Fix test runner to use shell execution for cross-platform compatibility (#19)

Switch from execFile to exec (shell mode) so npx, npm, and other
platform-specific commands resolve correctly on both Windows and Unix.
Remove ENOENT-specific handling (shell handles this internally).

Co-authored-by: unknown <that-github-user@github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: 3 people

src/scoring/test-runner.test.ts

@@ -36,10 +36,9 @@ describe("runTests", () => {
   });
 
   it("returns failure for non-existent command", async () => {
-    const result = await runTests(1, "nonexistent-command-xyz", "/tmp");
+    const result = await runTests(1, "nonexistent-command-xyz", ".");
     assert.equal(result.passed, false);
-    assert.equal(result.exitCode, 127);
-    assert.ok(result.output.includes("Command not found"));
+    assert.ok(result.exitCode !== 0);
   });
 
   it("returns success for passing command", async () => {

src/scoring/test-runner.ts

@@ -1,10 +1,10 @@
-import { execFile } from "node:child_process";
+import { exec as execCb } from "node:child_process";
 import { access } from "node:fs/promises";
 import { join } from "node:path";
 import { promisify } from "node:util";
 import type { TestResult } from "../types.js";
 
-const exec = promisify(execFile);
+const exec = promisify(execCb);
 
 const TEST_TIMEOUT_MS = 120_000;
 
@@ -62,7 +62,8 @@ export async function runTests(
   }
 
   try {
-    const { stdout, stderr } = await exec(cmd, args, {
+    // Use shell execution so npx, npm, etc. resolve correctly on all platforms
+    const { stdout, stderr } = await exec(testCmd, {
       cwd: worktreePath,
       timeout: TEST_TIMEOUT_MS,
       env: { ...process.env, CI: "true" },
@@ -92,16 +93,6 @@ export async function runTests(
       };
     }
 
-    // Command not found
-    if (typeof e.code === "string" && e.code === "ENOENT") {
-      return {
-        agentId,
-        passed: false,
-        output: `Command not found: ${cmd}. Is it installed?`,
-        exitCode: 127,
-      };
-    }
-
     return {
       agentId,
       passed: false,