Awesome Agent Skills

A curated list of agent skills, resources, and tools for building customizable AI workflows across Codex, Claude Code, and other platforms.

Goal: Help you discover, design, and reuse agent skills (prompts, tools, patterns, and automations) that are practical, composable, and production-ready.

Awesome Agent Skills

What is a Skill?

A skill is a reusable workflow unit for an AI agent—typically a combination of:

Clear objective: What success looks like and how to measure it
Input/output schema: Well-defined data structures and formats
Steps & constraints: Explicit instructions and boundaries
Tool usage conventions: How and when to use specific tools
Tests & evaluation criteria: Methods to verify correctness and quality

Skills enable AI agents to perform specialized tasks consistently and reliably across different platforms and contexts.

Skill Categories

Document Processing

Skills for creating, editing, analyzing, and transforming documents across various formats including Word, PDF, PowerPoint, Excel, Markdown, EPUB, and more.

Common use cases: Document generation, format conversion, content extraction, metadata manipulation, batch processing

TBD: Skill list for this category (contributors can add entries here)

Development & Code Tools

Skills focused on programming, testing, debugging, code review, browser automation, artifact building, and development workflow optimization.

Common use cases: Code refactoring, test generation, debugging assistance, code review automation, web testing, build optimization

TBD: Skill list for this category (contributors can add entries here)

Data & Analysis

Skills for working with databases, CSV/JSON files, data transformation, analysis, and visualization.

Common use cases: Database queries, data cleaning, format conversion, statistical analysis, report generation

TBD: Skill list for this category (contributors can add entries here)

Business & Marketing

Skills for brand management, competitive analysis, domain research, internal communications, and marketing content creation.

Common use cases: Brand guidelines enforcement, competitor research, content strategy, domain name ideation, internal announcements

TBD: Skill list for this category (contributors can add entries here)

Communication & Writing

Skills for content writing, editing, meeting analysis, article extraction, research assistance, and communication optimization.

Common use cases: Content drafting, research synthesis, meeting insights, writing improvement, citation management

TBD: Skill list for this category (contributors can add entries here)

Creative & Media

Skills for image generation, video processing, GIF creation, design work, and visual art production.

Common use cases: Image enhancement, video downloading, animation creation, visual design, media optimization

runapi-cli-skill - Run AI image, video, music/audio, and LLM model jobs from agent workflows through RunAPI.

Productivity & Organization

Skills for file organization, task management, workflow automation, and productivity enhancement.

Common use cases: File system organization, invoice processing, task tracking, automation setup, workflow optimization

TBD: Skill list for this category (contributors can add entries here)

Collaboration & Project Management

Skills for Git workflows, code review processes, team coordination, and project management tasks.

Common use cases: Version control automation, pull request management, code review, team synchronization, project planning

TBD: Skill list for this category (contributors can add entries here)

Security & Systems

Skills for digital forensics, threat hunting, security analysis, and system administration.

Common use cases: Forensic investigation, security auditing, threat detection, system analysis, incident response

TBD: Skill list for this category (contributors can add entries here)

By Platform

Codex

Codex is OpenAI's AI-powered coding assistant that works across CLI and IDE extensions. Codex Skills extend the platform with task-specific capabilities through bundled instructions, resources, and optional scripts.

Skills are packaged as folders that encapsulate reusable workflows; for the shared structure, see Creating Skills → Skill Structure.

Installation Locations:

Skills load hierarchically (highest precedence first):

Scope	Location	Use Case
REPO	`$CWD/.codex/skills`	Task-specific for current directory
REPO	`$CWD/../.codex/skills`	Shared for nested repositories
REPO	`$REPO_ROOT/.codex/skills`	Organization-wide skills
USER	`$CODEX_HOME/skills`	Personal cross-project skills
ADMIN	`/etc/codex/skills`	System-level for all users
SYSTEM	Bundled	Default skills

Creating Skills:

Using the Skill Creator:

$skill-creator

The built-in tool guides you through defining triggers and structure.

Manual Creation: Follow the shared template in Creating Skills → Basic Skill Template. Common Codex frontmatter limits are name max 100 chars and description max 500 chars; restart Codex to load.

Invoking Skills:

Explicit: Type $skill-name or use /skills command
Implicit: Codex auto-activates based on description match

Installing Additional Skills:

$skill-installer linear
$skill-installer notion-spec-to-implementation

For docs, see Resources → Official Documentation.

See: skills/codex/ for Codex-specific skill examples.

Claude Code

Claude Code is Anthropic's official CLI tool that brings Claude's capabilities directly into your terminal and development workflow. Agent Skills extend Claude with task-specific capabilities through organized instructions, scripts, and resources.

What Are Agent Skills:

Agent Skills are modular capabilities that extend Claude's functionality. Each Skill packages instructions, metadata, and optional resources that Claude automatically uses when relevant. Skills follow the open agent skills standard and work across Claude.ai, Claude Code, Claude API, and Claude Agent SDK.

Why Use Skills:

Specialize Claude: Tailor capabilities for specific domain tasks
Reduce Repetition: Build once, use automatically
Compose Functionality: Combine Skills for complex workflows

For the shared Skill structure, see Creating Skills → Skill Structure.

YAML Frontmatter (Required):

---
name: your-skill-name
description: What this Skill does and when to use it
---

Requirements:

name: Max 64 chars, lowercase/numbers/hyphens only, no XML tags, no reserved words
description: Max 1024 chars, non-empty, no XML tags

Progressive Disclosure - Three Loading Levels:

Skills use a three-level loading system to efficiently manage context:

Level	When Loaded	Token Cost	Content
Level 1: Metadata	Always (at startup)	~100 tokens per Skill	`name` and `description` from YAML frontmatter
Level 2: Instructions	When Skill triggered	<5k tokens	SKILL.md body with instructions and guidance
Level 3+: Resources	As needed	Virtually unlimited	Bundled files accessed via bash, no context cost until read

Claude only loads what's needed for each task, keeping context usage efficient.

Installation Locations:

For Claude Code (User-Level):

# Create user-level skills directory
mkdir -p ~/.config/claude-code/skills/

# Copy skill folder to the skills directory
cp -r your_skill_name ~/.config/claude-code/skills/

# Verify the skill is recognized
claude --list-skills

For Project-Specific:

# Create project-level skills directory
mkdir -p .claude/skills/

# Copy skill to project
cp -r your_skill_name .claude/skills/

Skills in .claude/skills/ are only available for that specific project, while skills in ~/.config/claude-code/skills/ are available globally.

Creating Skills:

Follow the shared template in Creating Skills → Basic Skill Template and respect the name / description constraints above; then copy into the appropriate directory and validate with claude --list-skills.

How Skills Are Discovered:

Startup: Load all Skill metadata (name + description)
Task Matching: Requests matching a description auto-trigger the Skill
On-demand loading: Read SKILL.md and other bundled resources only when needed

Best Practices:

Keep SKILL.md body under 500 lines
Use progressive disclosure for large Skills
Write descriptions that clearly state what and when
Test with Haiku, Sonnet, and Opus if using multiple models
Only use Skills from trusted sources (security)

For docs, see Resources → Official Documentation.

See: skills/claude_code/ for Claude Code-specific skill examples.

kiro-cli

TBD

Getting Started

For Beginners

Choose your platform: Codex or Claude Code
Browse categories: Find skills relevant to your use case
Install a skill: Follow platform-specific instructions
Test it out: Run the skill on sample data
Customize: Adapt the skill to your specific needs

For contributing, see Contributing.

Creating Skills

Skill Structure

Each skill is organized as a folder containing:

skill_name/
├── SKILL.md          # Required: Skill instructions and metadata
├── scripts/          # Optional: Executable scripts
├── references/       # Optional: Reference documents/data
└── assets/           # Optional: Templates and static assets

Basic Skill Template

---
name: my-skill-name
description: A clear, concise description of what this skill does
platforms: [codex, claude-code]
tags: [category, type, use-case]
version: 1.0.0
---

# Skill Name

## Objective
What this skill accomplishes and what "done" looks like.

## When to Use This Skill
- Use case 1
- Use case 2
- Use case 3

## Inputs
- **Input A**: Description and format
- **Input B**: Description and format

## Outputs
- **Output A**: Description and format
- **Output B**: Description and format

## Workflow
1. Step 1: Detailed description
2. Step 2: Detailed description
3. Step 3: Detailed description

## Constraints & Guardrails
- What not to do
- Safety considerations
- Security notes
- Stop conditions

## Examples

### Example 1
**Input:**
\```
[Example input]
\```

**Expected Output:**
\```
[Example output]
\```

## Evaluation Checklist
- [ ] Correctness: Output matches expected format and content
- [ ] Completeness: All required fields are present
- [ ] Safety: No security or privacy issues
- [ ] Reproducibility: Same inputs produce same outputs

Best Practices

Be Specific: Clear, unambiguous instructions
Handle Edge Cases: Account for unusual inputs
Document Assumptions: State what you expect
Include Examples: Real-world usage demonstrations
Test Thoroughly: Verify across different scenarios
Keep It Simple: Avoid unnecessary complexity
Version Control: Track changes and updates
Security First: Never expose sensitive data

Design Patterns

This section is a grab-bag of reusable workflow templates for designing Skills. Use them to improve testability, repeatability, and safety; you generally only need 1–2 patterns per Skill.

Plan → Execute → Verify

Break tasks into three phases: planning the approach, executing the steps, and verifying the results.

When to use: Tasks that can drift without clear acceptance checks.

Spec-First Workflows

Define inputs, outputs, and constraints before execution. This ensures clarity and prevents scope creep.

When to use: Output format matters or requirements are likely to change.

Tool-First Workflows

Prefer structured tools over free-form text when possible. Tools provide better error handling and validation.

When to use: You have CLIs/APIs/tests/formatters available to validate results.

Critic / Reviewer Loop

Implement self-checking against rubrics or checklists. Have skills review their own output for quality.

When to use: Outputs must strictly follow a checklist or spec.

Minimal Diff

For code edits, make the smallest safe changes necessary. Avoid refactoring beyond the immediate need.

When to use: You want to reduce risk and avoid unintended side effects.

RAG-Lite

When researching, quote and cite sources. Maintain traceability from conclusions back to evidence.

When to use: You need to verify facts, quote documents, or reduce hallucination risk.

Composability

Design skills to work together. Outputs from one skill should be usable as inputs to another.

When to use: You plan to chain multiple Skills into a workflow.

Idempotency

Skills should produce the same results when run multiple times with the same inputs.

When to use: You want reliable re-runs, regression checks, and predictable outputs.

Evaluation & Quality

Quality Checklist

Every skill should address these dimensions:

Clear I/O Contract: Well-defined inputs and outputs with examples
Edge Cases: Handling of boundary conditions and errors
Determinism: Predictable behavior where possible
Stop Conditions: Clear criteria for completion or failure
Security Notes: Awareness of secrets, injection risks, and data safety
Example Run: At least one complete example with expected output
Documentation: Clear instructions for users and maintainers

Testing Recommendations

Unit Tests: For individual skill components
Integration Tests: For skill interactions with tools
Golden Files: Reference outputs for regression testing
User Acceptance: Real-world validation with target users

Quality Metrics

Success Rate: Percentage of successful executions
Error Rate: Frequency of failures or exceptions
Performance: Execution time and resource usage
User Satisfaction: Feedback from skill users

Security & Safety

Best Practices

Never paste secrets: Use environment variables and secret managers
Treat external content as untrusted: Validate and sanitize all inputs
Least-privilege tool access: Grant minimum necessary permissions
Log and review actions: Maintain audit trails for automated operations
Validate outputs: Check results before committing or publishing
Rate limiting: Implement throttling for API-dependent skills
Data privacy: Respect user data and privacy regulations

Common Risks

Prompt Injection: Malicious inputs that hijack skill behavior
Data Leakage: Accidental exposure of sensitive information
Unintended Actions: Skills performing operations beyond their scope
Resource Exhaustion: Skills consuming excessive compute or memory

Contributing

We welcome contributions! Please see CONTRIBUTING.md for detailed guidelines.

Quick Start

Fork this repository
Create a new branch for your skill
Follow the skill template structure
Add your skill to the appropriate category
Submit a pull request with a clear description

Pre-Submit Checklist

Check for duplicates or merge opportunities
Include at least one reproducible full example (inputs + expected outputs)
Validate basics on relevant platforms (loads successfully and completes the workflow)
Document what/when/inputs-outputs/constraints clearly

Resources

Official Documentation

Codex Skills Guide - Official Codex Skills documentation
Creating Codex Skills - Guide for creating Codex Skills
Claude Agent Skills Overview - Complete guide to Agent Skills
Claude Skills Quickstart - Getting started with Skills
Claude Skills Best Practices - Writing effective Skills
Claude Skills API Guide - Using Skills with Claude API

License

This repository is licensed under the MIT License.

Individual skills may have different licenses - please check each skill's folder for specific licensing information.

Note: This is a curated list designed to help you discover and create agent skills across multiple platforms. Skills are meant to be practical, composable, and production-ready. Start with a use case, find or create a skill, and iterate based on real-world feedback.

Awesome Agent Skills

Contents

What is a Skill?

Skill Categories

Document Processing

Development & Code Tools

Data & Analysis

Business & Marketing

Communication & Writing

Creative & Media

Productivity & Organization

Collaboration & Project Management

Security & Systems

By Platform

Codex

Claude Code

kiro-cli

Getting Started

For Beginners

Creating Skills

Skill Structure

Basic Skill Template

Best Practices

Design Patterns

Plan → Execute → Verify

Spec-First Workflows

Tool-First Workflows

Critic / Reviewer Loop

Minimal Diff

RAG-Lite

Composability

Idempotency

Evaluation & Quality

Quality Checklist

Testing Recommendations

Quality Metrics

Security & Safety

Best Practices

Common Risks

Contributing

Quick Start

Pre-Submit Checklist

Resources

Official Documentation

License

Yorumlar (0)