@@ -158,7 +158,7 @@ The session mode controls how independently the agent works. The model and reaso
 
 The model control is below the prompt box, next to the mode selector. It shows **Auto** or a model name. Open it to find **Model**, **Effort**, and **Context window**. If the **Auto** switch is on, these settings are hidden. Turn it off to see them. Some app versions show separate model and reasoning controls.
 
-With **Auto**, the Copilot app chooses a model for your task. This is different from **Autopilot**, which controls how independently the agent works.
+With **Auto**, the Copilot app chooses a model for your task and shows which model it selected in the conversation, alongside a short explanation of why it made that choice. This is different from **Autopilot**, which controls how independently the agent works.
 
 For a repository tour or a short explanation, keep **Auto** or the current model and its default reasoning effort. For a difficult bug or a change across several files, consider a model suited to complex coding and a higher effort level. Higher effort does not guarantee a correct answer.
 
 
@@ -152,6 +152,9 @@ In Chapter 02, you attached an issue to practice giving Copilot focused context.
 
 1. Open the **Changes** tab and inspect the diff. Confirm that the change is focused on the filtered-book data flow. If it isn't, ask Copilot to revise it before continuing.
 
+   > [!TIP]
+   > If the agent changed a file you didn't intend, you can discard that file's changes directly from the diff view. Look for the discard action in the **Changes** tab. A confirmation dialog appears before the file is restored, so you can review the decision first.
+
 #### Validate the Fix
 
 1. Review the output in the session's **Terminal** tab. If Copilot already ran these commands after the fix, you don't need to run them again. Otherwise, run:
 
@@ -227,6 +227,9 @@ The agent profile is a Markdown file with YAML frontmatter, like a skill:
 
 This agent cannot run the tests or save its own report. That restriction is intentional: match an agent's tools to its task, and add nothing beyond that. Use the default agent when you need those actions.
 
+> [!TIP]
+> After a custom agent finishes its response, it may suggest next steps. These appear as a pre-filled prompt you can review and edit before sending. You can use a suggestion to switch smoothly to another agent or continue the conversation without retyping your intent.
+
 ### Checkpoint
 
 After these exercises, you should have: