Star200+

AI Agent Skill Repository

AI Agent Skills Directory

Browse reusable skills for Codex, Claude Code, Cursor, finance, research, web scraping, PPT, football analytics, data, marketing, design, and more.

Finance Stock analysisQuant Trading researchSlides PPT generationDocs PDF parsingExtract Web scraping
16
Shown
10
Tracks
48
Targets
Resolve APITask in, selected skill and install plan out.Skills APIMachine-readable shortlist with trust and safety signals.llms.txtA crawler-friendly entry point for agent surfaces.

Browse by scenario

Real skills, grouped by the work your agent needs to finish.

Each directory entry links to real skill pages with GitHub adoption, trust score, install handoff, risk notes, and agent-readable metadata. Use these as starting points when you want a shortlist before asking an agent to install anything.

Design to deployment

Frontend and UI skills

Design direction, Figma implementation, UI review, React performance, browser QA, and safe preview deployments.

  • Emil Design Engineering13K stars

    This skill encodes Emil Kowalski's philosophy on UI polish, component design, animation decisions, and the in…

    Trust 91Quality 100design-creative
  • Review Animations18K stars

    Reviews animation and motion code against a high craft bar derived from Emil Kowalski's design engineering ph…

    Trust 91Quality 100design-creative
  • Apple Design18K stars

    Apple's approach to interface design and fluid, physical motion, translated for the web. Use when building or…

    Trust 91Quality 100design-creative

Codex, Claude Code, Cursor

Coding agent skills

Code review, repo inspection, testing, planning, shipping, and engineering workflows for coding agents.

  • Skills122 stars

    Agent Skills for the Venice.ai API. One folder per surface area, each with a SKILL.md for agent runtimes (Cur…

    Trust 84Quality 87utility
  • Seedance 2.0 Prompt Writing3.1K stars

    Write precise Seedance 2.0 prompts for multimodal video, camera movement, editing, music, and product storyte…

    Trust 79Quality 92design-creative
  • Claude Code Vision Skill92 stars

    为 Claude Code 赋能多模态视觉能力，支持豆包、通义千问、GPT-4o 等模型，用于截图 / UI / 图表分析；适配 DeepSeek 等无视觉底座，搭配 browser-harness 可做前端布局自动化…

    Trust 75Quality 71coding-agents

Documents and knowledge

Research and RAG skills

Research, recent-events briefings, PDF parsing, markdown conversion, RAG ingestion, and knowledge workflows.

  • MarkEverythingDown339 stars

    Convert files (PDF, image, Word, PPT, Excel, notebooks, code snippets) to markdown using powerful multimodal…

    Trust 71Quality 46document-processing
  • Haystack26K stars

    Open-source AI orchestration framework for building context-engineered, production-ready LLM applications. De…

    Trust 90Quality 100data
  • Ruby Llm4.0K stars

    One delightful Ruby framework for every major AI provider. Build AI agents, chatbots, RAG apps, and multimoda…

    Trust 88Quality 98rag-knowledge

Crawlers and extraction

Web scraping skills

Crawling, scraping, extraction, browser automation, structured data capture, and website-to-markdown workflows.

  • Versatile OCR Program682 stars

    Multi-modal OCR pipeline optimized for ML training (text, figure, math, tables, diagrams)

    Trust 77Quality 69document-processing
  • Unilm22K stars

    Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities

    Trust 86Quality 97document-processing
  • MixTeX Latex OCR1.6K stars

    MixTeX multimodal LaTeX, ZhEn, and, Table OCR. It performs efficient CPU-based inference in a local offline o…

    Trust 81Quality 66document-processing

Slides and decks

PPT and presentation skills

Presentation generation, editable PPTX, slide decks, speaker notes, and visual storytelling workflows.

  • VLM2Vec655 stars

    This repo contains the code for "VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Ta…

    Trust 80Quality 76rag-knowledge
  • MultiBench631 stars

    [NeurIPS 2021] Multiscale Benchmarks for Multimodal Representation Learning

    Trust 71Quality 48robotics-iot

Prompts, B-roll, explainers

Video creation skills

Video-generation prompts, B-roll, Vox-style explainers, camera direction, captions, and creative-production workflows.

  • InternGPT3.2K stars

    InternGPT (iGPT) is an open source demo platform where you can easily showcase your AI models. Now it support…

    Trust 81Quality 75media-automation
  • Phantom1.5K stars

    Phantom: Subject-Consistent Video Generation via Cross-Modal Alignment

    Trust 82Quality 79media-automation
  • HunyuanCustom1.2K stars

    HunyuanCustom: A Multimodal-Driven Architecture for Customized Video Generation

    Trust 79Quality 73media-automation

Images, video, UI

Design and creative skills

Image, video, creative production, UI design, multimodal generation, and visual workflow skills.

  • Lance1.2K stars

    A 3B-active-parameter native unified multimodal model for image and video understanding, generation, and edit…

    Trust 87Quality 93media-automation
  • ARIS Movie Director46 stars

    Agentic, long-horizon visual generation: a fuzzy story → a cross-model-audited image-based movie. Brings ARIS…

    Trust 77Quality 77utility
  • Reconstruction Alignment406 stars

    [ICLR 2026] Official repo of paper "Reconstruction Alignment Improves Unified Multimodal Models". Unlocking t…

    Trust 77Quality 74media-automation

Analysis and pipelines

Data and analytics skills

Data analysis, analytics, ETL, notebooks, databases, tables, charts, and reporting workflows.

  • TDC1.3K stars

    Therapeutics Commons (TDC): Multimodal Foundation for Therapeutic Science

    Trust 79Quality 69geo-science
  • BioReason395 stars

    BioReason: Incentivizing Multimodal Biological Reasoning within a DNA-LLM Model | NeurIPS '25

    Trust 78Quality 74geo-science
  • Farmvibes AI874 stars

    FarmVibes.AI: Multi-Modal GeoSpatial ML Models for Agriculture and Sustainability

    Trust 75Quality 54geo-science

Popular directories

AI agent skills libraryA real library of reusable skills with trust, audit, install, and agent metadata.Open source AI agent skillsOpen-source skill candidates filtered by task fit, GitHub quality, and install safety.Best Codex skillsRepo inspection, code review, testing, web data, and safe install workflows for Codex.Best AI agent skills for stock analysisFinance and market research skills with trust, install, and maintenance signals.Best Codex skills for web scrapingCrawler and extraction skills that agents can shortlist before installing.Best Claude Code skills for PDF parsingDocument conversion, table extraction, OCR, and markdown workflows.Best AI agent skills for football analyticsMatch data, scouting, dashboards, and World Cup analytics workflows.OpenAgentSkill vs skills.shCompare task resolution, trust signals, install handoffs, and registry APIs.OpenAgentSkill vs agentskills.ioCompare directories, agent-readable metadata, audits, and discovery workflows.

Supply tracks

Build the registry by domain, not just by count.

Coding

Coding and developer agents

204

Code review, repo analysis, testing, CI, GitHub, DevOps, and developer workflow skills.

201 quality187 maintained

Research

Research and knowledge work

89

Deep research, source comparison, literature review, RAG, knowledge search, and reports.

87 quality81 maintained

Presentation

Presentation and deck workflows

1

PPTX generation, HTML slides, pitch decks, speaker notes, and presentation workflow skills.

1 quality1 maintained

Finance

Finance and quant workflows

13

Market data, SEC filings, portfolio analysis, quant research, backtesting, and risk workflows.

13 quality13 maintained

Marketing

Marketing and growth automation

12

SEO, content operations, lead generation, CRM, email automation, analytics, and growth workflows.

12 quality12 maintained

Design

Design and creative production

92

Design assets, images, video, audio, multimodal media, presentation, and creative production skills.

86 quality73 maintained

Data

Data, BI, and analytics

36

CSV, SQL, notebooks, dashboards, data pipelines, BI, ETL, and spreadsheet analysis.

36 quality33 maintained

Legal

Legal, policy, and compliance

10

Contract analysis, privacy, policy review, compliance checks, governance, and document risk review.

10 quality10 maintained

Education

Education and tutoring

5

Tutoring, course generation, quizzes, learning analytics, classrooms, and teaching workflows.

5 quality5 maintained

World Cup

Football and World Cup analytics

2

Football data, World Cup dashboards, xG, match prediction, scouting, and sports analytics.

2 quality2 maintained

High-intent entry points

Start from the task, not a keyword list.

These shortcuts use the same trust, supply, and relevance signals as the registry API, so humans and agents land on a useful shortlist faster.

Agent-readable index
FinanceStock analysisQuantTrading researchSlidesPPT generationDocsPDF parsingExtractWeb scrapingCreativeDesign workflowSportsFootball analytics

Decision filters

Choose by scenario, quality, and trust signals.

Showing 1-16 of 451 ranked candidates matching "mulit-modal"

Best blend of quality, stars, freshness, and agent usage

1

Modal Client

STRONG · 75TRUST · 77SAFE · REVIEWEDDESIGN

SDK libraries for Modal

$ npx skills add modal-labs/modal-client
481 stars52 quality77 trustReviewed with permission notes1mo since pushNeeds review
QualitySolid option that is likely worth shortlisting for production workflows.
TrustGood trust signals with a few areas worth checking before rollout.Review: Quality score needs review
Safety gateUsable candidate, but the agent should surface permission and audit notes before installation.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonmachine-learning
by modal-labsDetailsQuick view
2

Multimodal RAG Survey

PROMISING · 63TRUST · 74SAFE · REVIEWEDCODING

A Survey on Multimodal Retrieval-Augmented Generation

$ npx skills add llm-lab-org/Multimodal-RAG-Survey
524 stars46 quality74 trustReviewed with permission notes5mo since pushNeeds review
QualityUseful candidate, but compare it with alternatives before adopting.
TrustGood trust signals with a few areas worth checking before rollout.Review: License is unclear
Safety gateUsable candidate, but the agent should surface permission and audit notes before installation.

Scenario GitHub automation

CLI + Codex · 4 targets

rag
by llm-lab-orgDetailsQuick view
3

Multi Modality Arena

NEEDS REVIEW · 43TRUST · 69SAFE · EXPERIMENTALDESIGN

Chatbot Arena meets multi-modality! Multi-Modality Arena allows you to benchmark vision-language models side-by-side while providing images as inputs. Supports MiniGPT-4…

$ npx skills add OpenGVLab/Multi-Modality-Arena
567 stars38 quality69 trustExperimentalOpenAI Agents2y since pushNeeds review
QualityInspect the repository carefully before adding it to an agent workflow.Check: Repository looks stale
TrustPotentially useful, but at least one trust signal needs human inspection.Review: License is unclear
Safety gateSparse or mixed signals. Useful for discovery, but not for autonomous installation.

Scenario Multimodal media

OpenAI Agents + CLI · 4 targets

pythonchatbot
by OpenGVLabDetailsQuick view
4

ModAL

VERIFIEDPROMISING · 68TRUST · 76SAFE · EXPERIMENTALDESIGN

A modular active learning framework for Python

$ npx skills add modAL-python/modAL
2.4K stars50 quality76 trustExperimental2y since pushNeeds review
QualityUseful candidate, but compare it with alternatives before adopting.Check: Repository looks stale
TrustGood trust signals with a few areas worth checking before rollout.Review: Repository looks stale
Safety gateSparse or mixed signals. Useful for discovery, but not for autonomous installation.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonmachine-learning
by modAL-pythonDetailsQuick view
5

MultiModalMamba

STRONG · 75TRUST · 78SAFE · REVIEWEDDESIGN

A novel implementation of fusing ViT with Mamba into a fast, agile, and high performance Multi-Modal Model. Powered by Zeta, the simplest AI framework ever.

$ npx skills add kyegomez/MultiModalMamba
473 stars52 quality78 trustReviewed1mo since pushSafe to try
QualitySolid option that is likely worth shortlisting for production workflows.
TrustGood trust signals with a few areas worth checking before rollout.Review: Quality score needs review
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonmachine-learning
by kyegomezDetailsQuick view
6

Multimodal Garment Designer

NEEDS REVIEW · 42TRUST · 67SAFE · BLOCKEDCODING

This is the official repository for the paper "Multimodal Garment Designer: Human-Centric Latent Diffusion Models for Fashion Image Editing". ICCV 2023

$ npx skills add aimagelab/multimodal-garment-designer
446 stars37 quality67 trustBlocked for auto-install2y since pushRisky
QualityInspect the repository carefully before adding it to an agent workflow.Check: Repository looks stale
TrustPotentially useful, but at least one trust signal needs human inspection.Review: License is unclear
Safety gateThis skill should not be selected by an agent without explicit human security review.

Scenario Coding agents

CLI + Codex · 4 targets

pythoncomputer-vision
by aimagelabDetailsQuick view
7

UI-TARS Desktop

VERIFIEDEXCELLENT · 100TRUST · 85SAFE · REVIEWEDCODING

Run multimodal agents that operate desktop interfaces

$ npx skills add bytedance/UI-TARS-desktop
37.0K stars75 quality85 trustReviewed with permission notes2mo since pushNeeds review
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustGood trust signals with a few areas worth checking before rollout.Review: License is unclear
Safety gateUsable candidate, but the agent should surface permission and audit notes before installation.

Scenario Coding agents

CLI + Codex · 4 targets

ui-tarsdesktoptypescript
by bytedanceDetailsQuick view
8

Crawl4AI

VERIFIEDEXCELLENT · 100TRUST · 90SAFE · REVIEWEDRESEARCH

Web crawling built for AI

$ npx skills add unclecode/crawl4ai
4 agent calls100% success73.1K stars80 quality90 trustReviewedClaude Code + OpenAI Agents18d since pushSafe to try31.0K installs
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustStrong OpenAgentSkill Trust Score across adoption, recent maintenance, license clarity, documentation, dependency/…Review: Documentation summary is thin
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Research agents

Claude Code + OpenAI Agents · 4 targets

claudegpt-4langchaincrewaiopenclaw
by unclecodeDetailsQuick view
9

Vllm Omni

VERIFIEDEXCELLENT · 100TRUST · 87SAFE · REVIEWEDDESIGN

A framework for efficient model inference with omni-modality models

$ npx skills add vllm-project/vllm-omni
5.2K stars68 quality87 trustReviewed2mo since pushSafe to try
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustStrong OpenAgentSkill Trust Score across adoption, recent maintenance, license clarity, documentation, dependency/…Review: Documentation summary is thin
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonimage-generation
by vllm-projectDetailsQuick view
10

Seedance 2.0 Prompt Writing

VERIFIEDEXCELLENT · 92TRUST · 79SAFE · REVIEWEDDESIGN

Write precise Seedance 2.0 prompts for multimodal video, camera movement, editing, music, and product storytelling.

$ npx skills add dexhunter/seedance2-skill --skill seedance-prompt-en
3.1K stars61 quality79 trustReviewedClaude Code + OpenAI Agents6mo since pushSafe to try
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.Check: Model capabilities, asset limits, and platform rules can change; verify them be…
TrustGood trust signals with a few areas worth checking before rollout.Review: Model capabilities, asset limits, and platform rules can change; verify them be…
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Video creation

Claude Code + OpenAI Agents · 4 targets

claude-codecodexcursorseedance-2.0
by dexhunterDetailsQuick view
11

Parlor

VERIFIEDEXCELLENT · 95TRUST · 87SAFE · REVIEWEDDESIGN

On-device, real-time multimodal AI. Have natural voice and vision conversations with an AI that runs entirely on your machine. Powered by Gemma 4 E2B and Kokoro.

$ npx skills add fikrikarim/parlor
1.9K stars65 quality87 trustReviewed2mo since pushSafe to try
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustStrong OpenAgentSkill Trust Score across adoption, recent maintenance, license clarity, documentation, dependency/…Review: Documentation summary is thin
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Multimodal media

CLI + Codex · 4 targets

htmlspeech
by fikrikarimDetailsQuick view
12

Quillman

VERIFIEDEXCELLENT · 90TRUST · 85SAFE · REVIEWEDDESIGN

A voice chat app

$ npx skills add modal-labs/quillman
1.2K stars60 quality85 trustReviewed2mo since pushSafe to try
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustGood trust signals with a few areas worth checking before rollout.Review: Documentation summary is thin
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonspeech
by modal-labsDetailsQuick view
13

HunyuanImage 3.0

VERIFIEDSTRONG · 81TRUST · 79SAFE · REVIEWEDDESIGN

HunyuanImage-3.0: A Powerful Native Multimodal Model for Image Generation

$ npx skills add Tencent-Hunyuan/HunyuanImage-3.0
3.1K stars59 quality79 trustReviewed with permission notes6mo since pushNeeds review
QualitySolid option that is likely worth shortlisting for production workflows.
TrustGood trust signals with a few areas worth checking before rollout.Review: License is unclear
Safety gateUsable candidate, but the agent should surface permission and audit notes before installation.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonimage-generation
by Tencent-HunyuanDetailsQuick view
14

Docarray

VERIFIEDEXCELLENT · 90TRUST · 84SAFE · REVIEWEDRESEARCH

Represent, send, store and search multimodal data

$ npx skills add docarray/docarray
3.1K stars59 quality84 trustReviewed4mo since pushSafe to try
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustGood trust signals with a few areas worth checking before rollout.Review: Documentation summary is thin
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario RAG and knowledge

CLI + Codex · 4 targets

pythonsemantic-search
by docarrayDetailsQuick view
15

Lance

VERIFIEDEXCELLENT · 93TRUST · 87SAFE · REVIEWEDDESIGN

A 3B-active-parameter native unified multimodal model for image and video understanding, generation, and editing.

$ npx skills add bytedance/Lance
1.2K stars63 quality87 trustReviewed2mo since pushSafe to try
QualityHigh-confidence pick with strong adoption and healthy maintenance signals.
TrustStrong OpenAgentSkill Trust Score across adoption, recent maintenance, license clarity, documentation, dependency/…Review: Documentation summary is thin
Safety gateGood audit and safety signals with no high-risk permission hints in public metadata.

Scenario Multimodal media

CLI + Codex · 4 targets

pythonimage-generation
by bytedanceDetailsQuick view
16

Open Flamingo

VERIFIEDSTRONG · 71TRUST · 80SAFE · REVIEWEDRESEARCH

An open-source framework for training large multimodal models.

$ npx skills add mlfoundations/open_flamingo
4.1K stars52 quality80 trustReviewed with permission notes2y since pushNeeds review
QualitySolid option that is likely worth shortlisting for production workflows.Check: Repository looks stale
TrustGood trust signals with a few areas worth checking before rollout.Review: Repository looks stale
Safety gateUsable candidate, but the agent should surface permission and audit notes before installation.

Scenario Research agents

CLI + Codex · 4 targets

pythoncomputer-vision
by mlfoundationsDetailsQuick view

Page 1

Showing the strongest 16 results to keep the registry fast for humans and agents. Refine by use case, platform, stars, or search query for a narrower shortlist.

Try the agent resolve API