Build-Assistant

This skill provides templates, validation scripts, and documentation for building Claude Code framework components following the multiagent framework standards.

Instructions

When Creating a New Agent

1. Read the agent template:

   templates/agents/agent.md.template

2. Read the agent example:

   templates/agents/agent-example.md

3. Read documentation:

   docs/claude-code-agents.md

   (if available)
4. Create agent file with:
   • Proper frontmatter (name, description, tools, model, color)
   • Clear process steps
   • Key rules and output format
5. Validate using:

   scripts/validate-agent.sh <agent-file>

When Creating a Slash Command

1. Read the command template:

   templates/commands/command.md.template

2. Read the command example:

   templates/commands/command-example.md

3. Read documentation:

   docs/01-claude-code-slash-commands.md

4. Create command file with:
   • Frontmatter (description, argument-hint, allowed-tools)
   • Task invocation with proper subagent_type
   • Success criteria and notes
5. Validate using:

   scripts/validate-command.sh <command-file>

When Creating a Skill

1. Read the skill template:

   templates/skills/SKILL.md.template

2. Read the skill example:

   templates/skills/skill-example/SKILL.md

3. Read documentation:

   docs/02-claude-code-skills.md

4. Read decision guide:

   docs/04-skills-vs-commands.md

5. Create SKILL.md with:
   • Frontmatter with "Use when" trigger context
   • Step-by-step instructions
   • Concrete examples
   • Requirements
6. Validate using:

   scripts/validate-skill.sh <skill-directory>

When Creating a Plugin

1. Read the plugin template:

   templates/plugins/plugin.json.template

2. Read the plugin example:

   templates/plugins/example-plugin/

3. Read documentation:

   docs/03-claude-code-plugins.md

4. Create plugin structure with:

   • .claude-plugin/plugin.json

     manifest
   • README.md with components list
   • commands/, skills/, agents/ as needed
5. Validate using:

   scripts/validate-plugin.sh <plugin-directory>

Choosing Between Skills and Commands

Consult

docs/04-skills-vs-commands.md

to decide:

• Use Skill when: Claude should discover it automatically, complex capability, multiple supporting files
• Use Command when: User explicitly triggers it, simple orchestration, workflow shortcut

Available Resources

Templates

Agents:

• templates/agents/agent.md.template

  - Standard agent template with frontmatter

• templates/agents/agent-example.md

  - Complete working example

Commands:

• templates/commands/command.md.template

  - Standard slash command template

• templates/commands/command-example.md

  - Complete working example

Skills:

• templates/skills/SKILL.md.template

  - Standard skill template

• templates/skills/skill-example/SKILL.md

  - Git commit helper example

• templates/skills/README.md

  - Skills template documentation

Plugins:

• templates/plugins/plugin.json.template

  - Plugin manifest template

• templates/plugins/example-plugin/

  - Complete plugin example with commands and skills

Validation Scripts

• scripts/validate-agent.sh

  - Validates agent frontmatter and structure

• scripts/validate-command.sh

  - Validates command frontmatter and structure

• scripts/validate-skill.sh

  - Validates SKILL.md frontmatter and "Use when" context

• scripts/validate-plugin.sh

  - Validates plugin manifest and structure

• scripts/test-build-system.sh

  - Comprehensive build system test suite

Documentation

• docs/01-claude-code-slash-commands.md

  - Slash command reference

• docs/02-claude-code-skills.md

  - Skills reference with frontmatter fields

• docs/03-claude-code-plugins.md

  - Plugin architecture and structure

• docs/04-skills-vs-commands.md

  - Decision guide for choosing component type

Requirements

• Templates must exist in

  templates/

  directory
• Validation scripts must be executable
• Documentation files should be available in

  docs/

• Follow Claude Code standards for frontmatter
• Include "Use when" context in skill descriptions

Best Practices

1. Always validate - Run validation scripts after creation
2. Follow templates - Use provided templates as starting point
3. Read examples - Study working examples before creating new components
4. Check documentation - Consult docs for standards and patterns
5. Test thoroughly - Use test-build-system.sh for comprehensive testing

---

Generated from: multiagent-build plugin build-assistant skill Purpose: Standardize framework component creation across multiagent ecosystem Load when: Creating agents, commands, skills, or plugins