype, html style, 设计, 图表, 原型, poster design
| Skill | 说明 |
|---|---|
ai-drawio | 图表生成能力已整合到 **PPT 生成 Skill** 中,作为 PPT 自动生成图表的核心模块。 |
antv-infographic | 基于 [antvis/Infographic](https://github.com/antvis/Infographic)(npm 包 `@antv/info |
base-ui | 把 Base UI 的使用流程固化为可复用 runbook,支持自动触发与落盘文档。 |
beautiful-mermaid | 把 Mermaid 图表的“美观渲染 + 终端 ASCII 输出”流程固化为可复用 runbook,并提供主题与配色的标准做法。 |
better-icons | Better Icons |
canvas-design | Canvas Design |
design-prompts | 此 Skill 集成了 [Design Prompts](https://www.designprompts.dev/) 的设计提示词库功能,提供丰富的 UI/ |
design-taste-frontend | 1. **This skill refines existing UI — it doesn't create from scratch.** For gree |
diceui | 来源: - 文档:[`https://diceui.com/docs`](https://diceui.com/docs) - 仓库(MIT):[`https: |
drawio-diagram | Drawio Diagram |
excalidraw-diagram-skill | Excalidraw Diagram Skill |
figma-vectorizer | 把“图片转高质量矢量 SVG + Figma 落地”的流程固化为可复用 runbook,支持本地批量处理与可审计输出。 |
frontend-design | This skill guides creation of distinctive, production-grade frontend interfaces |
frontend-html-theme | `frontend-html-theme` 是 AI-tools 的 **单一事实源** HTML 主题 Skill:把“任何由 AI-tools 生成/导出的 |
html-bloomberg-style | Html Bloomberg Style |
html-document-style | Html Document Style |
html-economist-style | Html Economist Style |
html-mckinsey-style | Html Mckinsey Style |
html-style-router | Html Style Router |
impeccable | This skill guides creation of distinctive, production-grade frontend interfaces |
json-canvas | Json Canvas |
kumo-ui | 把 Cloudflare Kumo UI 的安装、使用、CLI 与组件注册表流程固化为可复用 runbook,支持自动触发与落盘文档。 |
lobe-icons | 你想要的是:把 LobeHub 的 AI/LLM 模型图标 **一键下载到本地**,并提供一个“图标库 Skill”用于 **搜索 / 导出 / 复用**。 |
motionsites-landing | Motionsites Landing |
nanobanana-image-gen | Nanobanana Image Gen |
obsidian-visual-skills | 该 Skill 用于把 Obsidian 的可视化技能包安装到你的 Obsidian Vault 或任意项目目录中,以便 Claude Code 能按需使用(以 |
pake-native-app | Pake Native App |
pencil-dev | 把 Pencil 的设计到代码工作流固化为可复用 runbook 与组件文档。 |
pixelle-video | 提供 Pixelle-Video 的一键安装与运行入口,落盘到项目目录,便于审计和复用。 |
pptgen-drawio | Pptgen Drawio |
product-design | Product Design |
qiaomu-design-advisor | Qiaomu Design Advisor |
qiaomu-mondo-poster-design | Qiaomu Mondo Poster Design |
ralph-wiggum | 把 Ralph Wiggum 方法的“需求/规划/构建”三阶段固定为可版本化工件,避免只存在于聊天里的临时流程。 |
redesign-existing-projects | When applied to an existing project, follow this sequence: |
svgl-shadcn-ui | 此 Skill 集成了两个强大的前端资源库: |
svglogo-icons | 来源:[`https://svglogo.top/?view=original`](https://svglogo.top/?view=original) |
tanstack-start-dashboard | 参考模板仓库:[`Kiranism/tanstack-start-dashboard`](https://github.com/Kiranism/tanstac |
ui-skills | 来源: - https://www.ui-skills.com/ - https://github.com/ibelick/ui-skills |
ui-ux-pro-max | 将 UI/UX Pro Max 的多平台设计体系固化为可复用 runbook 与 doc pack,覆盖安装、启用、使用与验收路径。 |
visual-style | Create, extract, and apply portable visual design systems. A `visual-style.md` f |
web-interface-guidelines | 来源: - https://vercel.com/design/guidelines - https://github.com/vercel-labs/web- |
ljg-card | Ljg Card |
ljg-cards | Ljg Cards |
ljg-infograph | Ljg Infograph |
codex-slides | Codex Slides |
codex-figma | Codex Figma |
codex-figma-implement-design | Codex Figma Implement Design |
codex-figma-use | Codex Figma Use |
codex-figma-design-rules | Codex Figma Design Rules |
codex-figma-generate-design | Codex Figma Generate Design |
codex-figma-generate-library | Codex Figma Generate Library |
codex-frontend-skill | Codex Frontend Skill |
html-slides-aurora-glass | Html Slides Aurora Glass |
html-slides-midnight-tech | Html Slides Midnight Tech |
html-slides-router | Html Slides Router |
html-slides-swiss-clean | Html Slides Swiss Clean |
html-slides-terracotta | Html Slides Terracotta |
/workflow-meta-swarm工作流编排、Agent 协作、任务规划、n8n 自动化、Skill 管理
触发词: workflow, pipeline, plan, n8n, agent, worktree, sprint, tdd, 工作流, 管线, 计划, 流程自动化, advisor, 顾问, skill create
| Skill | 说明 |
|---|---|
activity-visual-log | 目标:自动生成“当天活动可视化日志”,包含两类信息源: |
advisor-council | Advisor Council |
agent-eval-system | 将“任务 / 评分器 / 过程记录 / 结果状态”落盘为可追踪资产,建立可复用的评估基线。 |
agent-pack-n-go | Clone your OpenClaw Agent to a new device. Agent handles everything via SSH remo |
agent-teams-swarm | Put the 51 SOP pipelines (6 SDLC phases) into multi-agent execution mode. Instea |
bmad-method | 把 BMAD 的三类资产落到项目仓库中,形成可追踪、可复用、可审计的交付流程: |
brainstorming | 1. **Don't skip brainstorming for \"simple\" tasks.** The most expensive bugs co |
capability-evolver | Capability Evolver |
claude-hud | 该 Skill 用于安装 **Claude HUD**(Claude Code statusline 插件),展示上下文使用量、工具活动、运行中的代理和 tod |
cli-relay | > All AI CLI tools -> localhost:8317 -> Multiple upstream providers |
composio-skills | Composio Skills |
confidence-check | Confidence Check |
continue | Continue |
craft-agents-oss | 提供 Craft Agents OSS 的一键安装与运行入口,落盘到项目目录,便于审计和复用。 |
creating-skills | 1. **description is for the MODEL, not humans.** Claude decides whether to trigg |
dispatching-parallel-agents | When you have multiple unrelated failures (different test files, different subsy |
everything-claude-code | 把 Everything Claude Code 的配置集合(agents/skills/hooks/commands/rules/MCP)固化为可复用 run |
executing-plans | Load plan, review critically, execute tasks in batches, report for review betwee |
finishing-a-development-branch | Guide completion of development work by presenting clear options and handling ch |
full-output-enforcement | Treat every task as production-critical. A partial output is a broken output. Do |
harness-engineering | Harness Engineering |
manus-sandbox | 将 Manus Sandbox 的关键机制与最佳实践沉淀为可复用的文档包(index/design/api/config/runbook),默认落盘到项目 `d |
multi-agent-orchestration | This skill imports the core multi-agent and PM2 orchestration commands from the |
multi-role-brainstorm | A structured workflow for comprehensive project planning that simulates a cross- |
n8n-automation | N8N Automation |
oc-wrapped | 将你的本地 OpenCode 使用数据生成“年度回顾”(类似 Spotify Wrapped),输出一张可分享的 PNG 图片,并把证据(命令/日志/产物)落盘 |
planning-with-files | 用本地三份文件作为“外置工作记忆”,避免长任务在对话变长后发生目标漂移: |
ralph-loop | Claude 的官方 Ralph loop 依赖 Stop hook(Codex 不具备同等机制)。本 Codex skill 提供一个可落地的等价实现: |
receiving-code-review | Code review requires technical evaluation, not emotional performance. |
reflection | 这个 Skill 用于“阶段性反思/复盘/沉淀”:把你在任务过程中遇到的问题、修复思路、验证方式与下一步计划,按统一结构写入文件,避免信息散落在对话里不可审计、 |
requesting-code-review | Dispatch superpowers:code-reviewer subagent to catch issues before they cascade. |
retro | Retro |
skill-group-loop | Skill Group Loop |
skills-marketplace | 此 Skill 集成了 [Skills Marketplace (skillsmp.com)](https://skillsmp.com/),这是一个拥有 ** |
sprint-pipeline-v2 | Sprint Pipeline V2 |
stop-slop | Stop Slop |
strategic-compact | Strategic Compact |
subagent-driven-development | Execute plan by dispatching fresh subagent per task, with two-stage review after |
supermemory | Supermemory |
systematic-debugging | **NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST.** |
telos-framework | 把“你是谁、你想改变什么、如何行动”落盘为可复用上下文,形成 AI 可理解的稳定语境,并支持定期自我升级。 |
test-driven-development | Write the test first. Watch it fail. Write minimal code to pass. |
using-git-worktrees | Git worktrees create isolated workspaces sharing the same repository, allowing w |
using-superpowers | <EXTREMELY-IMPORTANT> If you think there is even a 1% chance a skill might apply |
verification-before-completion | **NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE.** |
workflow-router | 当你输入一些典型“任务场景”表达(例如 `新任务:...` / `重构任务...` / `功能优化...`)时,自动: |
writing-plans | Write comprehensive implementation plans assuming the engineer has zero context |
writing-skills | **Writing skills IS Test-Driven Development applied to process documentation.** |
ljg-clip | Ljg Clip |
ljg-roundtable | Ljg Roundtable |
ljg-skill-map | Ljg Skill Map |
ljg-xray-skill | Ljg Xray Skill |
ljg-xray-prompt | Ljg Xray Prompt |
ljg-calendar | Ljg Calendar |
claude-opus-chatstyle | A working style for serious collaboration. It should sound like a sharp human pa |
goal | Goal |
openai-prompt-packs | Openai Prompt Packs |
/code-quality-swarmCode review, debugging, testing, architecture, best practices, git workflows
触发词: debug, code review, test, architecture, refactor, git, e2e, maestro, vibe debug, best practices, 调试, 代码审查, 测试, 重构
| Skill | 说明 |
|---|---|
adversarial-review | Adversarial Review |
agent-browser | Linux 可选: |
agentlens | Agentlens |
api-design-reviewer | Api Design Reviewer |
bug-hunt | Bug Hunt |
classical-language-safety-audit | Classical Language Safety Audit |
claude-code-best-practices | 把 Anthropic 团队 Claude Code 最佳实践 PDF 转为可复用输出流程,配图表格由 CoWork 生成。 |
claude-code-boris-tips | 把 Boris Cherny 的 Claude Code 十个使用技巧整理为可复用的自动化 runbook,形成统一的提示词与操作模板。 |
claude-code-insights | 把 Claude Code `/insights` 的使用流程固化为可复用 runbook,并支持自动触发与落盘建议。 |
clawdbot-codespaces | 把 Clawdbot/OpenClaw 在 GitHub Codespaces 的部署流程固化为可复用 runbook,并覆盖飞书 Bot 对接与协议选择。 |
code | > **AI 驱动的代码生成系统** - 支持代码生成、审查、重构、API脚手架、测试生成 |
code-review | Code Review |
commit-helper | Commit Helper |
eigent | 目标:把开源项目 [`eigent-ai/eigent`](https://github.com/eigent-ai/eigent) 以 **可审计、可脚本化* |
engineering-protocol | 本 Skill 把 discountry 的《Software Engineering Protocol》做成**可执行资产**:用于生成可复制的 checkl |
fastapi-templates | Fastapi Templates |
frontend-testing | Frontend Testing |
git-commit | Creates git commits following Conventional Commits format with proper type, scop |
github-pr-creation | Creates Pull Requests with task validation, test execution, and Conventional Com |
github-pr-merge | Merges Pull Requests after validating pre-merge checklist and handling post-merg |
github-pr-review | Resolves Pull Request review comments with severity-based prioritization, fix ap |
gitnexus-debugging | Gitnexus Debugging |
gitnexus-exploring | Gitnexus Exploring |
gitnexus-impact-analysis | Gitnexus Impact Analysis |
gitnexus-refactoring | Gitnexus Refactoring |
go-best-practices | Go Best Practices |
maestro-e2e | Maestro E2E |
makepad-skills | 该 Skill 用于把 Makepad 的 Claude Skills 套件安装到你的 Rust/Makepad 项目中,便于在 Claude Code 里获得 |
new-api-endpoint | New Api Endpoint |
postmortem | 该 Skill 用于把“传统软件工程 Postmortem(尸检 / RCA)”流程落成可执行的工作流资产: |
react-best-practices | 该 Skill 用于安装 **React/Next.js 性能优化最佳实践** 规则集,来源于 Vercel 工程团队的 agent skills。适用于组件编 |
rust-coding | Rust Coding |
senior-architect | Senior Architect |
skill-publisher | Skill Publisher |
skill-vetter | Security-first vetting protocol for AI agent skills. **Never install a skill wit |
swarm-ops | Swarm Ops |
codex-gh-address-comments | Codex Gh Address Comments |
test-runner | Test Runner |
testing | > **AI 驱动的测试系统** - 支持E2E测试、API测试、视觉回归、性能测试 |
vibe-debug | If you cannot write a test that reproduces the bug, you do not understand the bu |
x-article-publisher-skill | 该 Skill 用于把上游仓库的 Claude Skill 安装到你的项目(或个人目录)中,便于 Claude Code 自动加载使用。 |
/pm-strategy-swarmProduct strategy, business models, pricing, competitive frameworks, roadmaps
触发词: product strategy, business model, pricing, roadmap, canvas, SWOT, porter, positioning, go-to-market, north star, 产品策略, 商业模式, 定价
| Skill | 说明 |
|---|---|
ansoff-matrix | Ansoff Matrix |
beachhead-segment | Beachhead Segment |
business-model | Business Model |
growth-loops | Growth Loops |
gtm-strategy | Gtm Strategy |
identify-assumptions-existing | Identify Assumptions Existing |
identify-assumptions-new | Identify Assumptions New |
lean-canvas | Lean Canvas |
metrics-dashboard | Metrics Dashboard |
monetization-strategy | Monetization Strategy |
north-star-metric | North Star Metric |
opportunity-solution-tree | Opportunity Solution Tree |
outcome-roadmap | Outcome Roadmap |
pestle-analysis | Pestle Analysis |
pm-cmd-business-model | Pm Cmd Business Model |
pm-cmd-growth-strategy | Pm Cmd Growth Strategy |
pm-cmd-north-star | Pm Cmd North Star |
pm-cmd-pre-mortem | Pm Cmd Pre Mortem |
pm-cmd-pricing | Pm Cmd Pricing |
pm-cmd-setup-metrics | Pm Cmd Setup Metrics |
pm-cmd-stakeholder-map | Pm Cmd Stakeholder Map |
pm-cmd-strategy | Pm Cmd Strategy |
pm-cmd-transform-roadmap | Pm Cmd Transform Roadmap |
pm-cmd-value-proposition | Pm Cmd Value Proposition |
porters-five-forces | Porters Five Forces |
positioning-ideas | Positioning Ideas |
pre-mortem | Pre Mortem |
pricing-strategy | Pricing Strategy |
prioritization-frameworks | Prioritization Frameworks |
prioritize-assumptions | Prioritize Assumptions |
prioritize-features | Prioritize Features |
product-name | Product Name |
product-strategy | Product Strategy |
product-vision | Product Vision |
stakeholder-map | Stakeholder Map |
startup-canvas | Startup Canvas |
swot-analysis | Swot Analysis |
value-prop-statements | Value Prop Statements |
value-proposition | Value Proposition |
/game-dev-swarm48-Agent 游戏工作室: 团队编排 + 游戏设计 + 引擎专家 + QA + 发布
触发词: game, 游戏, godot, unity, unreal, game design, team-combat, team-narrative, level design, 游戏开发
| Skill | 说明 |
|---|---|
gs-brainstorm | Gs Brainstorm |
gs-code-review | Gs Code Review |
gs-sprint-plan | Gs Sprint Plan |
gs-retrospective | Gs Retrospective |
start | Start |
setup-engine | Setup Engine |
design-system | Design System |
design-review | Design Review |
map-systems | Map Systems |
prototype | Prototype |
team-combat | Team Combat |
team-narrative | Team Narrative |
team-level | Team Level |
team-ui | Team Ui |
team-audio | Team Audio |
team-polish | Team Polish |
team-release | Team Release |
estimate | Estimate |
scope-check | Scope Check |
gate-check | Gate Check |
milestone-review | Milestone Review |
asset-audit | Asset Audit |
balance-check | Balance Check |
perf-profile | Perf Profile |
playtest-report | Playtest Report |
bug-report | Bug Report |
tech-debt | Tech Debt |
hotfix | Hotfix |
changelog | Changelog |
patch-notes | Patch Notes |
release-checklist | Release Checklist |
launch-checklist | Launch Checklist |
localize | Localize |
reverse-document | Reverse Document |
architecture-decision | Architecture Decision |
project-stage-detect | Project Stage Detect |
onboard | Create or improve onboarding experiences that help users understand, adopt, and |
/content-creation-swarm论文撰写、文档生成、技术写作、翻译、润色、字幕、发布
触发词: write, article, paper, translate, proofread, document, content, subtitle writing, 撰写, 文章, 论文写作, 翻译, 润色, 字幕写作
| Skill | 说明 |
|---|---|
baoyu-translate | Baoyu Translate |
codegen-diagram | Codegen Diagram |
codegen-doc | Codegen Doc |
compliance-docs | Compliance Docs |
document | > **AI 驱动的文档处理系统** - 支持Word、PDF、Excel、Markdown等文档的生成与处理 |
grammar-check | Grammar Check |
makeownsrt | Makeownsrt |
md-report-summary | Md Report Summary |
nlp | > **AI 驱动的文本处理系统** - 支持情感分析、实体识别、关键词提取、文本摘要 |
paper-write | Paper Write |
post-to-wechat | 该 Skill 将 [`JimLiu/baoyu-skills`](https://github.com/JimLiu/baoyu-skills) 的 **po |
prompt-engineering-100 | 这个 Skill 的目标是:把你的一句话需求,自动转成 **可执行、可验证、可复用** 的 Prompt。 |
prompt-fill | 把 Prompt Fill 的变量填空与一键优化流程固化为可复用 runbook 与组件文档。 |
qiaomu-x-article-publisher | Qiaomu X Article Publisher |
qiaomu-x-article-publisher-github | Qiaomu X Article Publisher Github |
rateyourmusic-genres | 把 RateYourMusic 的 genres 页面整理为可复用研究 runbook,输出结构化的流派与子流派清单。 |
research-paper-writing | Research Paper Writing |
skill-prompt-convert | Skill Prompt Convert |
summarize | Summarize |
wechat-article-search | Wechat Article Search |
wechat-article-writer | Wechat Article Writer |
wechat-reader | Wechat Reader |
x-reader | > Any URL -> Platform Detection -> Fetch Content -> Unified Output |
ljg-word | Ljg Word |
ljg-explain-words | Ljg Explain Words |
ljg-word-flow | Ljg Word Flow |
ljg-title | Ljg Title |
ljg-writes | Ljg Writes |
ljg-plain | Ljg Plain |
codex-doc | Codex Doc |
codex-pdf | Codex Pdf |
/pm-execution-swarmPRDs, sprints, stories, launch plans, release notes, OKRs
触发词: prd, sprint, okr, launch, user stories, release notes, product, pm, feature, backlog, triage, 产品, 需求, 迭代
| Skill | 说明 |
|---|---|
brainstorm-experiments-existing | Brainstorm Experiments Existing |
brainstorm-experiments-new | Brainstorm Experiments New |
brainstorm-ideas-existing | Brainstorm Ideas Existing |
brainstorm-ideas-new | Brainstorm Ideas New |
brainstorm-okrs | Brainstorm Okrs |
competitor-analysis | Competitor Analysis |
compliance | - 把你的合规需求描述,转换为 **可审计/可复核** 的交付骨架:目标、约束、政策映射、验证方法、上线条件。 |
create-prd | Create Prd |
gtm-motions | Gtm Motions |
pm-cmd-analyze-test | Pm Cmd Analyze Test |
pm-cmd-brainstorm | Pm Cmd Brainstorm |
pm-cmd-generate-data | Pm Cmd Generate Data |
pm-cmd-meeting-notes | Pm Cmd Meeting Notes |
pm-cmd-plan-launch | Pm Cmd Plan Launch |
pm-cmd-plan-okrs | Pm Cmd Plan Okrs |
pm-cmd-privacy-policy | Pm Cmd Privacy Policy |
pm-cmd-proofread | Pm Cmd Proofread |
pm-cmd-sprint | Pm Cmd Sprint |
pm-cmd-test-scenarios | Pm Cmd Test Scenarios |
pm-cmd-triage-requests | Pm Cmd Triage Requests |
pm-cmd-write-prd | Pm Cmd Write Prd |
pm-cmd-write-query | Pm Cmd Write Query |
pm-cmd-write-stories | Pm Cmd Write Stories |
privacy-policy | Privacy Policy |
product-tri-ownership | 将“产品三人组(Product Owner / Tech Owner / Quality Owner)+ AI Agents 执行”的协作框架落盘为可复用资产, |
release-notes | Release Notes |
rpi-implement | Rpi Implement |
rpi-plan | Rpi Plan |
test-scenarios | Test Scenarios |
tri-ownership | 把“头狼+狼群”的端到端工作拆分为 PO/TO/QO 三角色协作,并输出可执行的流程资产,提升 AI 时代的交付质量与节奏。 |
wwas | Wwas |
/research-learning-swarm知识学习、论文研究、自动迭代、RSS 聚合、知识站点、笔记本
触发词: learn, knowledge research, knowledge, rss, digest, notebooklm, autoresearch, 学习, 研究, 知识, 论文研究, obsidian, 学习材料, OREAS
| Skill | 说明 |
|---|---|
ak-rss-digest | Ak Rss Digest |
anthropic-official-skills | 该 Skill 用于把 Anthropic 官方仓库的 Claude Skills 套件安装到你的项目(或个人目录)中,便于 Claude Code 自动加载使 |
anything-to-notebooklm | 把多源内容(微信/网页/YouTube/PDF/Markdown/搜索)接入 NotebookLM 的流程固化为可复用 runbook 与组件文档。 |
bestblogs-daily-digest | Bestblogs Daily Digest |
bookauthority-reading | 把 BookAuthority 的权威非虚构书单筛选流程固化为可复用 runbook,帮助快速定位顶级思想领袖书单与垂直榜单。 |
continuous-learning-v2 | Continuous Learning V2 |
ebook-to-notebooklm | 把合法来源的电子书导入 NotebookLM,生成思维导图、音频与视频脚本。 |
github-skill-arsenal | 把 GitHub 开源项目系统化封装成 Skills,形成可复用的超级技能库。 |
karpathy-autoresearch | Karpathy Autoresearch |
knowledge-site-creator | Knowledge Site Creator |
learning-digest-v2 | Learning Digest V2 |
notebooklm | Interact with Google NotebookLM to query documentation with Gemini's source-grou |
obsidian-bases | Obsidian Bases |
obsidian-markdown | Obsidian Markdown |
onecontext | Onecontext |
skill-from-masters | 把 GitHub 上成熟项目的“核心知识”抽象为可复用技能,而不是简单包装工具。 |
summarize-interview | Summarize Interview |
summarize-meeting | Summarize Meeting |
verification-loop | Verification Loop |
ljg-paper | Ljg Paper |
ljg-paper-flow | Ljg Paper Flow |
ljg-xray-paper | Ljg Xray Paper |
ljg-xray-book | Ljg Xray Book |
ljg-xray-article | Ljg Xray Article |
ljg-learn | Ljg Learn |
ljg-explain-concept | Ljg Explain Concept |
ljg-rank | Ljg Rank |
ljg-the-one | Ljg The One |
ljg-travel | Ljg Travel |
/ljg-cognitive-swarmCogOS 8-layer cognitive skill system: Intake→Decompose→Synthesize→Create→Reflect→Meta. 约束驱动设计,认知碰撞理念
触发词: ljg, 刘力刚, 认知, 降秩, X光, xray, 铸, cast, 白话, plain, 解剖, 拆书, 论文X光, 概念解剖, 单词解剖, 信息图, 阅读卡, 剪藏, 圆桌, 本质, 投资分析, 关系分析, 旅行研究, 技能拓扑, skill map, 标题
| Skill | 说明 |
|---|---|
ljg-fetch | Ljg Fetch |
ljg-clip | Ljg Clip |
ljg-x-download | Ljg X Download |
ljg-xray-paper | Ljg Xray Paper |
ljg-xray-book | Ljg Xray Book |
ljg-xray-article | Ljg Xray Article |
ljg-xray-skill | Ljg Xray Skill |
ljg-xray-prompt | Ljg Xray Prompt |
ljg-paper | Ljg Paper |
ljg-paper-flow | Ljg Paper Flow |
ljg-learn | Ljg Learn |
ljg-explain-concept | Ljg Explain Concept |
ljg-explain-words | Ljg Explain Words |
ljg-word | Ljg Word |
ljg-word-flow | Ljg Word Flow |
ljg-rank | Ljg Rank |
ljg-the-one | Ljg The One |
ljg-plain | Ljg Plain |
ljg-roundtable | Ljg Roundtable |
ljg-invest | Ljg Invest |
ljg-relationship | Ljg Relationship |
ljg-card | Ljg Card |
ljg-cards | Ljg Cards |
ljg-infograph | Ljg Infograph |
ljg-title | Ljg Title |
ljg-writes | Ljg Writes |
ljg-travel | Ljg Travel |
ljg-calendar | Ljg Calendar |
ljg-skill-map | Ljg Skill Map |
/finance-tax-swarm29-Skill finance/tax expert team for bookkeeping, compliance, and advisory. Router + 27 specialists across 3 phases.
触发词: 财税, 税务, 会计, 记账, 申报, 发票, 社保, 公积金, 增值税, 企业所得税, 个人所得税, 合规, 审计, 汇算清缴, 工商, 注册, 注销, 代账, 税收优惠, 税务筹划, 出口退税, 转让定价, 关税, 印花税, 房产税, 土增税, 合并报表, 外币, 内审, 成本核算, 政府补贴, finance, tax, accounting, compliance, bookkeeping, CPA, CTA
| Skill | 说明 |
|---|---|
team-finance-tax | 意图分类 -> 专家分发 -> 协作编排 -> 结果综合。所有财税类问题的入口,根据用户意图路由到最合适的单一专家或编排多专家协作。简单事实类问题(如\"增值税 |
ft-business-analyst | 理解客户经营场景 -> 识别涉税事项 -> 拆解业务流 -> 输出结构化业务事件给会计师和税务师。作为财税处理的第一环节,确保后续专家拿到的是准确的业务事实而非 |
ft-accountant | 凭证录入 / 科目映射 / 期末结账 / 报表编制 / CAS 准则适用。将业务分析师输出的结构化业务事件翻译为准确的会计语言(借贷分录),确保账务处理符合企业 |
ft-tax-advisor | 18 税种判定 / 税率适用 / 纳税申报(增值税+企业所得税+个税+附加税) / 政策解读。负责将会计师的账务数据转化为准确的纳税申报,确保依法纳税、应享尽享 |
ft-compliance-auditor | 风险识别 / 合规规则匹配 / 金税四期预警应对 / 账务质检 / 整改建议。作为财税处理的最后一道防线,在会计师做完账、税务师算完税之后进行全面检查,确保业财 |
ft-social-insurance | 五险一金核算 / 基数年度调整(每年7月) / 工伤生育报销 / 残保金申报 / 灵活用工方案。确保企业在社保公积金领域的合规性,同时合理利用政策降低用工成本。 |
ft-business-registration | 企业注册 / 工商变更 / 年报年检 / 简易注销 / 经营范围调整 / 股东变更。处理企业全生命周期中与市场监督管理局相关的所有行政事务,并协调变更对税务和社 |
ft-invoice-manager | 进项认证/勾选抵扣/数电票(全电发票)/异常发票排查/进项转出判定/留抵退税申报/发票真伪验证。代理记账公司日常接触最多的税务凭证环节,直接影响客户增值税税负和 |
ft-regulation-lookup | KG MCP tool 的消费者。教模型何时查询知识图谱、如何构造查询、怎样解读返回结果。不做法规解读或税务筹划,只负责精准检索并返回结构化法规数据。 |
ft-compliance-checklist | 按行业+企业规模+纳税人类型自动生成结构化合规检查清单。生成清单后交给 ft-compliance-auditor 执行检查。本技能只负责\"列出该查什么\", |
ft-period-calendar | 申报截止日计算/节假日延期规则/提醒生成。代理记账公司最基础也最容易出错的工具——逾期申报的滞纳金和罚款是完全可以避免的损失。 |
ft-financial-statement | 三大报表编制/关键比率计算/报表间勾稽关系验证/报表异常识别。代理记账最终输出的核心产品——报表质量直接反映记账水平。 |
ft-tax-accounting-gap | 50+常见税会差异的识别/分类(永久性vs暂时性)/调整方向(调增vs调减)/递延所得税计算。汇算清缴季(1-5月)最核心的专业能力——差异识别不准确直接导致多 |
ft-tax-planner | 企业税务筹划方案设计专家。负责优惠政策匹配、组织架构税务优化、节税方案测算、合法性边界判断。核心能力是在合法筹划与避税/偷税之间划出清晰的红线,确保每个方案都经 |
ft-international-tax | 跨境交易税务处理与合规专家。负责非居民企业所得税代扣代缴、税收协定优惠适用、境外税收抵免计算、常设机构判定、转让定价同期资料准备。核心能力是在复杂的跨境交易结构 |
ft-legal-counsel | 税务争议解决与涉税法律风险防控专家。负责行政复议策略设计、税务稽查应对、涉税刑事风险预判、行政处罚裁量分析、税务行政诉讼代理。核心能力是在税务争议的每个阶段帮助 |
ft-export-tax-rebate | 出口货物退(免)税政策适用与申报专家。负责免退/免抵退方式选择、退税率查询与差额处理、出口收入确认与汇率换算、单证备案管理、进料加工手册核销。核心能力是确保企业 |
ft-cost-accounting | 企业成本核算体系设计与执行专家。负责工业/商贸/服务业成本模型搭建、制造费用分配方案设计、标准成本差异分析、存货计价方法选择、在产品成本计算。核心能力是确保成本 |
ft-tax-incentive | 企业税收优惠政策条件判定与申报执行专家。负责高新技术企业认定辅导、小型微利企业标准判定、研发费加计扣除范围划定、软件企业增值税即征即退测算、加速折旧方案设计、优 |
ft-risk-assessment | 企业税务合规风险评估与预警专家。负责行业基准对标分析、异常指标自动检测、风险等级量化评估、预警报告生成与整改建议输出。核心能力是在税务机关发现问题之前帮助企业识 |
team-audit | 审计团队路由器,负责编排内部审计、合并报表、外币核算、转让定价等高级财务工作。根据业务场景将任务分配给4个专家角色,并监控审计全流程(计划-执行-报告-跟踪)的 |
ft-internal-audit | 内部审计专家角色,负责企业内控评价、业务流程审计、舞弊风险识别与应对。基于COSO内部控制框架,对关键业务流程的控制有效性进行评估,识别控制缺陷并提出整改建议。 |
ft-consolidation | 合并报表专家角色,负责集团合并财务报表编制全流程。核心工作包括合并范围判定、内部交易抵消、少数股东权益计算、权益法与成本法核算,以及同一控制与非同一控制下企业合 |
ft-fx-accounting | 外币核算专家角色,负责企业外币交易的初始确认、期末调整、汇兑损益计算,以及境外子公司外币报表折算。核心能力包括功能货币判定、货币性与非货币性项目区分、套期会计处 |
ft-transfer-pricing | 转让定价专家角色,负责关联交易定价的合规管理。核心工作包括独立交易原则的应用、五种定价方法的选择、同期资料三层文档的准备、预约定价安排(APA)的申请与谈判,以 |
ft-subsidy-application | 政府补贴申报专家角色,负责各类财政补贴的项目筛选、申报材料准备、评审应对和资金使用合规管理。核心能力覆盖研发补助、人才引进、产业升级、知识产权奖励、专精特新认定 |
ft-stamp-property-tax | 小税种专家角色,负责印花税、房产税、城镇土地使用税、契税、土地增值税、车船税、环保税等非主体税种的计算、申报和合规管理。虽然单个税种金额不大,但种类多、规则杂、 |
ft-customs-duty | 关税消费税专家角色,负责进出口环节的关税计算、进口增值税和消费税核算、HS编码归类、保税区政策运用和加工贸易管理。确保企业进出口税费计算准确、关务合规,并充分利 |
ft-digital-asset-tax | 数字资产与ESG税务专家角色,负责数据资产入表、AI模型成本归集与资本化判定、碳排放配额会计处理、碳交易损益核算、环保税合规申报,以及ESG信息披露中的税务关联 |
/pm-research-swarmUser research, market analysis, competitive intelligence, personas, journey maps
触发词: user research, market analysis, competitive, persona, interview, discovery, market scan, lenny, 用户研究, 市场分析, 竞品
| Skill | 说明 |
|---|---|
competitive-battlecard | Competitive Battlecard |
customer-journey-map | Customer Journey Map |
draft-nda | Draft Nda |
ideal-customer-profile | Ideal Customer Profile |
interview-script | Interview Script |
job-stories | Job Stories |
market-segments | Market Segments |
market-sizing | Market Sizing |
marketing-ideas | Marketing Ideas |
marketing-skills | Marketing Skills |
pm-cmd-analyze-cohorts | Pm Cmd Analyze Cohorts |
pm-cmd-analyze-feedback | Pm Cmd Analyze Feedback |
pm-cmd-battlecard | Pm Cmd Battlecard |
pm-cmd-competitive-analysis | Pm Cmd Competitive Analysis |
pm-cmd-discover | Pm Cmd Discover |
pm-cmd-draft-nda | Pm Cmd Draft Nda |
pm-cmd-interview | Pm Cmd Interview |
pm-cmd-market-product | Pm Cmd Market Product |
pm-cmd-market-scan | Pm Cmd Market Scan |
pm-cmd-research-users | Pm Cmd Research Users |
pm-cmd-review-resume | Pm Cmd Review Resume |
pm-cmd-tailor-resume | Pm Cmd Tailor Resume |
review-resume | Review Resume |
rpi-research | Rpi Research |
user-personas | User Personas |
user-segmentation | User Segmentation |
user-stories | User Stories |
/devops-infra-swarmCI/CD, deployment, monitoring, security, infrastructure patrol, release management
触发词: deploy, ci/cd, monitoring, security, infrastructure, release, oncall, alert, docker, migration, 部署, 基建, 监控, 告警, 迁移
| Skill | 说明 |
|---|---|
babysit-pr | Babysit Pr |
data-pipeline | Data Pipeline |
deploy-preview | Deploy Preview |
docker-optimizer | Docker Optimizer |
enterprise-agent-ops | Enterprise Agent Ops |
infra-patrol | Infra Patrol |
new-migration | New Migration |
observability-setup | Observability Setup |
oncall-runner | Oncall Runner |
orphan-cleanup | Orphan Cleanup |
pipeline-triage | Pipeline Triage |
release-skills | Release Skills |
security-best-practices | Security Best Practices |
standup-post | Standup Post |
tailscale-network | Manage Tailscale mesh VPN from AI-tools: auto-connect on session start, check ne |
codex-gh-fix-ci | Codex Gh Fix Ci |
codex-security-ownership-map | Codex Security Ownership Map |
codex-security-threat-model | Codex Security Threat Model |
codex-cloudflare-deploy | Codex Cloudflare Deploy |
vmo-wrangler | 把 VMO Wrangler 的初始化流程固化为可复用 runbook,覆盖认证、模型提供商配置、项目创建与 Claude 插件安装。 |
weekly-recap | Weekly Recap |
lovable-ai-builder | Lovable Ai Builder |
/lark-suite-swarm飞书全套: 消息/文档/表格/日历/任务/知识库/会议
触发词: 飞书, lark, feishu, 多维表格, 知识库, 日程, 群消息
| Skill | 说明 |
|---|---|
lark-shared | 本技能指导你如何通过lark-cli操作飞书资源, 以及有哪些注意事项。 |
lark-im | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-doc | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-sheets | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-base | > **前置条件:** 先阅读 [`../lark-shared/SKILL.md`](../lark-shared/SKILL.md)。 > **执行前必做: |
lark-calendar | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-task | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-drive | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-wiki | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-mail | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-contact | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-vc | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-minutes | > **前置条件:** 先阅读 [`../lark-shared/SKILL.md`](../lark-shared/SKILL.md) 了解认证、全局参数和安 |
lark-whiteboard | > [!NOTE] > **环境依赖**:绘制画板需要 `@larksuite/whiteboard-cli`(画板 Node.js CLI 工具),以及 `l |
lark-event | > **前置条件:** 先阅读 [`../lark-shared/SKILL.md`](../lark-shared/SKILL.md) 了解认证、权限处理和安 |
lark-openapi-explorer | > **前置条件:** 先阅读 [`../lark-shared/SKILL.md`](../lark-shared/SKILL.md) 了解认证、身份切换和安 |
lark-skill-maker | 基于 lark-cli 创建新 Skill。Skill = 一份 `SKILL.md`,教 AI 用 CLI 命令完成任务。 |
lark-workflow-meeting-summary | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
lark-workflow-standup-report | **CRITICAL — 开始前 MUST 先用 Read 工具读取 [`../lark-shared/SKILL.md`](../lark-shared/SK |
/data-analytics-swarm数据分析、A/B 测试、用户分群、情感分析、SQL 查询、仪表盘
触发词: analytics, data, ab test, cohort, sql, sentiment, excel, 数据分析, A/B, 报表
| Skill | 说明 |
|---|---|
ab-test-analysis | Ab Test Analysis |
analyze-feature-requests | Analyze Feature Requests |
cohort-analysis | Cohort Analysis |
data | > **AI 驱动的数据分析系统** - 支持数据清洗、统计分析、可视化、SQL生成 |
database-designer | Database Designer |
dummy-dataset | Dummy Dataset |
excel-analysis | Excel Analysis |
invoice-ocr | 把“发票/小票/收据”的图片或 PDF,抽取为**结构化 JSON**,并做**质量评分 + 数学一致性校验 + 自动修复**。 |
postgresql-best-practices | Postgresql Best Practices |
sentiment-analysis | Sentiment Analysis |
sql-queries | Sql Queries |
ljg-invest | Ljg Invest |
ljg-relationship | Ljg Relationship |
codex-spreadsheet | Codex Spreadsheet |
codex-jupyter-notebook | Codex Jupyter Notebook |
/web-search-swarm多引擎检索、实时搜索、趋势查询、全网聚合、深度研究
触发词: search, find, lookup, web research, osint, 搜索, 检索, 查找, 调研
| Skill | 说明 |
|---|---|
ask-search | Ask Search |
deep-research | 基于 [Claude-Code-Deep-Research](https://github.com/AnkitClassicVision/Claude-Code |
grok-search | Grok Search |
grok-twitter-search | Grok Twitter Search |
infinisynapse-market | 把 InfiniSynapse Tool Market 的工具下载与 Agent 调用流程固化为可复用 runbook 与组件文档。 |
multi-search-engine | Multi Search Engine |
nuwa-world-osint-human-research | Nuwa World Osint Human Research |
osint-framework | 基于 [`lockfale/OSINT-Framework`](https://github.com/lockfale/OSINT-Framework) 的 ` |
search-first | Search First |
tavily-search | Tavily Search |
trends24 | Trends24 |
web-multi-search | Web Multi Search |
/advertising-swarmFull-funnel marketing: avatar extraction → messaging → creative → conversion
触发词: paid marketing, advertising, ad copy, campaign, funnel, conversion, headline
| Skill | 说明 |
|---|---|
avatar-extraction | You are a buyer psychologist. Your job is to define a real, specific person — no |
schwartz-awareness-mapper | You are a direct response strategist trained in Eugene Schwartz frameworks. |
offer-extraction | You are an offer strategist. You make the offer feel like a shortcut to a desire |
mechanism-builder | You are responsible for creating the “why this works” story. |
headline-matrix | You create attention and curiosity. |
ad-angle-multiplier | You multiply winning ideas. |
objection-crusher | You remove resistance to action. |
scroll-stopping-creative | You design for attention. |
conversion-path-builder | You reduce friction and increase conversion. |
full-funnel-campaign-orchestrator | You orchestrate specialists. |
generic-language-killer | You remove weak language. |
performance-diagnosis | You find the bottleneck. |
/algorithm-data-swarm算法工程、深度学习、大数据处理、可视化、量化策略
触发词: algorithm, deep learning, pytorch, ml, data science, bigdata, polars, dask, quantitative, trading, 算法, 深度学习, 大数据, 量化
| Skill | 说明 |
|---|---|
algo-core | Algo Core |
algo-dl | Algo Dl |
art | - 输入一句话/一段描述,生成 **可复现** 的 SVG 图形并落盘(不依赖第三方包)。 |
bigdata-core | Bigdata Core |
bigdata-ml | Bigdata Ml |
bigdata-viz | Bigdata Viz |
quantum-market-research | Analyze the current crypto market state for the Quantum X autonomous trading sys |
quantum-risk-report | Generate daily risk monitoring report for the Quantum X trading system. |
quantum-strategy-evolve | Drive the self-evolution loop for Quantum X trading strategies. |
rag-architect | Rag Architect |
valq-powerbi-planning | ValQ 的定位(从官网信息抽象): - 做 **sales/finance/operations** 场景规划、预算与预测 - 支持 **what-if si |
/animation-swarmGSAP + Remotion + Framer Motion animation skills for web and video
触发词: animation, gsap, timeline, scroll-trigger, draggable, motion, animate, tween
| Skill | 说明 |
|---|---|
gsap-core | Gsap Core |
gsap-timeline | Gsap Timeline |
gsap-scrolltrigger | Gsap Scrolltrigger |
gsap-plugins | Gsap Plugins |
gsap-react | Gsap React |
gsap-frameworks | Gsap Frameworks |
gsap-performance | Gsap Performance |
gsap-utils | Gsap Utils |
remotion-best-practices | Use this skills whenever you are dealing with Remotion code to obtain the domain |
jh3y-hover-cards | > 将 `jh3y` 的 CodePen 效果「context aware :hover cards w/ blur」做成可复用 Skill:**抓取缓存** |
/wechat-cn-social-swarm多平台 IM 群聊摘要管线(微信/企业微信/飞书)+ 小红书 + 公众号 + 朋友圈
触发词: wechat, 微信, 企业微信, 企微, wecom, 飞书群聊, feishu group, lark chat, 小红书, xiaohongshu, 公众号, telegram, discord, im, 群聊, 朋友圈, IM群消息, 微信群, 飞书群, 工作群
| Skill | 说明 |
|---|---|
claude-to-im | Claude To Im |
feishu-group | Feishu Group |
pi-mono | 把 pi-mono 的“七包分治”思路,以 **repo-first** 的方式落到你的项目中: |
wechat-db | Wechat Db |
wechat-decrypt | Wechat Decrypt |
wechat-multi | Wechat Multi |
wechat-report | Wechat Report |
wechat-sns | Wechat Sns |
wecom-db | Wecom Db |
xiaohongshu-automation | Xiaohongshu Automation |
/icon-design-swarm图标/视觉资产生成: macOS icon + 海报 + Canvas + 图标库检索
触发词: icon, 图标, app icon, macos icon, poster, 海报, logo, visual asset, 封面
| Skill | 说明 |
|---|---|
macos-icon-gen | Macos Icon Gen |
qiaomu-mondo-poster-design | Qiaomu Mondo Poster Design |
canvas-design | Canvas Design |
nanobanana-image-gen | Nanobanana Image Gen |
better-icons | Better Icons |
lobe-icons | 你想要的是:把 LobeHub 的 AI/LLM 模型图标 **一键下载到本地**,并提供一个“图标库 Skill”用于 **搜索 / 导出 / 复用**。 |
svglogo-icons | 来源:[`https://svglogo.top/?view=original`](https://svglogo.top/?view=original) |
/baseDefault fallback group for unclassified tasks. Contains core utility skills for planning, review, and commit.
触发词:
| Skill | 说明 |
|---|---|
planning-with-files | 用本地三份文件作为“外置工作记忆”,避免长任务在对话变长后发生目标漂移: |
code-review | Code Review |
git-commit | Creates git commits following Conventional Commits format with proper type, scop |
brainstorming | 1. **Don't skip brainstorming for \"simple\" tasks.** The most expensive bugs co |
systematic-debugging | **NO FIXES WITHOUT ROOT CAUSE INVESTIGATION FIRST.** |
verification-before-completion | **NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE.** |
reflection | 这个 Skill 用于“阶段性反思/复盘/沉淀”:把你在任务过程中遇到的问题、修复思路、验证方式与下一步计划,按统一结构写入文件,避免信息散落在对话里不可审计、 |
管线:A->B->C 固定顺序执行,每步依赖上步结果。
/impeccable-pipelineProgressive UI/UX design refinement: audit → critique → fix → polish
触发词: UI design, UX polish, design system, visual refinement, layout
执行顺序:critique -> arrange -> typeset -> polish (最多 3 轮)
| Skill | 说明 |
|---|---|
teach-impeccable | Gather design context for this project, then persist it for all future sessions. |
frontend-design | This skill guides creation of distinctive, production-grade frontend interfaces |
critique | Conduct a holistic design critique, evaluating whether the interface actually wo |
audit | Run systematic quality checks and generate a comprehensive audit report with pri |
arrange | Assess and improve layout and spacing that feels monotonous, crowded, or structu |
colorize | Strategically introduce color to designs that are too monochromatic, gray, or la |
typeset | Assess and improve typography that feels generic, inconsistent, or poorly struct |
clarify | Identify and improve unclear, confusing, or poorly written interface text to mak |
animate | Analyze a feature and strategically add animations and micro-interactions that e |
adapt | Adapt existing designs to work effectively across different contexts - different |
harden | Strengthen interfaces against edge cases, errors, internationalization issues, a |
delight | Identify opportunities to add moments of joy, personality, and unexpected polish |
distill | Remove unnecessary complexity from designs, revealing the essential elements and |
bolder | Increase visual impact and personality in designs that are too safe, generic, or |
quieter | Reduce visual intensity in designs that are too bold, aggressive, or overstimula |
overdrive | Start your response with: |
normalize | Analyze and redesign the feature to perfectly match our design system standards, |
extract | Identify reusable patterns, components, and design tokens, then extract and cons |
optimize | Identify and fix performance issues to create faster, smoother user experiences. |
polish | **First**: Use the frontend-design skill for design principles and anti-patterns |
/browser-automation-pipeline浏览器自动化、表单填充、网页交互、截图、DOM 操作、桌面自动化
触发词: browser, automation, playwright, screenshot, form fill, dom, ghost, desktop, mobile, device, 浏览器, 自动化, 截图, 表单
| Skill | 说明 |
|---|---|
agent-browser-electron | Agent Browser Electron |
agent-browser-session | Agent Browser Session |
agent-browser-slack | Agent Browser Slack |
agent-reach | Install and configure upstream tools for 12+ platforms. After setup, call them d |
android-device-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
browser-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
chrome-bridge-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
continue-workflow | 固化 /continue 会话延续流程,降低上下文重建成本,提高任务完全达成率。 |
defuddle | Defuddle |
desktop-computer-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
flutter-development | Flutter Development |
ghost-os-automation | Ghost Os Automation |
harmonyos-device-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
ios-device-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
playwright-codex | Playwright Codex |
playwright-interactive | Playwright Interactive |
screenshot-codex | Screenshot Codex |
vphone-cli | Vphone Cli |
xiaohongshu-automation | Xiaohongshu Automation |
/web-scraping-pipelineWeb 数据采集、多平台爬虫、社交媒体数据、视频下载、内容提取
触发词: scrape, crawl, extract, download, apify, 爬虫, 采集, 提取, 下载
| Skill | 说明 |
|---|---|
agent-fetch | Agent Fetch |
agentic-seo | Agentic Seo |
ai-news | 把多个 AI 资讯来源统一聚合到一个本地文件,支持增量更新、标题中文翻译与定时运行。 |
apify-competitor-intelligence | Apify Competitor Intelligence |
apify-trend-analysis | Apify Trend Analysis |
apify-ultimate-scraper | Apify Ultimate Scraper |
downloadhd | Downloadhd |
firecrawl-skill | Firecrawl Skill |
news-aggregator-skill | Fetch real-time hot news from 28 sources, generate deep analysis reports in Chin |
pdf-layout-analysis | 基于 [`huridocs/pdf-document-layout-analysis`](https://github.com/huridocs/pdf-doc |
pdf-processing | Pdf Processing |
scrapling-crawler | This skill provides an advanced web scraping utility built on the [Scrapling](ht |
twitter-web-exporter | 将 Twitter/X Web 导出工具固化为可复用的 runbook 与组件文档,支持团队在浏览器内完成推文/书签/列表等数据导出。 |
video-download | Download video and audio from URLs using yt-dlp directly. No wrapper scripts nee |
x-tweet-fetcher | X Tweet Fetcher |
youtube-clipper-skill | 把 YouTube Clipper 的剪辑与字幕流程固化为可复用 runbook 与组件文档。 |
ljg-fetch | Ljg Fetch |
ljg-x-download | Ljg X Download |
/ontology-audit-swarm3-round 17-expert swarm for knowledge graph ontology design audit. Round 1: 6 strategic advisors (Hickey/Drucker/Munger/Brooks/Meadows/Research). Round 2: 5 domain experts (Tax Law/Accounting CAS/Practice/KG Design/Tax Risk). Round 3: 6 business deep-dive (CPA/Planning/Lifecycle/Financial Analysis/Cross-border/Payroll).
触发词: 本体设计, 本体审计, ontology, 知识图谱架构, KG设计, 蜂群审计, 专家团审查
| Skill | 说明 |
|---|---|
team-finance-tax | 意图分类 -> 专家分发 -> 协作编排 -> 结果综合。所有财税类问题的入口,根据用户意图路由到最合适的单一专家或编排多专家协作。简单事实类问题(如\"增值税 |
ft-tax-advisor | 18 税种判定 / 税率适用 / 纳税申报(增值税+企业所得税+个税+附加税) / 政策解读。负责将会计师的账务数据转化为准确的纳税申报,确保依法纳税、应享尽享 |
ft-accountant | 凭证录入 / 科目映射 / 期末结账 / 报表编制 / CAS 准则适用。将业务分析师输出的结构化业务事件翻译为准确的会计语言(借贷分录),确保账务处理符合企业 |
ft-compliance-auditor | 风险识别 / 合规规则匹配 / 金税四期预警应对 / 账务质检 / 整改建议。作为财税处理的最后一道防线,在会计师做完账、税务师算完税之后进行全面检查,确保业财 |
ft-risk-assessment | 企业税务合规风险评估与预警专家。负责行业基准对标分析、异常指标自动检测、风险等级量化评估、预警报告生成与整改建议输出。核心能力是在税务机关发现问题之前帮助企业识 |
ft-tax-planner | 企业税务筹划方案设计专家。负责优惠政策匹配、组织架构税务优化、节税方案测算、合法性边界判断。核心能力是在合法筹划与避税/偷税之间划出清晰的红线,确保每个方案都经 |
ft-international-tax | 跨境交易税务处理与合规专家。负责非居民企业所得税代扣代缴、税收协定优惠适用、境外税收抵免计算、常设机构判定、转让定价同期资料准备。核心能力是在复杂的跨境交易结构 |
ft-legal-counsel | 税务争议解决与涉税法律风险防控专家。负责行政复议策略设计、税务稽查应对、涉税刑事风险预判、行政处罚裁量分析、税务行政诉讼代理。核心能力是在税务争议的每个阶段帮助 |
ft-social-insurance | 五险一金核算 / 基数年度调整(每年7月) / 工伤生育报销 / 残保金申报 / 灵活用工方案。确保企业在社保公积金领域的合规性,同时合理利用政策降低用工成本。 |
ft-financial-statement | 三大报表编制/关键比率计算/报表间勾稽关系验证/报表异常识别。代理记账最终输出的核心产品——报表质量直接反映记账水平。 |
ft-tax-accounting-gap | 50+常见税会差异的识别/分类(永久性vs暂时性)/调整方向(调增vs调减)/递延所得税计算。汇算清缴季(1-5月)最核心的专业能力——差异识别不准确直接导致多 |
ft-transfer-pricing | 转让定价专家角色,负责关联交易定价的合规管理。核心工作包括独立交易原则的应用、五种定价方法的选择、同期资料三层文档的准备、预约定价安排(APA)的申请与谈判,以 |
ft-cost-accounting | 企业成本核算体系设计与执行专家。负责工业/商贸/服务业成本模型搭建、制造费用分配方案设计、标准成本差异分析、存货计价方法选择、在产品成本计算。核心能力是确保成本 |
ft-invoice-manager | 进项认证/勾选抵扣/数电票(全电发票)/异常发票排查/进项转出判定/留抵退税申报/发票真伪验证。代理记账公司日常接触最多的税务凭证环节,直接影响客户增值税税负和 |
ft-business-analyst | 理解客户经营场景 -> 识别涉税事项 -> 拆解业务流 -> 输出结构化业务事件给会计师和税务师。作为财税处理的第一环节,确保后续专家拿到的是准确的业务事实而非 |
ft-business-registration | 企业注册 / 工商变更 / 年报年检 / 简易注销 / 经营范围调整 / 股东变更。处理企业全生命周期中与市场监督管理局相关的所有行政事务,并协调变更对税务和社 |
ft-internal-audit | 内部审计专家角色,负责企业内控评价、业务流程审计、舞弊风险识别与应对。基于COSO内部控制框架,对关键业务流程的控制有效性进行评估,识别控制缺陷并提出整改建议。 |
/video-media-pipelineAI 视频生成、视频编辑、字幕、语音合成、头像视频、换脸
触发词: video, media, tts, avatar, heygen, subtitle, audio, speech, 视频, 音频, 语音, 字幕
| Skill | 说明 |
|---|---|
ai-video-gen | Generate AI videos from text prompts. Supports multiple providers (VEO 3.1, Klin |
avatar-video | Create AI avatar videos with full control over avatars, voices, scripts, scenes, |
create-video | Generate complete videos from a text prompt. Describe what you want and the AI h |
faceswap | Swap a face from a source image into a target video using GPU-accelerated AI pro |
heygen | > **This skill is deprecated.** Use the focused skills instead: > - **`create-vi |
libtv-skill | 通过 agent-im 的 OpenAPI 创建会话、发送消息(生图、生视频、编辑视频等)、上传图片/视频文件,并查询会话消息进展。 |
opencode-glm47-remotion | 把 OpenCode + GLM-4.7 + Remotion 的高燃短视频剪辑流程固化为可复用 runbook 与组件文档。 |
pdf2video | 把 PDF 转视频的 Remotion 工作流固化为可复用 runbook 与组件文档。 |
text-to-speech | Generate speech audio files from text using HeyGen's in-house Starfish TTS model |
video-constrict | 基于 [Constrict](https://github.com/Wartybix/Constrict) 的视频压缩能力,将视频压缩到指定的目标文件大小。非常 |
video-download | Download video and audio from URLs using yt-dlp directly. No wrapper scripts nee |
video-edit | Edit videos locally by running ffmpeg/ffprobe directly. No wrapper scripts neede |
video-translate | Translate and dub existing videos into multiple languages, preserving lip-sync a |
video-understand | Understand video content locally using ffmpeg for frame extraction and Whisper f |
yt-search-download | Yt Search Download |
/content-publish-pipeline内容全渠道发布: 写作 -> 排版 -> 推送 (公众号 + X + Markdown)
触发词: publish, 公众号发布, wechat article, 发布文章, 排版, content publish, blog post, newsletter publish
| Skill | 说明 |
|---|---|
wewrite | Wewrite |
wechat-article-writer | Wechat Article Writer |
wechat-article-search | Wechat Article Search |
wechat-reader | Wechat Reader |
qiaomu-x-article-publisher | Qiaomu X Article Publisher |
md-report-summary | Md Report Summary |
baoyu-translate | Baoyu Translate |
grammar-check | Grammar Check |
stop-slop | Stop Slop |
humanizer | You are a writing editor that identifies and removes signs of AI-generated text |
/project-promotion-pipeline项目推广三件套: 传统 SEO + AI 搜索 GEO + 外链自动化
触发词: seo, geo, promotion, 推广, 外链, backlink, marketing, gtm launch, launch
| Skill | 说明 |
|---|---|
agentic-seo | Agentic Seo |
marketing-ideas | Marketing Ideas |
marketing-skills | Marketing Skills |
positioning-ideas | Positioning Ideas |
product-name | Product Name |
gtm-strategy | Gtm Strategy |
gtm-motions | Gtm Motions |
/business-diagnosis-pipeline栋哥商业诊断工具箱: 商业模式诊断 + 对标分析 + 内容创作 + 执行力 + 概念拆解
触发词: 商业, 诊断, 对标, 模仿, 内容诊断, 执行力, 拖延, 拆概念, 商业模式, dbs, 帮我看看
| Skill | 说明 |
|---|---|
dbs | Dbs |
dbs-diagnosis | Dbs Diagnosis |
dbs-benchmark | Dbs Benchmark |
dbs-content | Dbs Content |
dbs-unblock | Dbs Unblock |
dbs-deconstruct | Dbs Deconstruct |
/skill-ops-pipelineOpenClaw runtime upgrades, skills supply-chain audit, registry hygiene, auto-trigger routing, and workflow automation.
触发词: openclaw, clawhub, skill update, skills update, skill group, skills registry, skill audit, supply chain, 自动触发, 自动调用, 技能更新, 技能组, 技能审计, 供应链审计
| Skill | 说明 |
|---|---|
openclaw-best-practices | 本 Skill 目标是把社区里“跑得久以后才总结出来的 5 条基本原则”固化成一份可执行的检查清单,避免重复踩坑与供应链风险。 |
dna-inheritance | This skill explains how to treat a shared git repository as an agent \"DNA\" sys |
planning-with-files | 用本地三份文件作为“外置工作记忆”,避免长任务在对话变长后发生目标漂移: |
using-superpowers | <EXTREMELY-IMPORTANT> If you think there is even a 1% chance a skill might apply |
find-skills | This skill helps you discover and install skills from the open agent skills ecos |
writing-skills | **Writing skills IS Test-Driven Development applied to process documentation.** |
/stitch-design-pipelineFull-auto design pipeline: PRD -> Stitch prototype -> Gemini icon -> code export -> UI polish
触发词: stitch, 设计原型, design prototype, generate screen, stitch loop
执行顺序:enhance-prompt -> stitch-skills -> stitch-loop -> design-md -> react-components -> stitch-design-pipeline
| Skill | 说明 |
|---|---|
enhance-prompt | You are a **Stitch Prompt Engineer**. Your job is to transform rough or vague UI |
stitch-skills | 把 Stitch MCP 的 design-md 与 react-components 能力固化为可复用 runbook 与组件文档,降低设计到代码的交付成本。 |
stitch-loop | You are an **autonomous frontend builder** participating in an iterative site-bu |
design-md | You are an expert Design Systems Lead. Your goal is to analyze the provided tech |
react-components | You are a frontend engineer focused on transforming designs into clean React cod |
stitch-design-pipeline | Stitch Design Pipeline |
/wechat-content-pipeline公众号全流程: 热点→选题→写作→去AI→排版→推送
触发词: 公众号写作, 推文, 微信文章, 写公众号
| Skill | 说明 |
|---|---|
wewrite | Wewrite |
wechat-article-writer | Wechat Article Writer |
wechat-article-search | Wechat Article Search |
wechat-reader | Wechat Reader |
/codex-relay-pipelineDual-agent relay: Claude Code + Codex CLI cross-model collaboration
触发词: codex review, relay, cross-model, dual agent
| Skill | 说明 |
|---|---|
codex-review | Codex Review |
code-review | Code Review |
adversarial-review | Adversarial Review |
/ios-dev-pipelineCLI-first SwiftUI development with xcodebuild + XcodeBuildMCP
触发词: iOS, SwiftUI, watchOS, macOS app, Xcode, Apple Watch
| Skill | 说明 |
|---|---|
ios-swiftui-expert | Ios Swiftui Expert |
ios-device-automation | > **CRITICAL RULES — VIOLATIONS WILL BREAK THE WORKFLOW:** > > 1. **Never run mi |
| 命令 | 说明 |
|---|---|
/tmux-bridge | 多 Agent 跨 tmux session 协作工具 |
/commit | Git 提交(自动生成 message) |
/pr | 创建 Pull Request |
/push-all | 暂存 + 提交 + 推送 |
/relay-plan /relay-code /relay-codex /relay-review | Claude+Codex 文件中继协作 |
/rpi:plan /rpi:research /rpi:implement | RPI 三阶段(研究->规划->实现) |
/telegram:access /discord:access | 频道管理 |
aider | Aider - 开源AI结对编程助手 |
amp | Amp CLI - ampcode.com |
augment | Augment CLI (Auggie) - Augment Code AI编程代理 |
claude | Claude Code - Anthropic编程助手 |
codex | OpenAI Codex - OpenAI官方编程助手 |
codex-openclaw | Quiet Codex CLI wrapper for OpenClaw's `codex-cli/*` backend |
copilot | GitHub Copilot CLI - GitHub官方编程助手 |
droid | Factory Droid - Factory AI编程代理 |
gemini | Google Gemini CLI - Google官方AI助手 |
opencode | OpenCode CLI - sst/opencode |
af | AI Tools 统一入口 |
ai | AI Tools 统一入口 |
ai-workflow | AI-Tools LangGraph Workflow CLI - LangGraph 工作流统一入口 |
ait | AI Tools launcher entrypoint (alternative name). |
a2a | A2A Agent Registry - list/discover/card/stats (233 Agent Car |
agent | Cursor Agent CLI - Cursor 官方 Agent 命令行 |
amazon | Amazon Q CLI - AWS官方AI助手 |
asc | App Store Connect CLI (asc) - release automation helper |
codegraph | CodeGraph OSS wrapper - commercial-friendly local knowledge |
codex-2p | Codex via Poe shortcut wrapper |
cursor-agent | Cursor Agent CLI - 兼容命令 cursor-agent |
embed | Fleet Embedding Service - Unified CLI |
fleet-run | Unified project runner -- Brooks 3-layer contract (build/tes |
gemini-acp | Wrapper to launch gemini CLI with ACP protocol enabled |
github | GitHub CLI 工具集 |
github-mcp-server | GitHub MCP Server - GitHub API 的 MCP 协议服务端 |
gitnexus | GitNexus wrapper - knowledge graph operations via AI-tools l |
graph | Graph wrapper - alias to CodeGraph OSS via AI-tools launcher |
interpreter | Open Interpreter - 自然语言全能执行环境 |
jarvis | Terminal Jarvis - 多AI工具统一接口 |
metrics | metrics -- Session cost/usage metrics viewer |
nebula | CogNebula wrapper - repository knowledge graph engine |
oi | Open Interpreter - 自然语言全能执行环境 (快捷命令: oi) |
router | 4-Layer Cascading Decomposition Router - route/strategy/exec |
ruler | Ruler - AI编程规则管理器 |
speckit | Spec Kit (Tier0) - Spec-Driven Development 工具箱 |
swarm | Swarm Intelligence Engine - status/health/init/simulate/allo |
MCP 服务器:让 Agent 通过标准接口访问外部系统。
| qmd | 文档全文搜索 |
| github | GitHub API |
| ghost-os | macOS 无障碍树(29 工具) |
| chrome-devtools | CDP + Lighthouse(40+) |
| claude-in-chrome | 浏览器扩展(16 工具) |
| excalidraw | 白板绘图 |
| agentation | Agent 标注工作流 |
| apple-docs | Apple 开发文档 |
| 21st-magic | UI 组件生成 |
| stitch | Google 设计原型 |
| nanobanana-mcp | Gemini 图片生成 |
| thesvg | 5600+ SVG 图标 |
| xcodebuild | Xcode 构建集成 |
| computer-use | 桌面自动化 |
| codex-cli | Codex CLI 代理 |
| Cloudflare | Workers/D1/KV/R2 |
| Gmail | 邮件读写 |
| Calendar | 日程管理 |
| telegram | Telegram 双向消息 |
| discord | Discord 通知 |
| swift-lsp | Swift 语言服务 |
| clangd-lsp | C/C++ 语言服务 |
全自动:PRD -> 原型 -> 图标 -> 代码。无需人工干预。
| 1 | PRD -> Stitch 提示词 |
| 2 | Stitch 多迭代原型 |
| 3 | Gemini Pro 图标/Logo |
| 4 | 导出 -> Figma/HTML 代码 |
| 5 | 3 轮 UI 打磨(强制) |
| 深色陷阱 | 必须显式指定 LIGHT + 色值 |
| CORS | 用 CDP a11y click 绕过 |
| 16 屏限制 | 导出前删多余屏 |
| 认证 | CLI ≠ 浏览器,失败回退 |
| 模型 | 必须 Gemini Pro,不用 Flash |
默认用最低层,失败了才升一层。
| 层 | 工具 | Token | 场景 |
|---|---|---|---|
| L0 | WebMCP | ~20 | Chrome 146+ 结构化调用 |
| L0.5 | bb-browser | ~100 | 已登录 Chrome 直连 |
| L1 | agent-browser-session | ~200 | DOM + 反检测 + 持久登录 |
| L2 | DevTools / Ghost OS | ~300 | CDP 调试 / macOS 桌面 |
| L3 | Midscene | ~2000 | 视觉定位(最后手段) |
| L4 | agent-fetch / Jina | ~50 | HTTP 提取(默认首选) |
| L5 | fleet-page-fetch | ~100 | 翻墙 / 反爬代理 |
| 蓝图 | 模式 | 说明 |
|---|---|---|
| full-review | 议会(4) | 安全+性能+架构+测试 |
| parallel-research | 蜂群(3) | 独立并行研究 |
| pipeline-release | 管线(3) | 门禁->部署->验证 |
| council-architecture | 议会(4) | PM+架构+安全+QA |
| watchdog-monitor | 看门狗(3) | 并行构建+质量监控 |
| advisor-council | 动态(2-8) | 13 维顾问决策 |
| D1 Drucker | 价值与客户 |
| D2 Jobs | 产品体验 |
| D3 Meadows | 系统动力学 |
| D4 Munger | 多元思维模型 |
| D5 Taleb | 反脆弱 |
| D6 Hickey | 简单性 |
| D7 Brooks | 项目管理 |
| D8 Musk | 执行速度 |
| D9 Catmull | 创意管理 |
| D10 Orwell | 清晰表达 |
| D11 Buffett | 长期价值 |
| D12 Hara | 留白美学 |
| D13 Godin | 市场传播 |
| verification_gate | 完成前两阶段检查 |
| loop_detector | L0-L4 循环检测升级 |
| context_monitor | 40%/50%/60% 三级预警 |
| trace_collector | 结构化执行追踪 |
| knowledge_compounder | 修复 -> 模式 -> DNA 胶囊 |
| staleness_detector | 过时引用检测+自修复 |
| semantic_skill_router | 语义相似度渐进披露 |
| quality_scorer | 5 维质量评分 A-F |
| structural_test | ArchUnit 层依赖验证 |
| auditor | 全面审计报告 |
| health | 统一健康报告 |
| session_state | 跨会话状态持久化 |
生成、摘要、Telegram、语音、社交、专家团
心跳、自愈、路由、同步、设备网格、代理
爬取、研究、知识图谱、安全审计、新闻
自优化、记忆整合、Skill 审计、知识胶囊固化
CI/CD、部署预览、测试、模型测试
知识图谱索引(KuzuDB)