/add-agent

Arguments: $ARGUMENTS

Goal: Add new agents (LLM-based, custom tool, or workflow) to a Google ADK project with proper configuration and integration.

Core Principles:

• Detect project structure before assuming configuration
• Support all three agent types (LLM, custom, workflow)
• Follow Google ADK best practices and patterns
• Validate agent configuration before finalizing

Phase 1: Discovery Goal: Understand the target project and agent requirements

Actions:

• Parse $ARGUMENTS for agent type and name
• If $ARGUMENTS is unclear or missing, use AskUserQuestion to gather:
  • What type of agent? (llm/custom/workflow)
  • What is the agent name?
  • What should the agent do?
  • Any specific tools or capabilities needed?
• Verify this is a Google ADK project:
  • Check for agent configuration patterns
  • Example: !{bash ls -la agents/ app/ 2>/dev/null || echo "Not found"}
• Load existing agent configurations for context
• Example: @agents/

Phase 2: Analysis Goal: Understand existing agent patterns and project structure

Actions:

• Find existing agent implementations:
  • Example: !{bash find agents/ -name ".py" -o -name ".ts" 2>/dev/null | head -5}
• Read sample agent files to understand patterns
• Identify agent registry location
• Determine required dependencies and imports
• Check for agent configuration files

Phase 3: Planning Goal: Design the agent implementation approach

Actions:

• Based on agent type (llm/custom/workflow), plan:
  • Required files and structure
  • Dependencies and imports
  • Configuration format
  • Integration points with existing agents
• Present plan to user:
  • Agent structure to be created
  • Files that will be modified
  • Configuration approach
  • Any trade-offs or considerations

Phase 4: Implementation Goal: Create the agent with proper configuration

Actions:

Task(description="Add Google ADK agent", subagent_type="google-adk-agent-builder", prompt="You are the google-adk-agent-builder agent. Add a new agent to this Google ADK project for $ARGUMENTS.

Context from discovery:

• Agent type and name parsed from arguments
• Existing project structure analyzed
• Agent patterns identified

Requirements:

• Create agent implementation file
• Configure agent according to type (LLM/custom/workflow)
• Add proper imports and dependencies
• Register agent in agent registry
• Follow Google ADK best practices
• Include appropriate error handling
• Add documentation and comments

Agent Types:

• LLM: Uses language model for reasoning and responses
• Custom: Uses custom tools and functions
• Workflow: Orchestrates multiple steps or sub-agents

Expected output:

• Agent implementation file created
• Agent registered and configured
• Documentation included
• Summary of capabilities and usage")

Phase 5: Verification Goal: Ensure the agent is properly configured

Actions:

• Verify agent file was created
• Check agent registration
• Validate configuration syntax if applicable
• Example: !{bash python -m py_compile agents/*.py 2>&1 || true}
• Test agent can be imported/loaded
• Confirm no syntax errors

Phase 6: Summary Goal: Document what was accomplished

Actions:

• Summarize agent creation:
  • Agent type and name
  • Files created or modified
  • Key capabilities added
  • How to use the agent
• Provide next steps:
  • Testing the agent
  • Configuring additional tools if needed
  • Integration with other agents
  • Deployment considerations