skills-knowledge

# Skills Knowledge Base

Create, audit, and refine Claude Code skills following the 2026 unified skills-commands paradigm.

## 🚨 MANDATORY: Read BEFORE Creating Skills

**CRITICAL**: You MUST read and understand these URLs:

### Primary Documentation (MUST READ)
- **[MUST READ] Official Skills Guide**: https://code.claude.com/docs/en/skills
  - **Tool**: `mcp__simplewebfetch__simpleWebFetch`
  - **When to Read**: Before any skill creation or modification
  - **Content**: Skills architecture, YAML frontmatter, progressive disclosure
  - **Cache**: 15 minutes minimum

- **[MUST READ] Agent Skills Specification**: https://github.com/agentskills/agentskills
  - **Tool**: `mcp__simplewebfetch__simpleWebFetch`
  - **When to Read**: Before implementing any skill following the open standard
  - **Content**: Technical requirements, open standard specifications
  - **Cache**: 15 minutes minimum

### Examples & Patterns (SHOULD READ)
- **[SHOULD READ] Anthropic Skills Repository**: https://github.com/anthropics/skills
  - **Tool**: `mcp__simplewebfetch__simpleWebFetch`
  - **Content**: Official skill implementations, working examples

- **[SHOULD READ] Community Examples**: https://github.com/agentskills/examples
  - **Tool**: `mcp__simplewebfetch__simpleWebFetch`
  - **Content**: Diverse skill implementations, real-world patterns

### ⚠️ BLOCKING RULES
- **DO NOT proceed** until you've fetched and reviewed Primary Documentation
- **REQUIRED** to understand progressive disclosure before creating skills

## Core Principles

**Skills** are the PRIMARY building blocks for Claude Code plugins.

**Self-Sufficient Design**:
- Autonomous: Complete 80-95% of tasks without user questions
- Self-Sufficient: Work standalone AND enhance other workflows
- Discoverable: Auto-load through description trigger matching
- Progressive Disclosure: Tier 1 (metadata) → Tier 2 (instructions) → Tier 3 (resources)

## Skills vs Commands (2026)

**Unified Paradigm**: Commands and skills are unified -