test(coding-agent): activated eval in skill workflow fixture

Keep the skill keyword workflowz contract on the active eval-and-task path.

packages/coding-agent/test/agent-session-skill-keywords.test.ts

@@ -22,7 +22,7 @@ type ObservedSkillTurn = {
 	texts: string[];
 };
 
-// 4644 gates the workflowz notice on an active `task` tool; keep one active so
+// Workflowz requires active `task` and `eval` tools; keep both active so
 // keyword steering exercises the notice path.
 const mockTaskTool: AgentTool = {
 	name: "task",
@@ -32,6 +32,14 @@ const mockTaskTool: AgentTool = {
 	execute: async () => ({ content: [{ type: "text" as const, text: "ok" }] }),
 };
 
+const mockEvalTool: AgentTool = {
+	name: "eval",
+	label: "Eval",
+	description: "Mock eval tool",
+	parameters: type({}),
+	execute: async () => ({ content: [{ type: "text" as const, text: "ok" }] }),
+};
+
 describe("AgentSession skill prompt keyword steering", () => {
 	let tempDir: TempDir;
 	let authStorage: AuthStorage | undefined;
@@ -53,7 +61,7 @@ describe("AgentSession skill prompt keyword steering", () => {
 			initialState: {
 				model,
 				systemPrompt: ["Test"],
-				tools: [mockTaskTool],
+				tools: [mockTaskTool, mockEvalTool],
 				messages: [],
 			},
 			convertToLlm,