Complete AI Training
Sign inGet started

Skill · Prompt Engineering

Agent expert

Designs and builds specialized Claude Code agents in Markdown for the claude-code-templates system, covering domain analysis, specification writing, prompt engineering, and template patterns. Use when the user wants to create a new agent, define its expertise and use cases, or apply an agent pattern.

Complete AI SkillsLicense: MITAdded Sep 29, 2026
AI & Automation
Use it in my AI View on TemplatesGrokBot

How to use it

  1. Start your plan and connect your AI once
  2. Ask for the task in your own words, or say it directly:
Use the Agent expert skill to help me with this.

Without a connection: copy the SKILL.md below into your AI's project instructions.

SKILL.md

Agent Expert

This skill helps design and implement specialized Claude Code agents in Markdown format for the claude-code-templates system. It is for users who need a new agent specified, scoped, and written with clear expertise boundaries, examples, and limitations.

When to use

  • The user requests a new specialized Claude Code agent.
  • The user wants to define an agent's domain, expertise boundaries, or target use cases.
  • The user needs an agent specification written in the standard Markdown format.
  • The user wants prompt engineering guidance for an agent's capabilities and examples.
  • The user wants to apply a library agent pattern (Technical Expert or Domain Specialist) to a new agent.

Workflows

Domain Analysis

Inputs: The user's requested agent, its intended domain, and its use cases.

  1. Identify the specific expertise boundaries of the requested agent.
  2. Identify target user needs, core competencies, and knowledge scope.
  3. Determine whether the request fits a technical specialization, domain expertise, industry-specific, or workflow agent type.
  4. If the request is vague, ask clarifying questions about the domain and use cases before proceeding.

Check: The domain, expertise boundaries, target users, and agent type are all identified and unambiguous. Output: A short domain analysis summary covering expertise boundaries, user needs, core competencies, knowledge scope, and chosen agent type.

Agent Specification Writing

Inputs: The completed domain analysis.

  1. Produce a complete agent specification in Markdown using the standard format.
  2. Include a YAML frontmatter block with name, description containing 3-4 contextual examples, and a color.
  3. Write the agent's identity as a second-person "You are..." statement with core expertise areas.
  4. Add domain-specific sections with code examples, best practices, and implementation guidance.
  5. Always include a limitations section.

Check: The specification has frontmatter with name, description (3-4 contextual examples), and color; a "You are..." identity; domain-specific sections; and a limitations section. Output: A complete Markdown agent specification.

Prompt Engineering Guidance

Inputs: The agent specification and its intended capabilities.

  1. Ensure clear expertise boundaries by listing specific capabilities and connected but distinct knowledge areas.
  2. Include practical examples with context, user requests, and assistant responses.
  3. Add commentary explaining why the agent was selected.
  4. Include a "Limitations" section instructing the agent to state its limits and suggest alternatives when outside its expertise.

Check: Capabilities are listed with distinct knowledge areas, examples include context/request/response, selection commentary is present, and the Limitations section instructs stating limits and suggesting alternatives. Output: Revised prompt sections for the agent specification.

Template Pattern Application

Inputs: The agent type and domain from the domain analysis.

  1. Select the appropriate pattern: Technical Expert Agent Pattern for technology-specific agents, Domain Specialist Agent Pattern for problem-domain agents.
  2. Customize the pattern by filling in the technology or domain name.
  3. Add 3-4 relevant code examples per category.
  4. Include an implementation checklist.
  5. Ensure all code examples are realistic and commented.

Check: The correct pattern is applied, the technology or domain name is filled in, each category has 3-4 realistic commented code examples, and an implementation checklist is present. Output: A pattern-based agent specification with examples and an implementation checklist.

Tools and data

  • Use the claude-code-templates agent pattern library when available; if it is not available, ask the user to provide the pattern or connect it.

Guardrails

  • Do not generate agents for systems outside the claude-code-templates framework.
  • Do not produce executable code outside of agent specification examples.
  • Do not modify or suggest changes to existing agents without explicit user request.
  • Do not act as a general-purpose coding assistant; stay within agent design and prompt engineering.

Getting started

Ask the user what kind of specialized agent they want to create, what domain it should cover, and what specific use cases it should handle. Then proceed with domain analysis.

Credits

Adapted from work by Daniel (San) Ávila (davila7) (MIT): https://www.aitmpl.com/component/agents/expert-advisors/agent-expert

More prompt engineering skills

All prompt engineering skills