# skill.rbuild.ai > A curated library of Claude Code skills — plain Markdown files that give AI agents deep, reusable expertise on specific tasks. Download, install in ~/.claude/skills/, and invoke with /skill-name. Covers Claude Code, VS Code, Microsoft Copilot Studio, Dataverse, GitHub Copilot, Cursor, and Windsurf. Built by Ragnar Pitla (Principal PM, Microsoft Agentic Team | Founder, RBuild.ai). ## Key Pages - [Skill Library](https://skill.rbuild.ai/): Browse 470+ skills across personal and business domains - [Learn: Install on any platform](https://skill.rbuild.ai/learn): Step-by-step guides for Claude Code, VS Code, Copilot Studio, Dataverse, GitHub Copilot, Cursor, and Windsurf - [What is a Skill?](https://skill.rbuild.ai/learn#what-is-a-skill): Deep dive — anatomy, trigger system, quality bar, comparison to prompts and plugins - [Create a Skill](https://skill.rbuild.ai/create): 3-question wizard that generates a build prompt + downloads the Skill Creator ## Skill Creator The fastest way to build a production-quality skill is to use the Skill Creator skill itself: Download: https://skill.rbuild.ai/skill-creator-rbuild.md Install: mkdir -p ~/.claude/skills/skill-creator-rbuild && mv skill-creator-rbuild.md ~/.claude/skills/skill-creator-rbuild/SKILL.md Invoke: /skill-creator-rbuild [skill-name] [description or requirements] The Skill Creator follows Anthropic's official best practices: proper frontmatter, trigger phrases in the description field, three-level progressive disclosure (frontmatter → SKILL.md body → references/), and a 500-line limit. It generates complete SKILL.md files with optional reference folders and example scenario files. ## What Is a Skill? A skill is a plain Markdown file with YAML frontmatter that gives a Claude Code agent specific, reusable expertise. Skills live in ~/.claude/skills//SKILL.md (personal, all projects) or .claude/skills//SKILL.md (project-only). Frontmatter format: --- name: skill-slug description: "Third-person. What it does + trigger phrases. Max 1024 chars." user-invokable: true metadata: author: Your Name version: "1.0" --- The description field is the discovery engine — Claude reads only this to decide whether to activate the skill. It must include specific trigger phrases ("Use when user says 'review PR', 'security audit'..."). Skills load in three stages: 1. Frontmatter (~100 tokens) — always loaded at startup 2. SKILL.md body (<500 lines) — loaded when skill triggers 3. references/ folder (unlimited) — loaded on demand ## Skill Library Structure Skills are organized into two tabs: - Personal: software engineering, AI/agent development, frontend, backend, content creation, productivity, security, devops, language-specific - Business: Copilot Studio, D365 Finance & Operations, MCP (Model Context Protocol), Power Platform, enterprise AI, supply chain, industry verticals Each skill in the index has: slug, name, description, tab, domain, difficulty (starter/intermediate/advanced), source_type, tags, version, is_featured, is_coming_soon, raw_url, file_path. Skills marked "Coming Soon" are running evals before publish. Every published skill has been validated for trigger accuracy, step correctness, platform fidelity, and includes an example scenario. ## Platform Installation Quick Reference | Platform | Path | Format | |----------|------|--------| | Claude Code (personal) | ~/.claude/skills//SKILL.md | YAML frontmatter + Markdown | | Claude Code (project) | .claude/skills//SKILL.md | YAML frontmatter + Markdown | | VS Code native | .github/copilot-instructions.md | Markdown | | Copilot Studio | UI upload (Knowledge tab) | PDF, Word, Markdown, TXT | | Dataverse | Power Apps maker portal (Business Skills table) | Natural language text | | GitHub Copilot | .github/copilot-instructions.md | Markdown | | Cursor | .cursor/rules/*.mdc | Markdown | | Windsurf | .windsurf/rules/*.md | Markdown | ## Skill Authoring Rules (Quick Reference) - description: third-person, includes trigger phrases, max 1024 chars, no XML tags - name: kebab-case, max 64 chars, no "claude" or "anthropic" - SKILL.md body: under 500 lines — move detail to references/ - No over-explaining things Claude already knows - No time-sensitive dates or version numbers - Each step has a discrete action and a discrete output - Include Common Mistakes section to prevent bad responses ## Contact LinkedIn: https://www.linkedin.com/in/ragnarpitla/ RBuild.ai: https://rbuild.ai YouTube: https://www.youtube.com/@RagnarPitla