In this module, you will put your Module 1 test set to work. You will test an agent against its happy path first, then design edge cases, adversarial inputs, and simulated tool failures or incomplete data. You will also learn why repeatability matters: an agent that passes once but fails on a repeat run isn't reliable. In a role play, you will explain a repeatability problem to an operations lead. In the module's hands-on lab, you will run a structured test cycle on your existing agent (or the supplied sample agent), recording inputs, outputs, and pass/fail results as evidence. By the end of this module, you will have documented test results that become the baseline for your capstone.

In this module, you will focus on the risks that make an otherwise functional agent unsafe to rely on. You will apply least-privilege permission and data-privacy practices, recognize what prompt injection looks like, and learn to spot sensitive-data exposure before it becomes a problem. In a scenario-based activity, you right-size an agent's access, and in a role play, you probe a workflow for prompt injection and over-automation risk. You will also learn to design human-in-the-loop approval points, deciding which actions run automatically, where approval belongs, and what the agent does when no one responds. By the end of this module, you will be able to evaluate a scenario for safety and security risk and recommend a concrete mitigation.

In this module, you will turn your Module 2 test results into action. You will learn to read logs and traces to debug a failure back to its root cause, practicing in a guided dialogue, then refine instructions and workflow logic to fix what's broken. In the module's hands-on lab, you will implement at least two improvements to your Course 2 agent (or the supplied sample agent) and rerun the relevant tests to confirm they worked, comparing before-and-after results. The module closes with what to monitor after release and how to manage changes responsibly. By the end of this module, you will have documented, tested evidence of improvement to carry into your capstone.

In this final module, you will consolidate everything from this course into one Capstone Deployment Brief for your Course 2 agent (or the supplied sample agent). Part 1 consolidates your evaluation criteria, test cases, acceptance thresholds, baseline results, repeatability evidence, and identified risks from Modules 1 and 2, plus a workflow diagram. Part 2 consolidates the improvements, retest evidence, guardrails, data and permission boundaries, and human-approval points developed in Modules 3 and 4 and uses it to produce a deployment-readiness recommendation covering monitoring and change management. Parts 1 and 2 are submitted together. The module then closes with a course summary, glossary, a self-review checklist, and a wrap-up marking the end of the AI Agents for Dummies specialization.