Unnamed Skill

Create specialized Claude Code agents (subagents) with proper YAML frontmatter, tool permissions, and domain expertise. Use when: creating new agents, designing agent prompts, configuring agent tools and models. Triggers: create agent, new agent, agent template, subagent, agent design.

$ Instalar

git clone https://github.com/petbrains/mvp-builder /tmp/mvp-builder && cp -r /tmp/mvp-builder/.claude/skills/agent-creator ~/.claude/skills/mvp-builder

// tip: Run this command in your terminal to install the skill


name: agent-creator description: Create specialized Claude Code agents (subagents) with proper YAML frontmatter, tool permissions, and domain expertise. Use when: creating new agents, designing agent prompts, configuring agent tools and models. Triggers: create agent, new agent, agent template, subagent, agent design. allowed-tools: Read, Write

Agent Creator

Create specialized Claude Code agents in .claude/agents/ directory.

Template: See references/template.md for YAML structure, format rules, tools, skills, models, color, and examples.

Process

ANALYZE → DESIGN → IMPLEMENT → VALIDATE

Step 1: Analyze

Determine agent requirements:

  • What specific domain or task?
  • What tools needed?
  • What model appropriate?
  • Clear trigger conditions?

Step 2: Design

Agent Type

TypePatternExample
Technical[technology]-expertreact-expert, postgres-expert
Domain[domain]-specialistsecurity-specialist, performance-specialist
Process[process]-agentcode-review-agent, refactor-agent

Color by Domain

DomainColor
Frontendblue
Backendgreen
Securityred
Performanceyellow
Testingpurple
DevOpsgray

Step 3: Implement

Create file .claude/agents/<name>.md following @.claude/skills/agent-creator/references/template.md exactly.

Important: Follow Format Rules from template.md for YAML fields.

Step 4: Validate

Check before saving:

  • Name matches filename (kebab-case)
  • Description explains when to use
  • Tools minimal (only what needed)
  • YAML format per template.md Format Rules
  • Clear expertise boundaries
  • Constraints defined

Decision Rules

Tool Restriction

Restrict tools when:

  • Agent should only read, not modify
  • Limited to specific commands (e.g., Bash(git:*))
  • Security-sensitive domain

Grant all tools when:

  • General-purpose agent
  • Needs full file system access
  • Complex multi-step workflows