awesome-skillkit

English | 中文 | 日本語

🌐 Browse / download: site source repo (enable GitCode Pages for a live site, already synced to v0.19.0): gitcode.com/badhope/skillkit-site. Per-pack and full _all.zip are on each platform's Releases (see "Platform sync status" below). Deploy guide: docs/DEPLOY-SITE.md

Curated scene packs for AI tools. Each pack = one real-world scenario, containing multiple hand-picked skills. Download a zip → unzip → drag the skill folders into your AI tool's skills directory → it just works.

Positioning

The scenario is the answer — grounded in platform + tool.

flowchart LR
    A[Real-world scenario] --> B[Scene pack<br/>1 pack = 1 scenario]
    B --> C[Group of collaborating skills<br/>2–18 of them]
    C --> D[Drop into AI tool<br/>skills dir]
    D --> E[Works in new session]
    style A fill:#eaf2ff,stroke:#5b8def
    style E fill:#eafaea,stroke:#4caf72

Loading

• Every pack maps to a concrete scenario ("review a PR", "build CI/CD", "post to my blog"), not a broad domain.
• Every pack bundles the skills that work together for that scenario — from a focused pair to an 18-skill suite (Content Publishing Automation covers 16 Chinese platforms end-to-end).
• Every skill's source is clearly attributed (upstream curated / self-authored / open-source distilled).

By the numbers

143 skills · 36 scene packs · 18 chain domains / 58 skill chains · v0.19.0 · Apache-2.0

Source breakdown:

pie title Skill sources (n=143)
    "Self-authored scenario skills" : 105
    "Upstream curated (alirezarezvani/claude-skills, MIT)" : 36
    "Other open-source distilled" : 10

Loading

Pack size distribution (█ = 1 skill, 36 packs total):

场景包	技能数	规模
AI Research & Writing	19	███████████████████
Content Publishing Automation	18	██████████████████
Office Productivity	8	████████
Data, ML & Scientific Computing	7	███████
AI Video Pipeline	6	██████
AI Agent Development	5	█████
Code Review	5	█████
AI Media Toolkit	4	████
Image Studio	4	████
Memory Systems	4	████
Toolsmith	4	████
Video Design Studio	4	████
Visual Design Studio	4	████
Workspace Integrations	4	████
System Architecture	3	███
Audio Studio	3	███
CI/CD Pipeline	3	███
Code Planning & Generation	3	███
Containers & Orchestration	3	███
De-AI Writing	3	███
Edu Craft	3	███
GitHub Collaboration	3	███
Growth Marketing	3	███
Homework Autopilot	3	███
Incident Response & SRE	3	███
Infrastructure as Code	3	███
Skill Forge	3	███
API Development & Testing	2	██
Database Design & Management	2	██
Data Viz Studio	2	██
Knowledge Base	2	██
Security & Secrets	2	██
Test-Driven Development	2	██
Viral Entertainment	2	██
Chat Prompt Craft	1	█
Performance Profiling	1	█

Getting started (30 seconds)

1. 📦 Download the zip for the scene you need from Releases (or run python3 build.py to build dist/*.zip locally).
2. 📂 Unzip — you get multiple skill folders (each with SKILL.md).
3. 🧲 Drag the folders into your AI tool's skills directory:
   • Claude Code: ~/.claude/skills/ (global) or .claude/skills/ (project)
   • Other tools with skills support: use their skills directory
4. 🚀 Start a new session — works immediately, no config.

Scene pack catalog

工程与编程

Scene pack	Skills	Scenario	Source
AI Agent Development	5	生产级 Agent、多智能体、MCP、特性开关、自评估	上游
Code Review	5	PR 审查、代码质量、依赖审计、技术债	上游
System Architecture	3	系统架构、零停机迁移、monorepo	上游
CI/CD Pipeline	3	CI/CD 流水线、发布门、spec 驱动开发	上游
Code Planning & Generation	3	模糊需求→结构化计划→生成→失败诊断	自研
Containers & Orchestration	3	Dockerfile、compose、Helm、K8s operator	上游
GitHub Collaboration	3	并行 worktree、约定式 changelog、PR 审查	上游
Incident Response & SRE	3	事故指挥、runbook、SLO/错误预算	上游
Infrastructure as Code	3	Terraform 模式、可观测性、K8s	上游
API Development & Testing	2	REST API 设计审查、契约/集成测试	上游
Database Design & Management	2	库设计、ERD、迁移、SQL 优化	上游
Security & Secrets	2	密钥库、环境变量卫生	上游
Test-Driven Development	2	单测、fixture、mock、红绿重构、Playwright 流程测试	上游
Chat Prompt Craft	1	对话 AI 提示词工程：五要素公式、agent 系统提示、反向约束	自研
Performance Profiling	1	Node/Python/Go 的 CPU/内存/IO 剖析	上游

研究与写作

Scene pack	Skills	Scenario	Source
AI Research & Writing	19	从问题到成稿：多轮研究、选题、大纲、草稿、润色、SEO、图表、LaTeX	自研
De-AI Writing	3	AI 痕迹审计、人声改写、个人声纹档案（降低 AI 感）	自研

内容发布

Scene pack	Skills	Scenario	Source
Content Publishing Automation	18	16+ 中文平台文章/视频发布、编辑、跨平台分发、AI 封面	自研

视频创作

Scene pack	Skills	Scenario	Source
AI Video Pipeline	6	短视频全链路：脚本→配音→对口型→剪辑→字幕→封面→发布	自研
AI Media Toolkit	4	文/图生视频、生图、生乐、封面（本地生成网关）	自研
Video Design Studio	4	视频前期：分镜、镜头配方、prompt 工程、风格锚点	自研
Viral Entertainment	2	会说话宝宝播客、龙崽 meme 短片（角色一致性）	自研

图像与设计

Scene pack	Skills	Scenario	Source
Image Studio	4	图像创作工作台：prompt、重绘、扩图、超分	自研
Visual Design Studio	4	brief→spec→prompt→layout 审计，设计总监两遍工作流	自研

音频

Scene pack	Skills	Scenario	Source
Audio Studio	3	播客链：脚本→配音→发布（Kokoro/Qwen3-TTS）	自研

数据科学

Scene pack	Skills	Scenario	Source
Data, ML & Scientific Computing	7	ETL、特征工程、建模求解、仿真、可视化、ML 流水线	自研

办公生产力

Scene pack	Skills	Scenario	Source
Office Productivity	8	PPT、Excel、Word、PDF、简历、纪要、内部通讯	自研

增长营销

Scene pack	Skills	Scenario	Source
Growth Marketing	3	电商营销链：文案框架、活动策划、渠道适配	自研

教育

Scene pack	Skills	Scenario	Source
Edu Craft	3	精通教学链：课程→练习→费曼讲解	自研
Homework Autopilot	3	一键作业完成（有温度版，降低冷血感）	自研

记忆系统

Scene pack	Skills	Scenario	Source
Memory Systems	4	长期记忆：设计/抽取/管理/检索（mem0/letta 蒸馏）	自研

工具与自动化

Scene pack	Skills	Scenario	Source
Toolsmith	4	文件整理、批量重命名、格式转换、定时任务（全部 dry-run 优先）	自研

元技能

Scene pack	Skills	Scenario	Source
Skill Forge	3	技能生成、规范校验（CI 门禁）、技能检索与装配	自研

外部集成

Scene pack	Skills	Scenario	Source
Workspace Integrations	4	Notion / 飞书·钉钉·企业微信 / Jira·Linear·GitHub Issues / 云盘归档	自研

个人知识库

Scene pack	Skills	Scenario	Source
Knowledge Base	2	笔记库构建（索引/检索/体检）、知识图谱抽取与导出	自研

数据可视化

Scene pack	Skills	Scenario	Source
Data Viz Studio	2	CSV 剖析 → 仪表盘生成（零外部依赖）、图表选择词库	自研

Detail lexicons (the differentiator)

Generation-quality skills live or die on how detailed the prompt is. We ship high-density detail lexicons — term + effect/mood + when-to-use + example — to consult before writing a prompt:

Lexicon	Domain	Coverage
cinematography-lexicon.md	视频	17 种转场 / 动作动词空间语义 / 微表情表演 / 速度节奏 / 五模型方言 / 迭代修复对照

Representative skill chains (pipelines, not single points)

flowchart LR
    subgraph Short-video pipeline
    S[video-script-writer] --> V[video-voice-synth]
    V --> L[video-lip-sync] --> E[video-editor]
    E --> SUB[video-subtitles] --> T[video-thumbnail] --> P[Publish]
    end
    subgraph Research-writing chain
    R[deep-research] --> O[article-outliner] --> D[article-drafter]
    D --> C[content-editor] --> Q[seo-optimizer]
    end
    subgraph Publishing chain
    W[Article/Video] --> A[ai-cover-generator] --> X[cross-post-orchestrator]
    X --> Z[16+ platforms]
    end

Loading

skill_chains.json ships 18 chain domains / 58 chains, pinning "who runs first, who hands off to whom" so cross-module calls never get lost.

Platform sync status

Platform	Repo	Code sync	Release / assets	Status
GitCode	badhope/awesome-skillkit	✅ through 0.19.0	✅ created	OK
Gitee	badhope/awesome-skillkit	✅ through 0.19.0	✅ v0.19.0, 36 zip assets	OK
GitHub	x33834/awesome-skillkit	⚠️ push from local	⚠️ none	Sandbox egress ACL cuts TLS; run git push origin main --follow-tags locally (token needs repo+workflow)
Site	badhope/skillkit-site	✅ synced v0.19.0	—	Enable GitCode Pages to go live

Directory layout

packs/              # scene pack definitions (one dir per scenario; pack.json = metadata + skill list + sources)
skills/             # single source of truth for all skills (multi-level taxonomy)
  ├─ programming/   # upstream curated (alirezarezvani, 33 programming skills)
  ├─ writing/       # self-authored scenario skills
  ├─ video/ design/ audio/ marketing/ education/ scenarios/ …
  └─ skill_chains.json  # 13 domains / 48 chains
dist/               # build output: one zip per pack (gitignored)

Sources & attribution

Two tracks, fully attributed per-skill in manifest.json, each packs/*/pack.json, and SOURCES.md:

• Self-authored (105): skills/writing/, scenarios/, design/, audio/ etc. — China-platform automation, video/image/audio pipelines, de-AI writing, memory systems, homework autopilot, all original workflows with executable lint scripts + unit tests, dry-run by default.
• Upstream curated (36, MIT): alirezarezvani/claude-skills for programming/engineering skills.
• Other distilled (10): methodology distilled from Anthropic public skills docs, mem0/letta/Claude memory tool, Kokoro/Qwen3-TTS ecosystem — all credited in references/sources-and-methodology.md, zero content copied.

Build & release

python3 build.py                           # single build entry: dist/*.zip per pack + dist/_all.zip
python3 tools/release.py 0.19.0 --commit   # validate CHANGELOG → bump → commit → tag
git push origin main --follow-tags            # push code + three version tags
# Create the release on Gitee / GitCode and upload dist/*.zip (manifest.json version is the single source of truth)

License

Apache License 2.0 © 2026 Morningstar202604