Helpful Bob Prompts for watsonx Orchestrate Development¶

This guide contains useful prompts to help you work with Bob throughout the workshop and beyond.

Workshop Task Pattern¶

In the workshop, Bob is expected to do the work with you—not only explain it. Adapt this compact pattern for each artifact:

Inspect the relevant project files and custom rules. Consult the watsonx
Orchestrate documentation MCP for ADK 2.14.0 syntax. Create or update the
requested artifacts, show me what changed, and validate them. After I review
them, import them into the active draft environment using the existing .venv
and verify they are listed. Suggest a small test set with normal, edge, and
failure cases, then start the appropriate chat or test command.

Review generated files and commands before approving them. Never paste API keys or other secrets into Bob chat; enter them only in dedicated credential prompts.

General Agent Development¶

Understanding Concepts¶

Bob, explain what watsonx Orchestrate agents are and how they differ from regular chatbots

Bob, what are the key components of a watsonx Orchestrate agent?

Bob, explain the difference between native, external, and assistant agents

Getting Started¶

Bob, help me set up my watsonx Orchestrate development environment

Bob, show me how to verify my watsonx Orchestrate connection is working

Bob, what's the basic structure of an agent YAML file?

Creating Agents¶

Basic Agent Creation¶

Bob, create a YAML file for a customer service agent that can help with orders and refunds

Bob, create an agent that acts as a Python programming tutor

Bob, build an agent specification for a travel booking assistant

Improving Agent Instructions¶

Bob, review these agent instructions and suggest improvements: [paste instructions]

Bob, make these agent instructions more specific and actionable: [paste instructions]

Bob, add error handling guidance to these agent instructions: [paste instructions]

Agent Configuration¶

Bob, explain when to use hide_reasoning in an ADK 2.14.0 native-agent YAML

Bob, help me choose the right LLM model for my agent

Bob, what's the difference between agent styles (default, react, planner)?

Building Tools¶

Tool Creation¶

Bob, create a Python tool that checks order status given an order ID

Bob, build a tool that validates email addresses and returns detailed feedback

Bob, create a tool that calls the OpenWeather API to get current weather

Tool Improvement¶

Bob, add input validation to this tool: [paste tool code]

Bob, improve the error handling in this tool: [paste tool code]

Bob, add type hints and better documentation to this tool: [paste tool code]

Tool Debugging¶

Bob, why isn't my agent calling this tool? Here's the agent YAML and tool code: [paste both]

Bob, this tool is returning an error. Can you help debug it? [paste error and code]

Bob, review this tool for potential issues: [paste tool code]

Working with Knowledge Bases¶

Knowledge Base Setup¶

Bob, create a knowledge base YAML file for FAQ documents

Bob, help me structure my knowledge base for best retrieval performance

Bob, explain how to connect a knowledge base to my agent

Knowledge Base Optimization¶

Bob, how should I format my documents for optimal knowledge base performance?

Bob, what's the best way to organize multiple knowledge bases?

Agent Collaboration¶

Creating Collaborator Agents¶

Bob, create a specialized escalation agent that handles complex customer issues

Bob, build a technical support agent that can collaborate with a billing agent

Bob, design an agent hierarchy for a customer service system

Routing and Delegation¶

Bob, improve the routing logic in these agent instructions: [paste instructions]

Bob, help me decide when my agent should delegate to a collaborator

Testing and Debugging¶

Test Case Generation¶

Bob, generate test cases for this customer support agent

Bob, create a test script that validates my order status tool

Bob, write test scenarios that cover edge cases for my refund tool

Debugging¶

Bob, my agent isn't responding correctly. Here's the conversation: [paste conversation]

Bob, analyze this error message and suggest a fix: [paste error]

Bob, why is my agent calling the wrong tool? [paste agent YAML and conversation]

Performance Optimization¶

Bob, how can I make my agent respond faster?

Bob, review my agent's instructions for efficiency: [paste instructions]

Bob, suggest ways to reduce token usage in my agent

Integration and Deployment¶

API Integration¶

Bob, create a tool that calls this REST API: [paste API documentation]

Bob, add authentication to this API tool: [paste tool code]

Bob, help me handle rate limiting in this API tool: [paste tool code]

Deployment¶

Bob, create a deployment checklist for my agent

Bob, help me test my agent before deploying to production

Bob, generate webchat embed code for my agent

Code Review and Refactoring¶

Code Review¶

Bob, review this agent YAML for best practices: [paste YAML]

Bob, check this tool for security issues: [paste tool code]

Bob, review my entire agent setup and suggest improvements: [paste all files]

Refactoring¶

Bob, refactor this tool to be more modular: [paste tool code]

Bob, split this complex agent into multiple specialized agents: [paste agent YAML]

Bob, improve the code organization in my tool: [paste tool code]

Documentation¶

Generating Documentation¶

Bob, create documentation for this tool: [paste tool code]

Bob, write a README for my agent project

Bob, generate API documentation for my custom tools

User Guides¶

Bob, create a user guide for interacting with my customer service agent

Bob, write instructions for deploying this agent

Advanced Topics¶

Custom Workflows¶

Bob, design a multi-step workflow using agent collaborators

Bob, create a planner-style agent with structured output

Bob, create a planner-style agent and keep internal reasoning hidden from users

MCP Servers¶

Bob, help me create an MCP server for my custom tools

Bob, explain how to connect an MCP server to watsonx Orchestrate

Model Configuration¶

Bob, create a custom model configuration for Azure OpenAI

Bob, set up a model policy with fallback models

Troubleshooting Common Issues¶

Connection Issues¶

Bob, I'm getting a connection refused error. How do I fix it?

Bob, my API key isn't working. What should I check?

Agent Not Responding¶

Bob, my agent isn't responding to user messages. Help me debug this.

Bob, the agent is giving generic responses instead of using tools. Why?

Tool Issues¶

Bob, my tool isn't being imported. What could be wrong?

Bob, the agent sees my tool but never calls it. How do I fix this?

Performance Issues¶

Bob, my agent is slow. How can I improve response time?

Bob, I'm hitting rate limits. What's the best approach?

Best Practices¶

Ask Bob for Best Practices¶

Bob, what are the best practices for writing agent instructions?

Bob, how should I structure my tools for maintainability?

Bob, what's the recommended way to handle errors in agents?

Bob, how do I make my agents more reliable in production?

Learning and Exploration¶

Exploring Features¶

Bob, what are some advanced features of watsonx Orchestrate I should know about?

Bob, show me examples of creative ways to use agent collaborators

Bob, what are some real-world use cases for watsonx Orchestrate agents?

Staying Updated¶

Bob, what's new in the latest version of watsonx Orchestrate?

Bob, are there any deprecated features I should be aware of?

Tips for Effective Bob Prompts¶

✅ DO:¶

Be specific: Include relevant code, error messages, or context
State your goal: Explain what you're trying to achieve
Provide examples: Show what you've tried or what you want
Ask follow-ups: Build on Bob's responses with clarifying questions
Request explanations: Ask Bob to explain the reasoning behind suggestions

❌ DON'T:¶

Be vague: "Bob, fix this" without context
Skip context: Not providing relevant code or configuration
Expect mind-reading: Bob needs information about your specific situation
Ignore suggestions: Bob's recommendations are based on best practices
Ask multiple unrelated questions: Focus on one topic at a time

Example Conversation Flow¶

Here's an example of an effective conversation with Bob:

You: Bob, I need to create a customer support agent that can check order status and process refunds.

Bob: [Provides initial agent structure]

You: Thanks! Can you also add a tool that checks order status? It should take an order ID and return the status, items, and delivery date.

Bob: [Creates order status tool]

You: Great! Now I'm getting an error when I import the tool. Here's the error: [paste error]

Bob: [Analyzes error and provides fix]

You: Perfect! Can you now update the agent YAML to use this tool?

Bob: [Updates agent configuration]

You: Excellent! Can you generate some test cases for this agent?

Bob: [Provides test scenarios]

Quick Reference¶

Most Common Prompts¶

Create: "Bob, create a [type] that does [function]"
Debug: "Bob, why is [thing] not working? Here's the [code/error]"
Improve: "Bob, improve [code/config] by [specific improvement]"
Explain: "Bob, explain [concept] in the context of [use case]"
Review: "Bob, review [code/config] for [issues/best practices]"

💡 Pro Tip: The more context you provide to Bob, the better the responses will be. Include relevant code, error messages, and explain what you're trying to achieve!

🎯 Remember: Bob is your AI pair programmer. Use Bob to learn, not just to get answers. Ask Bob to explain the reasoning behind suggestions!