From 8e691aca2bc956b4c8567e4a79fba7349ae6d2e0 Mon Sep 17 00:00:00 2001
From: Anton Gulin <i@anton.qa>
Date: Tue, 16 Jun 2026 23:52:38 -0700
Subject: [PATCH] fix(test): isolate run-loop mocks from run-eval suite

Move run-loop module mocks into the test body and run Bun tests with
--isolate so mock.module cannot leak across files and break run-eval
imports during publish prepack.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 plugin/package.json          |   2 +-
 plugin/test/run-loop.test.ts | 105 +++++++++++++++++------------------
 2 files changed, 52 insertions(+), 55 deletions(-)

diff --git a/plugin/package.json b/plugin/package.json
index 635ea1c..a315859 100644
--- a/plugin/package.json
+++ b/plugin/package.json
@@ -10,7 +10,7 @@
   "scripts": {
     "build": "node ./scripts/build.mjs",
     "test": "node --test test/*.test.mjs",
-    "test:ts": "bun test test/*.test.ts",
+    "test:ts": "bun test --isolate test/*.test.ts",
     "prepack": "npm run build && npm test && npm run test:ts"
   },
   "dependencies": {
diff --git a/plugin/test/run-loop.test.ts b/plugin/test/run-loop.test.ts
index 0245316..2fbf76d 100644
--- a/plugin/test/run-loop.test.ts
+++ b/plugin/test/run-loop.test.ts
@@ -1,9 +1,7 @@
-import { afterEach, expect, mock, test } from "bun:test"
+import { expect, mock, test } from "bun:test"
 
 import type { EvalItem, EvalOutput, EvalResultItem } from "../lib/run-eval"
 
-const calls: { evalResults: EvalOutput }[] = []
-
 const result = (
   query: string,
   overrides: Partial<EvalResultItem> = {},
@@ -19,59 +17,56 @@ const result = (
   ...overrides,
 })
 
-mock.module("../lib/utils", () => ({
-  parseSkillMd: () => ({
-    name: "warning-skill",
-    description: "original description",
-    fullContent: "skill content",
-  }),
-}))
+test("runLoop derives train warnings from train results and prints unique split warnings", async () => {
+  const calls: { evalResults: EvalOutput }[] = []
+
+  mock.module("../lib/utils", () => ({
+    parseSkillMd: () => ({
+      name: "warning-skill",
+      description: "original description",
+      fullContent: "skill content",
+    }),
+  }))
 
-mock.module("../lib/run-eval", () => ({
-  findProjectRoot: () => "/tmp/project",
-  buildEvalWarnings: (results: EvalResultItem[]) => {
-    const shouldTriggerResults = results.filter((r) => r.should_trigger)
-    if (shouldTriggerResults.length === 0) return []
-    return shouldTriggerResults.every((r) => r.triggers === 0 && r.errors === 0)
-      ? ["all-zero warning"]
-      : []
-  },
-  runEval: () => ({
-    skill_name: "warning-skill",
-    description: "original description",
-    results: [
-      result("train trigger"),
-      result("train negative", { should_trigger: false }),
-      result("test trigger"),
-    ],
-    warnings: [],
-    summary: {
-      passed: 1,
-      failed: 2,
-      total: 3,
-      run_errors: 0,
-      queries_with_errors: 0,
+  mock.module("../lib/run-eval", () => ({
+    findProjectRoot: () => "/tmp/project",
+    buildEvalWarnings: (results: EvalResultItem[]) => {
+      const shouldTriggerResults = results.filter((r) => r.should_trigger)
+      if (shouldTriggerResults.length === 0) return []
+      return shouldTriggerResults.every((r) => r.triggers === 0 && r.errors === 0)
+        ? ["all-zero warning"]
+        : []
     },
-  }),
-}))
+    runEval: () => ({
+      skill_name: "warning-skill",
+      description: "original description",
+      results: [
+        result("train trigger"),
+        result("train negative", { should_trigger: false }),
+        result("test trigger"),
+      ],
+      warnings: [],
+      summary: {
+        passed: 1,
+        failed: 2,
+        total: 3,
+        run_errors: 0,
+        queries_with_errors: 0,
+      },
+    }),
+  }))
 
-mock.module("../lib/improve-description", () => ({
-  improveDescription: (opts: { evalResults: EvalOutput }) => {
-    calls.push({ evalResults: opts.evalResults })
-    return "improved description"
-  },
-}))
+  mock.module("../lib/improve-description", () => ({
+    improveDescription: (opts: { evalResults: EvalOutput }) => {
+      calls.push({ evalResults: opts.evalResults })
+      return "improved description"
+    },
+  }))
 
-mock.module("../lib/report", () => ({
-  generateHtml: () => "<html></html>",
-}))
+  mock.module("../lib/report", () => ({
+    generateHtml: () => "<html></html>",
+  }))
 
-afterEach(() => {
-  calls.length = 0
-})
-
-test("runLoop derives train warnings from train results and prints unique split warnings", async () => {
-  const { runLoop } = await import("../lib/run-loop")
   const evalSet: EvalItem[] = [
     { query: "train trigger", should_trigger: true },
     { query: "train negative", should_trigger: false },
@@ -84,6 +79,7 @@ test("runLoop derives train warnings from train results and prints unique split
   }
 
   try {
+    const { runLoop } = await import("../lib/run-loop")
     await runLoop({
       evalSet,
       skillPath: "/tmp/skill/SKILL.md",
@@ -97,10 +93,11 @@ test("runLoop derives train warnings from train results and prints unique split
       agent: undefined,
       verbose: true,
     })
+
+    expect(calls[0]?.evalResults.warnings).toEqual(["all-zero warning"])
+    expect(errors.filter((line) => line === "Warning: all-zero warning")).toHaveLength(2)
   } finally {
     console.error = originalError
+    mock.restore()
   }
-
-  expect(calls[0]?.evalResults.warnings).toEqual(["all-zero warning"])
-  expect(errors.filter((line) => line === "Warning: all-zero warning")).toHaveLength(2)
 })