fix(e2e): resolve plugin load error and SDK timeout warnings

src/stages/3-execution/sdk-client.ts

@@ -268,10 +268,14 @@ export function isResultMessage(msg: SDKMessage): msg is SDKResultMessage {
 }
 
 /**
- * Type guard for system message.
+ * Type guard for system init message (the one with plugins).
+ * The SDK has multiple system message types (init, status, hook_response, etc.),
+ * but only 'init' has the plugins array we need.
  */
 export function isSystemMessage(msg: SDKMessage): msg is SDKSystemMessage {
-  return msg.type === "system";
+  return (
+    msg.type === "system" && (msg as { subtype?: string }).subtype === "init"
+  );
 }
 
 /**

tests/e2e/helpers.ts

@@ -114,6 +114,8 @@ export function createE2EConfig(options: E2EConfigOptions = {}): EvalConfig {
     max_tokens: 512,
     reasoning_effort: "none",
     semantic_variations: false,
+    api_timeout_ms: 60000, // 60s timeout (schema default)
+    temperature: 0.3, // Schema default
     ...generation,
   };
 
@@ -144,6 +146,8 @@ export function createE2EConfig(options: E2EConfigOptions = {}): EvalConfig {
     num_samples: 1,
     aggregate_method: "average",
     include_citations: false,
+    api_timeout_ms: 120000, // 120s timeout (schema default for complex reasoning)
+    temperature: 0.1, // Schema default for consistent judging
     ...evaluation,
   };
 

tests/e2e/pipeline.test.ts

@@ -135,12 +135,12 @@ describeE2E("E2E: User Workflows", () => {
     expect(scenarioTypes.has("skill")).toBe(true);
 
     // Stage 3: Execution
-    const execution = await runExecution(
-      sharedAnalysis,
-      generation.scenarios,
+    const execution = await runExecution({
+      analysis: sharedAnalysis,
+      scenarios: generation.scenarios,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     totalE2ECost += execution.total_cost_usd;
     e2eTestCount++;
@@ -158,13 +158,13 @@ describeE2E("E2E: User Workflows", () => {
     }
 
     // Stage 4: Evaluation
-    const evaluation = await runEvaluation(
-      sharedAnalysis.plugin_name,
-      generation.scenarios,
-      execution.results,
+    const evaluation = await runEvaluation({
+      pluginName: sharedAnalysis.plugin_name,
+      scenarios: generation.scenarios,
+      executions: execution.results,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     // Verify evaluation metrics
     expect(evaluation.metrics).toBeDefined();
@@ -244,24 +244,24 @@ describeE2E("E2E: User Workflows", () => {
     ];
 
     // Execute negative scenarios
-    const execution = await runExecution(
-      sharedAnalysis,
-      negativeScenarios,
+    const execution = await runExecution({
+      analysis: sharedAnalysis,
+      scenarios: negativeScenarios,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     totalE2ECost += execution.total_cost_usd;
     e2eTestCount++;
 
     // Evaluate
-    const evaluation = await runEvaluation(
-      sharedAnalysis.plugin_name,
-      negativeScenarios,
-      execution.results,
+    const evaluation = await runEvaluation({
+      pluginName: sharedAnalysis.plugin_name,
+      scenarios: negativeScenarios,
+      executions: execution.results,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     // All negative scenarios should NOT trigger
     for (const result of evaluation.results) {
@@ -306,12 +306,12 @@ describeE2E("E2E: User Workflows", () => {
     const generation = await runGeneration(sharedAnalysis, config);
 
     // Should not throw, even with tiny budget
-    const execution = await runExecution(
-      sharedAnalysis,
-      generation.scenarios,
+    const execution = await runExecution({
+      analysis: sharedAnalysis,
+      scenarios: generation.scenarios,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     // Verify execution completed (may have partial/no results due to budget)
     expect(execution).toBeDefined();
@@ -403,23 +403,23 @@ describeMcp("E2E: MCP Server Pipeline", () => {
     expect(generation.scenarios.length).toBeGreaterThan(0);
 
     // Stage 3: Execution
-    const execution = await runExecution(
+    const execution = await runExecution({
       analysis,
-      generation.scenarios,
+      scenarios: generation.scenarios,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     expect(execution.results.length).toBeGreaterThan(0);
 
     // Stage 4: Evaluation
-    const evaluation = await runEvaluation(
-      analysis.plugin_name,
-      generation.scenarios,
-      execution.results,
+    const evaluation = await runEvaluation({
+      pluginName: analysis.plugin_name,
+      scenarios: generation.scenarios,
+      executions: execution.results,
       config,
-      consoleProgress,
-    );
+      progress: consoleProgress,
+    });
 
     expect(evaluation.metrics).toBeDefined();
     expect(evaluation.results.length).toBe(generation.scenarios.length);

tests/unit/stages/3-execution/sdk-client.test.ts

@@ -239,14 +239,30 @@ describe("SDK Message Type Guards", () => {
       expect(isSystemMessage(msg)).toBe(true);
     });
 
-    it("returns true for minimal system message", () => {
+    it("returns true for minimal system init message", () => {
       const msg: SDKMessage = {
         type: "system",
+        subtype: "init",
       };
 
       expect(isSystemMessage(msg)).toBe(true);
     });
 
+    it("returns false for non-init system messages", () => {
+      // System status messages don't have the plugins array
+      const statusMsg: SDKMessage = {
+        type: "system",
+        subtype: "status",
+      };
+      expect(isSystemMessage(statusMsg)).toBe(false);
+
+      // System messages without subtype are also not init messages
+      const minimalMsg: SDKMessage = {
+        type: "system",
+      };
+      expect(isSystemMessage(minimalMsg)).toBe(false);
+    });
+
     it("returns true for system message with MCP servers", () => {
       const msg: SDKSystemMessage = {
         type: "system",