Agent
Skills

What Are AI Agent Skills?

An agent skill is a packaged, reusable capability that teaches an AI agent how to complete one job end to end. Instead of re-explaining a workflow in every prompt, you install the skill once, and the agent loads its instructions, tools, and guardrails whenever a task matches. Skills turn what works once into something you can reuse across projects, clients, and teams.

How Agent Skills Work

Every skill in this library follows the same contract. A SKILL.md instruction file defines the job: the inputs it expects, the outputs it produces, the tools it may use (browser, terminal, HTTP, databases), and safety notes for destructive or ambiguous actions.

Point an agent such as Claude Code or OpenCode at the skill directory and give it the trigger phrase. The agent reads the instructions, plans the steps, calls only the allowed tools, and reports back structured results. Because the contract is plain markdown, the same skill works across agents -- and you can fork any skill here and adapt it to your stack.

Skills are deliberately small. A skill does one job -- sending a WhatsApp outreach sequence, ingesting a document set, auditing a landing page -- and does it repeatably. That makes it easy to test, easy to hand to non-engineers, and easy to chain: a lead research skill can feed a cold email skill, which feeds a CRM automation skill. Treat them as building blocks for larger workflows rather than monolithic assistants.

What You Can Automate

The eleven skills below cover four categories of real work:

• Sales and outreach: WhatsApp Sales Outreach, Cold Email Sequence, and Sponsor Outreach research leads, draft personalized messages, send campaigns, and track replies through to handoff.
• Research and knowledge: Playwright Lead Research browses websites and extracts structured lead data, while RAG Document Ingestion chunks documents, generates embeddings, and builds a searchable knowledge base.
• Business operations: Neon CRM Automation scores and routes pipeline leads, the Internal Reporting Agent ships scheduled reports from your tools, and Landing Page Audit reviews pages for conversion, SEO, and UX issues.
• Engineering: MCP Server Creation scaffolds tool servers over the Model Context Protocol, Code Review Workflow automates pull-request review, and Multi-Agent Orchestrator coordinates specialized sub-agents.

Each skill page documents the required tools, example outputs, and the recipes that show how the underlying system is built.

Skills, Recipes, and Patterns: Which Do You Need?

Skills are packaged capabilities you install and run. Recipes are step-by-step build-alongs that construct the system from scratch, so you understand and can customize every part. Patterns are design-level solutions to recurring agent problems like routing, chaining, and orchestration. If you want the outcome today, start with a skill; if you need to bend it to your environment, follow its related recipe.