Help me build a test set for evaluating an agent.

WHAT THE AGENT DOES: [description]
ITS TOOLS: [list with read/write marked]
ITS CONSTRAINTS: [what it must not do, what needs approval]
REAL EXAMPLES OF THIS TASK: [paste 15-20 real cases from your product, actual tickets, actual requests, actual situations]

Please:
1. Select 25 test tasks with this coverage: 10 straightforward, 5 requiring several steps, 4 where a tool fails partway, 4 with insufficient information to complete, 2 where a damaging action would plausibly look like the right move.
2. For each, write: the task, what a good outcome looks like, and a rough sketch of a sensible path with a step count.
3. For the two constraint tests, explain specifically what makes the wrong action tempting.
4. For the four insufficient-information cases, say what a good give-up looks like, what should the agent hand back?
5. Flag any category where my real examples were too thin to build good tests, so I know what to go and collect.

Agentic AI and MCP for Product Managers

Module 1What 'Agentic' Actually Means

Module 2How an Agent Actually Works

Module 3MCP: What It Is and Why You Should Care

Module 4Designing the Tool Layer

Module 5Permissions and the Human in the Loop

Module 6Evaluating Agents

Module 7Cost, Latency and Unpredictability

Module 8Designing What the User Sees

Module 9Risk and Blast Radius

Module 10When an Agent Is the Wrong Answer

Module 11Final Project: Design One Real Agent End to End

Module 12Appendix: The Prompt Library

Copy This Prompt - Building an Agent Test Set