Skills Library

A categorized collection of 688 AI agent skills organized by domain and technology.

All skills live under SKILLS/ and follow the standard Agent Skills layout:

SKILLS/<CATEGORY>[/<SUBCATEGORY>]/<skill-folder>/
├── SKILL.md          # Required — the skill itself
├── references/       # Optional — supporting docs
├── scripts/          # Optional — executable helpers
├── evals/            # Optional — evaluation cases
└── ...

---

Categories

Category	Subcategories	Skills
AI	AGENTS, EVAL, IMAGE, LLM, ML, RAG	52
ARCHITECTURE	—	7
BUSINESS	ANALYTICS, FINANCE, LEGAL, MANAGEMENT, MARKETING, PRODUCT, STARTUP	38
CLOUD	AZURE, DEVOPS, KUBERNETES, SERVERLESS	29
DATA	—	22
DATABASE	—	17
DEVTOOLS	BUILD, DEBUGGING, DOCS, ENV, REFACTORING, REVIEW, TESTING	116
FRAMEWORKS	ALPINE, ANGULAR, ASTRO, ELECTRON, EXPO, REACT, VITE, VUE	32
GAMES	GENERAL, GODOT, HOI4, MINECRAFT, UNITY, UNREAL	24
LANGUAGES	BASH, CPP, CSHARP, ELIXIR, GO, HASKELL, JAVA, JAVASCRIPT, JULIA, KOTLIN, LUA, OTHER, PHP, PYTHON, RUBY, RUST, SCALA, SWIFT, TYPESCRIPT	83
META	—	26
MOBILE	ANDROID, CROSS, MAUI, REACT-NATIVE	14
OFFICE	—	9
RESEARCH	—	4
SECURITY	AUDIT, PATTERNS, PENTEST, REVERSE	28
SQERSTERS	—	3
WEB	3D, BACKEND, FRONTEND, FULLSTACK, OTHER, PERFORMANCE, SCRAPING, SEO, TESTING, UI	173
WIKI	—	6
WRITING	—	5

---

Category guide

Top-level categories and what lives in each subcategory.

AI — applied AI/ML skills

• AGENTS/ — agent frameworks, orchestration, tool design (LangChain, LangGraph, LlamaIndex, custom agents)
• EVAL/ — LLM evaluation, eval-driven development, LLM-as-judge
• IMAGE/ — image generation pipelines (ComfyUI, AI Studio)
• LLM/ — prompt engineering, structured output, LLMOps, local LLMs
• ML/ — classical ML, deep learning, MLOps (scikit-learn, PyTorch, JAX, pipelines)
• RAG/ — retrieval, vector search, embeddings, similarity patterns

ARCHITECTURE — system design

ADRs, DDD, event sourcing, pattern selection, trade-off analysis.

BUSINESS — non-engineering domains

• ANALYTICS/ — product analytics, event tracking
• FINANCE/ — quant analysis, risk metrics
• LEGAL/ — legal/compliance advisory
• MANAGEMENT/ — HR, IT mgmt, scheduling, estimation
• MARKETING/ — ads, growth, sales, ASO, launch strategy
• PRODUCT/ — product design, PM toolkits
• STARTUP/ — market sizing, MVP, financial modeling, metrics

CLOUD — cloud platforms & ops

• AZURE/ — Azure SDKs (AI, OpenAI, translation, transcription, projects)
• DEVOPS/ — CI/CD, IaC, observability, incident response
• KUBERNETES/ — k8s architecture and deployment
• SERVERLESS/ — serverless platforms, Trigger.dev

DATA — data engineering & science

pandas, big data, ETL, data quality, scientific computing, warehousing.

DATABASE — relational + NoSQL

SQL/NoSQL design, Postgres, ORMs, optimization, schema patterns.

DEVTOOLS — engineering workflow

• BUILD/ — MSBuild, binlogs, build perf, templates
• DEBUGGING/ — strategies, error handling, smart debug
• DOCS/ — code/API documentation generators
• ENV/ — devcontainers, git, GitHub, IDE, workspace
• REFACTORING/ — clean code, tech debt, modernization
• REVIEW/ — code review, audits
• TESTING/ — TDD, unit testing, test frameworks

FRAMEWORKS — frontend/app frameworks

• ALPINE/, ANGULAR/, ASTRO/, ELECTRON/, EXPO/, REACT/, VITE/, VUE/

GAMES — game development

• GENERAL/ — game design, engines, abstract strategy
• GODOT/ — Godot / GDScript
• HOI4/ — Hearts of Iron IV modding
• MINECRAFT/ — mods, plugins, datapacks, server admin, resource packs
• UNITY/ — Unity, ECS, AI-assisted creation
• UNREAL/ — Unreal Engine C++

LANGUAGES — per-language skills

One subfolder per language (Bash, C++, C#, Elixir, Go, Haskell, Java, JS, Julia, Kotlin, Lua, PHP, Python, Ruby, Rust, Scala, Swift, TypeScript, plus OTHER/).

META — about the skills themselves

Skill creation, scaffolding, auditing, brainstorming, project skill audits.

MOBILE — mobile platforms

• ANDROID/ — Android, Jetpack Compose
• CROSS/ — cross-platform mobile patterns
• MAUI/ — .NET MAUI
• REACT-NATIVE/ — React Native architecture

OFFICE — productivity / docs

PDF, Word, Excel, PowerPoint, Notion, Gmail, Google Calendar.

RESEARCH — academic research & writing

Coupled academic pipeline: deep research, paper writing, multi-perspective peer review, and end-to-end orchestration. Each skill ships its own shared/ protocols and scripts/ validators.

SECURITY — defense & offense

• AUDIT/ — security audits, compliance, hardening
• PATTERNS/ — JWT, GDPR, threat modeling, secure coding
• PENTEST/ — red team, fuzzing, SQL injection
• REVERSE/ — reverse engineering, anti-RE

SQERSTERS — my own skills

Skills I wrote myself. Take precedence on duplicate names when building the flat zip.

WEB — everything web

• 3D/ — three.js, WebGL, R3F
• BACKEND/ — APIs, GraphQL, Node, payments
• FRONTEND/ — UI architecture, design taste, PWA
• FULLSTACK/ — end-to-end app building
• OTHER/ — browser extensions, bots, weather, maps
• PERFORMANCE/ — web perf, profiling
• SCRAPING/ — browser automation, web/wiki/arxiv search
• SEO/ — SEO, AEO, GEO, schema markup
• TESTING/ — API mocking, web testing
• UI/ — design systems, components, CSS, animations, viz

WIKI — codebase documentation

Wiki architects, onboarding docs, changelogs, Q&A from source.

WRITING — natural-language work

Proofreading, scientific writing.

---

Tree

SKILLS/
├── AI/
│   ├── AGENTS/         agent frameworks, orchestration, tool use
│   ├── EVAL/           LLM evals, eval-driven dev, judging
│   ├── IMAGE/          image generation, ComfyUI
│   ├── LLM/            prompt engineering, LLMOps, structured output
│   ├── ML/             classical ML, deep learning, MLOps
│   └── RAG/            retrieval, vector search, embeddings
├── ARCHITECTURE/       ADRs, DDD, event sourcing, system design
├── BUSINESS/
│   ├── ANALYTICS/      product analytics, tracking
│   ├── FINANCE/        risk metrics, quant analysis
│   ├── LEGAL/          legal advisory
│   ├── MANAGEMENT/     HR, IT mgmt, scheduling, estimation
│   ├── MARKETING/      ads, growth, sales, ASO
│   ├── PRODUCT/        product design, PM toolkits
│   └── STARTUP/        market sizing, MVP, financial modeling
├── CLOUD/
│   ├── AZURE/          Azure SDKs (AI, OpenAI, translation, etc.)
│   ├── DEVOPS/         CI/CD, IaC, observability, incident response
│   ├── KUBERNETES/     k8s architecture and deployment
│   └── SERVERLESS/     serverless platforms, Trigger.dev
├── DATA/               pandas, big data, ETL, scientific computing
├── DATABASE/           SQL, NoSQL, Postgres, ORMs
├── DEVTOOLS/
│   ├── BUILD/          MSBuild, binlogs, templates, build perf
│   ├── DEBUGGING/      debugging strategies, error handling
│   ├── DOCS/           code/API documentation generators
│   ├── ENV/            devcontainers, git, build, IDE, GitHub, workspace
│   ├── REFACTORING/    clean code, tech debt, modernization
│   ├── REVIEW/         code review, audits
│   └── TESTING/        TDD, unit testing, test frameworks
├── FRAMEWORKS/
│   ├── ALPINE/         Alpine.js
│   ├── ANGULAR/        Angular ecosystem
│   ├── ASTRO/          Astro
│   ├── ELECTRON/       Electron desktop apps
│   ├── EXPO/           Expo / React Native tooling
│   ├── REACT/          React, Next.js
│   ├── VITE/           Vite
│   └── VUE/            Vue.js
├── GAMES/
│   ├── GENERAL/        game design, engines, strategy
│   ├── GODOT/          Godot / GDScript
│   ├── HOI4/           Hearts of Iron IV modding
│   ├── MINECRAFT/      mods, plugins, datapacks, server admin
│   ├── UNITY/          Unity development
│   └── UNREAL/         Unreal Engine C++
├── LANGUAGES/          per-language skills (see table above)
├── META/               skill creation, auditing, brainstorming
├── MOBILE/
│   ├── ANDROID/        Android, Jetpack Compose
│   ├── CROSS/          cross-platform mobile
│   ├── MAUI/           .NET MAUI
│   └── REACT-NATIVE/   React Native architecture
├── OFFICE/             PDF, office productivity, Excel, Word, PowerPoint, Notion, Gmail, Calendar
├── RESEARCH/           academic paper writing, peer review, deep research, pipeline
├── SECURITY/
│   ├── AUDIT/          security audits, compliance, hardening
│   ├── PATTERNS/       JWT, GDPR, threat modeling
│   ├── PENTEST/        red team, fuzzing, SQL injection
│   └── REVERSE/        reverse engineering, anti-RE
├── SQERSTERS/          my own skills (made by me — SQERSTERS)
├── WEB/
│   ├── 3D/             three.js, WebGL, R3F
│   ├── BACKEND/        APIs, GraphQL, Node, payments
│   ├── FRONTEND/       UI architecture, design taste, PWA
│   ├── FULLSTACK/      end-to-end app building
│   ├── OTHER/          browser extensions, bots, weather, maps
│   ├── PERFORMANCE/    web perf, profiling
│   ├── SCRAPING/       browser automation, web/wiki/arxiv search
│   ├── SEO/            SEO, AEO, GEO, schema markup
│   ├── TESTING/        API mocking, web testing
│   └── UI/             design systems, components, CSS, animations, viz
├── WIKI/               codebase wikis, onboarding docs
└── WRITING/            proofreading, scientific writing

---

Finding a skill

• By tech: jump to LANGUAGES/<LANG> or FRAMEWORKS/<NAME>.
• By task: pick the matching domain (e.g. debugging → DEVTOOLS/DEBUGGING, SEO → WEB/SEO).
• By name: search the tree — every folder is named 001-<skill-slug>.

Each folder contains a SKILL.md with the activation keywords and full instructions.

---

Layout rules

• One skill per folder.
• A skill goes in the single category that fits it best — no duplication.
• Subcategories are uppercase. Skill folders keep the 001- prefix.
• Top-level categories without a clear subcategory split (e.g. DATA, DATABASE, META) hold skills directly.

---

Steerings

The STEERINGS/ folder holds my personal AI steering rules — open-sourced so anyone can use or adapt them.

STEERINGS/
└── SQERSTERS/                my own steering set (made by me — SQERSTERS)
    ├── AGENTS.md             hard, non-negotiable rules — ask before any change
    ├── all_agent_new.md      global "always do / never do" rules for every task
    ├── code_mastery.md       how I write, review, and improve code
    ├── opensource.md         rules for open-source / external contributions
    └── project_privates.md   rules for private/personal projects

What each file does

File	Purpose	Always active?
AGENTS.md	Hard rules. The agent must ask for approval before creating, editing, deleting, running, refactoring, or installing anything — even on small changes, and even when the user says "update" or "add". Approval is never assumed; the agent must confirm before acting.	Yes
all_agent_new.md	Global workflow rules: activate ≥2 relevant skills per task, scan for errors, double-check output, keep the 3 core docs (README.md, TODO.md, CHANGELOG.md) under docs/, never fabricate, never act without authorization. Defines ANALYSIS / IMPLEMENTATION / REVIEW modes.	Yes
code_mastery.md	Code quality bar: KISS, single-responsibility functions, security defaults (no hardcoded secrets, parameterized queries), error handling, testing mindset, dependency hygiene, consistency with existing style.	Yes
opensource.md	When contributing to an external open-source project: don't touch upstream docs, find the root cause first, get explicit approval, ship a minimal fix, open a clean human-toned PR, respect licenses.	Conditional
project_privates.md	When the project is yours: you own the docs, can update README/TODO/CHANGELOG freely in scope, fix obvious bugs along the way — but new deps, deletions, architecture changes, and auth/crypto/payments still require approval.	Conditional

How they work together

• AGENTS.md, all_agent_new.md, and code_mastery.md are always active.
• opensource.md and project_privates.md are mutually exclusive — pick one based on the project:
  • Open-source / contributing to someone else's repo → use opensource.md.
  • Your own project (full ownership) → use project_privates.md.
• Precedence on conflict (from all_agent_new.md):
  1. Safety & user control (hard rules in AGENTS.md)
  2. Context rules (opensource.md or project_privates.md)
  3. Code quality (code_mastery.md)

Drop the files you want into your agent's steering folder (for example ~/.kiro/steering/) to use them.

---

Building a flat zip for installation

Some agents (Kiro included) only read skill folders at the root of their skills directory — they don't recurse into category subfolders. To install this whole library into such an agent, build a flat zip first:

# From the repo root
powershell -ExecutionPolicy Bypass -File scripts\Build-SkillsZip.ps1 -Force

This produces dist/skills.zip with every 001-* skill folder placed at the root of the archive (no AI/, WEB/, etc. wrappers). Extract the zip into the agent's skills directory and every skill is where the agent expects it.

Scripts

Script	What it does
scripts/Build-SkillsZip.ps1	Walks SKILLS/ recursively, finds every 001-* folder containing a SKILL.md, and packs them into a flat zip at dist/skills.zip.
scripts/Test-SkillsZip.ps1	Verifies the produced zip: every top-level entry is a 001-* folder, no nested 001-/001- paths, every skill has a SKILL.md, no empty folders. Exits non-zero on failure.

Build options

# Default — skip duplicates with a warning, output to dist/skills.zip
powershell -File scripts\Build-SkillsZip.ps1 -Force

# Keep both copies of any duplicate, suffixed by category path
powershell -File scripts\Build-SkillsZip.ps1 -OnDuplicate Suffix -Force

# Stop with an error if any duplicates exist (useful in CI)
powershell -File scripts\Build-SkillsZip.ps1 -OnDuplicate Fail

# Custom output location
powershell -File scripts\Build-SkillsZip.ps1 -OutFile C:\path\to\my-skills.zip -Force

# Verify the result
powershell -File scripts\Test-SkillsZip.ps1

Duplicate handling

SKILLS/SQERSTERS/ is given priority on duplicate skill names — if the same 001-<name> folder exists in both SQERSTERS/ and a community category, the SQERSTERS version is kept. Currently the only collision is 001-hoi4-skills (in GAMES/HOI4/ and SQERSTERS/).