Author high-quality Claude Code skills with proper frontmatter, trigger-optimized descriptions, progressive disclosure structure, and MCP tool integration. Includes file organization patterns, testing approaches, and performance tuning. Consult when building new skills, debugging activation issues, or refactoring existing skill content.
Creates Claude Code skills with proper structure, optimized triggers, and progressive disclosure patterns for reliable activation.
npx claudepluginhub aeyeops/aeo-skill-marketplaceThis skill inherits all available tools. When active, it can use any tool Claude has access to.
QUICK-REFERENCE.mdexamples/code-skill.mdexamples/multi-file-skill.mdexamples/simple-skill.mdreferences/anti-patterns.mdreferences/doc-retrieval.mdreferences/file-organization.mdreferences/mcp-tools.mdreferences/progressive-disclosure.mdreferences/testing.mdscripts/init_skill.pyscripts/package_skill.pyComprehensive guide for authoring high-quality Claude skills.
These rules are NON-NEGOTIABLE. Violating them wastes tokens and breaks skills.
See [guide.md], then guide.md MUST exist with content| Complexity | Max Files | When to Use |
|---|---|---|
| Simple | 1-3 | Most skills |
| Moderate | 4-6 | Multi-domain skills |
| Complex | 7-10 | Rare, justify each file |
| 15+ files | NEVER | You are over-engineering |
Detailed anti-pattern examples: See references/anti-patterns.md
Before creating skills, fetch current Anthropic specifications:
mcp__context7__resolve-library-id with "claude code skills"mcp__context7__get-library-docs with ID /websites/code_claude_en topic "skills"WebFetch url="https://code.claude.com/docs/en/skills"
prompt="Extract skill file structure, frontmatter requirements, and best practices"
Complete retrieval guide: See references/doc-retrieval.md
Your skill shares context with system prompt, conversation history, other skills, and user requests.
Loading Model:
name + description (~100 tokens per skill)Best Practice: Keep SKILL.md under 500 lines.
---
name: skill-name
description: What it does. Use when [triggers].
allowed-tools: Read, Grep, Glob # Optional: restrict tool access
---
| Field | Limit | Notes |
|---|---|---|
name | 64 chars | Lowercase, hyphens only. No "anthropic"/"claude" |
description | 1024 chars | Third person. Include WHAT + WHEN |
allowed-tools | Optional | Comma-separated tool list |
Use verb + -ing:
processing-pdfs not pdf-processoranalyzing-spreadsheets not spreadsheet-analyzergenerating-commits not commit-generator[What the skill does]. Use when [trigger conditions].
Good:
description: Extract text and tables from PDF files, fill forms, merge documents. Use when working with PDF documents or form automation.
Bad:
description: Helps with documents and files when needed.
Simple (1 file):
my-skill/
└── SKILL.md
With References (3-4 files):
my-skill/
├── SKILL.md
├── reference.md
└── examples.md
With Scripts (5-7 files):
pdf-skill/
├── SKILL.md
├── FORMS.md
└── scripts/
├── analyze_form.py
└── fill_form.py
Complete guide: See references/file-organization.md
Claude navigates skills like a filesystem, reading files only when needed.
Keep references one level deep from SKILL.md:
SKILL.md
├── references to → advanced.md ✓
├── references to → reference.md ✓
└── references to → examples.md ✓
Avoid nested references:
SKILL.md → advanced.md → details.md → more.md ✗
For files >100 lines: Include a table of contents.
Complete patterns: See references/progressive-disclosure.md
Always use fully qualified tool names: ServerName:tool_name
Use `Salesforce:query` for SOQL queries.
Use `BigQuery:bigquery_schema` for table schemas.
Use `mcp__context7__get-library-docs` for documentation.
Why: Tool names conflict across servers. Qualified names ensure correct invocation.
Complete guide: See references/mcp-tools.md
Gather concrete usage examples. Ask: "What would trigger this skill?"
Identify resources needed. STOP AND CHECK: Can this be 1-3 files?
Run python scripts/init_skill.py <skill-name> --path <output-dir>
Implement SKILL.md and referenced files. Every file you reference must exist.
Run python scripts/package_skill.py <skill-dir> to validate and package.
Test with realistic scenarios. Refine based on Claude's navigation patterns.
Use the three-agent pattern:
Watch for:
Complete guide: See references/testing.md
---).claude/skills/<name>/SKILL.md/)Only use skills from trusted sources:
Skills can direct Claude to invoke tools in unintended ways. Audit untrusted skills thoroughly.
processing-pdfs)Complete working examples:
The Claude CLI has direct plugin management commands (not just interactive /plugin):
claude plugin marketplace add <source> # Add marketplace from URL, path, or GitHub repo
claude plugin marketplace list # List configured marketplaces
claude plugin marketplace remove <name> # Remove a marketplace
claude plugin marketplace update [name] # Update marketplace(s)
claude plugin validate <path> # Validate plugin/marketplace manifest
claude plugin install <plugin> # Install plugin from marketplace
Supported marketplace sources:
/path/to/marketplace or ./relative/pathowner/repo (shorthand)https://github.com/owner/repo.gitNot supported: Azure DevOps URLs (use local clone instead)
Activates when the user asks about AI prompts, needs prompt templates, wants to search for prompts, or mentions prompts.chat. Use for discovering, retrieving, and improving prompts.
Search, retrieve, and install Agent Skills from the prompts.chat registry using MCP tools. Use when the user asks to find skills, browse skill catalogs, install a skill for Claude, or extend Claude's capabilities with reusable AI agent components.
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.