Skills Collection 2: 650+ Skill Resource Library for AI and Agent Developers

Introducing Skills Collection 2 maintained by pinkpixel-dev, the second edition of the AI and agent skill resource library. It contains 658 skill folders covering reusable workflows, security manuals, cloud implementation guides, scripts, and other resources. Complementing the first edition, it builds a systematic knowledge system and provides practical references for AI developers, agent builders, and operations personnel.

Original author/maintainer: pinkpixel-dev; Source platform: GitHub; Release/update date: 2026-05-28; Original link: https://github.com/pinkpixel-dev/skills-collection-2. This project is an extension of the first edition, with a larger scale. Complementing the first edition, it forms a systematic AI skill knowledge system and provides rich practical references for relevant personnel.

Adopts a decentralized directory structure: each skill is stored in an independent folder under the SKILLS directory, containing its own README, scripts, and other resources. Advantages include maintainability, discoverability, scalability, and independence. Each skill folder integrates Instructions (operation guides and best practices), References (links to papers, etc.), Scripts (automation scripts), and Assets (auxiliary materials), enabling one-stop resource organization and lowering the application threshold.

Covers four major areas: 1. Reusable workflows (multi-agent collaboration, prompt engineering standardization, model evaluation pipelines, RAG construction processes, etc.); 2. Security manuals (prompt injection protection, model output review, sensitive data processing, access control policies, etc.); 3. Cloud implementation guides (model deployment on mainstream cloud platforms, serverless function optimization, vector database selection, cost optimization, etc.); 4. Scripts and tools (dataset preprocessing, model benchmarking, API integration examples, log analysis scripts, etc.).

Browsing and discovery methods: Browse folders by name, GitHub keyword search, and explore References by association; Application process: Read README → View References → Run Scripts → Adapt and adjust → Production deployment; Community contribution methods: Submit new skills, improve documentation, share cases, report issues or suggestions.

Ecosystem positioning: Meta-resource library, technically neutral (not tied to a specific stack), composable, migratable, and continuously evolving. Community value: Aggregates scattered knowledge into reusable skill units, reduces learning costs, provides a common knowledge base for teams, and lowers communication costs and implementation risks.

Potential directions include multi-modal skills (images, audio, etc.), vertical domain skills (medical, legal, etc.), agent orchestration skills, AI security and governance skills. It is an important resource library for developers and teams to enhance their AI capabilities.