gsd-pi 2.82.0-dev.c22380fc3 → 2.82.0-dev.e7a7f1ed5

package/src/resources/extensions/gsd/tests/post-execution-checks.test.ts

@@ -127,6 +127,61 @@ import { b } from './b';
     assert.equal(imports.length, 2);
   });
 
+  test("ignores import-looking string literals in test fixtures", () => {
+    const source = `
+const rewritten = source.replace(
+  'import { normalizeZagrebBusinessDeadline } from "./cutoff";',
+  'const helper = true;'
+);
+
+import realThing from "./real-thing";
+`;
+    const imports = extractRelativeImports(source);
+    assert.deepEqual(imports, [
+      { importPath: "./real-thing", lineNum: 7 },
+    ]);
+  });
+
+  test("ignores import-looking lines inside template literals", () => {
+    const source = [
+      "const fixture = `",
+      "import missingThing from './missing-thing';",
+      "`;",
+      "",
+      "import realThing from './real-thing';",
+    ].join("\n");
+    const imports = extractRelativeImports(source);
+    assert.deepEqual(imports, [
+      { importPath: "./real-thing", lineNum: 5 },
+    ]);
+  });
+
+  test("ignores require() inside string literals", () => {
+    const source = [
+      'const fixture = "const x = require(\'./missing\');";',
+      "const otherFixture = 'const y = require(\"./also-missing\");';",
+      "const real = require('./real');",
+    ].join("\n");
+    const imports = extractRelativeImports(source);
+    assert.deepEqual(imports, [
+      { importPath: "./real", lineNum: 3 },
+    ]);
+  });
+
+  test("ignores require() inside template literals", () => {
+    const source = [
+      "const fixture = `",
+      "const x = require('./missing');",
+      "`;",
+      "",
+      "const real = require('./real');",
+    ].join("\n");
+    const imports = extractRelativeImports(source);
+    assert.deepEqual(imports, [
+      { importPath: "./real", lineNum: 5 },
+    ]);
+  });
+
   test("handles empty source", () => {
     const imports = extractRelativeImports("");
     assert.deepEqual(imports, []);
@@ -810,6 +865,37 @@ describe("runPostExecutionChecks", () => {
     }
   });
 
+  test("does not fail on import-looking strings in task key files", () => {
+    tempDir = join(tmpdir(), `post-exec-test-${Date.now()}`);
+    mkdirSync(tempDir, { recursive: true });
+    mkdirSync(join(tempDir, "tests"), { recursive: true });
+    writeFileSync(join(tempDir, "tests", "real-thing.ts"), "export default true;");
+    writeFileSync(
+      join(tempDir, "tests", "source-verifier.test.ts"),
+      `
+const rewritten = source.replace(
+  'import { normalizeZagrebBusinessDeadline } from "./cutoff";',
+  'const helper = true;'
+);
+
+import realThing from "./real-thing";
+assert.ok(realThing);
+`
+    );
+
+    try {
+      const task = createTask({
+        id: "T03",
+        key_files: ["tests/source-verifier.test.ts"],
+      });
+      const result = runPostExecutionChecks(task, [], tempDir);
+      assert.equal(result.status, "pass");
+      assert.deepEqual(result.checks, []);
+    } finally {
+      rmSync(tempDir, { recursive: true, force: true });
+    }
+  });
+
   test("returns fail status when blocking failure exists", () => {
     tempDir = join(tmpdir(), `post-exec-test-${Date.now()}`);
     mkdirSync(tempDir, { recursive: true });

package/src/resources/extensions/gsd/tests/post-unit-git-failure.test.ts

@@ -11,7 +11,7 @@ const source = readFileSync(
 
 test("postUnitPreVerification blocks on git action failure", () => {
   const failureBlock = extractSourceRegion(source, 'if (gitResult.status === "failed")');
-  assert.ok(failureBlock.includes('ctx.ui.notify(failureMsg, "error")'));
+  assert.ok(failureBlock.includes('ctx.ui.notify(failureMsg, opts?.softFailure ? "warning" : "error")'));
   assert.ok(failureBlock.includes("await pauseAuto(ctx, pi)"));
   assert.ok(failureBlock.includes('return "dispatched"'));
   assert.ok(!failureBlock.includes("git-action-failed-nonblocking"));

package/src/resources/extensions/gsd/tests/pre-execution-checks.test.ts

@@ -9,6 +9,7 @@
  *   2. File path consistency — files exist vs prior expected_output
  *   3. Task ordering — detect impossible read-before-create
  *   4. Interface contracts — contradictory function signatures
+ *   5. Verify commands — reject unsafe or non-runnable task verification
  */
 
 import { describe, test, mock } from "node:test";
@@ -22,6 +23,7 @@ import {
   checkFilePathConsistency,
   checkTaskOrdering,
   checkInterfaceContracts,
+  checkVerificationCommands,
   runPreExecutionChecks,
   normalizeFilePath,
   type PreExecutionResult,
@@ -812,6 +814,33 @@ function process(a: number): number
   });
 });
 
+describe("checkVerificationCommands", () => {
+  test("accepts pipe-free pytest Verify command", () => {
+    const results = checkVerificationCommands([
+      createTask({
+        id: "T01",
+        verify: "python3 -m pytest tests/ -q --tb=short",
+      }),
+    ]);
+
+    assert.deepEqual(results, []);
+  });
+
+  test("rejects piped pytest Verify command", () => {
+    const results = checkVerificationCommands([
+      createTask({
+        id: "T01",
+        verify: "python3 -m pytest tests/ -q --tb=short 2>&1 | tail -5",
+      }),
+    ]);
+
+    assert.equal(results.length, 1);
+    assert.equal(results[0]?.category, "tool");
+    assert.equal(results[0]?.blocking, true);
+    assert.match(results[0]?.message ?? "", /shell control syntax/);
+  });
+});
+
 // ─── runPreExecutionChecks Integration Tests ─────────────────────────────────
 
 describe("runPreExecutionChecks", () => {
@@ -847,6 +876,30 @@ describe("runPreExecutionChecks", () => {
     }
   });
 
+  test("returns fail status for unsafe Verify command before execution", async () => {
+    tempDir = join(tmpdir(), `pre-exec-test-${Date.now()}`);
+    mkdirSync(tempDir, { recursive: true });
+
+    try {
+      const tasks = [
+        createTask({
+          id: "T01",
+          verify: "python3 -m pytest tests/ -q --tb=short 2>&1 | tail -5",
+        }),
+      ];
+
+      const result = await runPreExecutionChecks(tasks, tempDir);
+
+      assert.equal(result.status, "fail");
+      assert.equal(result.checks.length, 1);
+      assert.equal(result.checks[0]?.category, "tool");
+      assert.equal(result.checks[0]?.blocking, true);
+      assert.match(result.checks[0]?.message ?? "", /Unsafe or non-runnable Verify command/);
+    } finally {
+      rmSync(tempDir, { recursive: true, force: true });
+    }
+  });
+
   test("returns fail status when blocking failure exists", async () => {
     tempDir = join(tmpdir(), `pre-exec-test-${Date.now()}`);
     mkdirSync(tempDir, { recursive: true });

package/src/resources/extensions/gsd/tests/prefs-wizard-coverage.test.ts

@@ -141,3 +141,62 @@ test("category summaries expose the wizard menu surface for configured prefs", (
   assert.match(summaries.integrations, /remote: C123/);
   assert.match(summaries.verification, /1 cmd/);
 });
+
+test("models wizard offers discovered models for enabled providers", async () => {
+  const dir = mkdtempSync(join(tmpdir(), "gsd-prefs-wizard-"));
+  const prefsPath = join(dir, "PREFERENCES.md");
+  const choices = [
+    "Models",
+    "local (2 models)",
+    "discovered-model",
+    "(keep current)",
+    "(keep current)",
+    "(keep current)",
+    "(keep current)",
+    "(keep current)",
+    "(keep current)",
+    "(keep current)",
+  ];
+  const ctx = {
+    modelRegistry: {
+      getAvailable: () => [{ provider: "local", id: "baseline-model" }],
+      getAllWithDiscovered: () => [
+        { provider: "local", id: "baseline-model" },
+        { provider: "local", id: "discovered-model" },
+        { provider: "disabled", id: "hidden-model" },
+      ],
+    },
+    ui: {
+      notify() {},
+      select: async (label: string, options: string[]) => {
+        const choice = choices.shift();
+        if (!choice && label === "GSD Preferences") return "── Save & Exit ──";
+        if (!choice && options.includes("(keep current)")) return "(keep current)";
+        if (!choice && options.includes("Done")) return "Done";
+        assert.ok(choice, `Unexpected prompt: ${label}`);
+        if (choice === "Models") {
+          const modelsOption = options.find((option) => option.startsWith("Models"));
+          assert.ok(modelsOption, "Expected Models category option");
+          return modelsOption;
+        }
+        assert.ok(options.includes(choice), `"${choice}" must be offered by "${label}"`);
+        assert.ok(!options.includes("hidden-model"), "models from disabled providers must not be offered");
+        return choice;
+      },
+      input: async () => null,
+    },
+    waitForIdle: async () => {},
+    reload: async () => {},
+  } as any;
+
+  try {
+    await handlePrefsWizard(ctx, "project", {}, { pathOverride: prefsPath });
+
+    assert.equal(choices.length, 0, "Expected all queued wizard choices to be consumed");
+    const saved = readFileSync(prefsPath, "utf-8");
+    assert.match(saved, /research:\s+local\/discovered-model/);
+    assert.doesNotMatch(saved, /hidden-model/);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});

package/src/resources/extensions/gsd/tests/prompt-loader.test.ts

@@ -0,0 +1,23 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { loadPrompt } from "../prompt-loader.ts";
+
+test("loadPrompt reports missing template variables with balanced braces", () => {
+  assert.throws(
+    () => loadPrompt("guided-discuss-milestone", {
+      milestoneId: "M001",
+      milestoneTitle: "Missing working directory",
+      structuredQuestionsAvailable: "false",
+      fastPathInstruction: "",
+      inlinedTemplates: "context template",
+      commitInstruction: "Do not commit during this test.",
+    }),
+    (error) => {
+      assert.ok(error instanceof Error);
+      assert.match(error.message, /template declares \{\{workingDirectory\}\} but no value was provided/);
+      assert.doesNotMatch(error.message, /\{\{workingDirectory\}\}\}/);
+      return true;
+    },
+  );
+});

package/src/resources/extensions/gsd/tests/provider-errors.test.ts

@@ -10,7 +10,12 @@ import assert from "node:assert/strict";
 import { classifyError, isTransient, isTransientNetworkError } from "../error-classifier.ts";
 import { pauseAutoForProviderError } from "../provider-error-pause.ts";
 import { resumeAutoAfterProviderDelay } from "../bootstrap/provider-error-resume.ts";
-import { MAX_TRANSIENT_AUTO_RESUMES, resetTransientRetryState } from "../bootstrap/agent-end-recovery.ts";
+import {
+  MAX_TRANSIENT_AUTO_RESUMES,
+  isTerminalDeletedWorktreeProviderError,
+  resetTransientRetryState,
+  shouldDeferTransientErrorToCoreRetry,
+} from "../bootstrap/agent-end-recovery.ts";
 import { _buildCancelledUnitStopReason } from "../auto/phases.ts";
 import { getNextFallbackModel } from "../preferences.ts";
 // Zero-import module — imported by path rather than through the package
@@ -399,6 +404,25 @@ test("pauseAutoForProviderError falls back to indefinite pause when not rate lim
   ]);
 });
 
+test("isTerminalDeletedWorktreeProviderError matches removed auto-worktree paths only", () => {
+  assert.equal(
+    isTerminalDeletedWorktreeProviderError('Path "/Users/dev/.gsd/projects/abc123/worktrees/M005" does not exist'),
+    true,
+  );
+  assert.equal(
+    isTerminalDeletedWorktreeProviderError('Path "/Users/dev/app/.gsd/worktrees/M005" does not exist'),
+    true,
+  );
+  assert.equal(
+    isTerminalDeletedWorktreeProviderError('Path "/Users/dev/app/src/file.ts" does not exist'),
+    false,
+  );
+  assert.equal(
+    isTerminalDeletedWorktreeProviderError('Path "/Users/dev/.gsd/projects/abc123/worktrees/M005" failed with EACCES'),
+    false,
+  );
+});
+
 // ── resumeAutoAfterProviderDelay ────────────────────────────────────────────
 
 test("resumeAutoAfterProviderDelay restarts paused auto-mode from the recorded base path", async () => {
@@ -659,3 +683,15 @@ test("agent-session retryable error regex matches server_error (underscore)", ()
   // "temporarily backed off" must NOT be matched (intentional exclusion #3429)
   assert.ok(!RETRYABLE_ERROR_RE.test("temporarily backed off"));
 });
+
+test("exhausted retry errors are not deferred back to core retry handling", () => {
+  const cls = classifyError("Retry failed after 3 attempts: 500 empty_stream: upstream stream closed before first payload");
+  assert.equal(cls.kind, "server");
+  assert.equal(
+    shouldDeferTransientErrorToCoreRetry(
+      cls,
+      "Retry failed after 3 attempts: 500 empty_stream: upstream stream closed before first payload",
+    ),
+    false,
+  );
+});

package/src/resources/extensions/gsd/tests/queue-reorder-ui.test.ts

@@ -0,0 +1,54 @@
+import { describe, test } from "node:test";
+import assert from "node:assert/strict";
+
+import { showQueueReorder } from "../queue-reorder-ui.ts";
+
+const fakeTheme = {
+  fg: (_color: string, text: string) => text,
+  bold: (text: string) => text,
+};
+
+describe("queue-reorder-ui", () => {
+  test("keeps cursor visible while scrolling long queue with arrow keys (#4656)", async () => {
+    const originalRowsDescriptor = Object.getOwnPropertyDescriptor(process.stdout, "rows");
+    Object.defineProperty(process.stdout, "rows", { value: 20, configurable: true });
+
+    try {
+      const pending = Array.from({ length: 20 }, (_, idx) => ({
+        id: `M${String(idx + 1).padStart(3, "0")}`,
+        title: `Milestone ${idx + 1}`,
+      }));
+
+      let resolved: { order: string[]; depsToRemove: Array<{ milestone: string; dep: string }> } | null = null;
+      let lastRender: string[] = [];
+
+      const ctx = {
+        hasUI: true,
+        ui: {
+          custom: async (factory: any) => {
+            const component = factory({ requestRender() {} }, fakeTheme, null, (value: any) => {
+              resolved = value;
+            });
+
+            for (let i = 0; i < 15; i++) component.handleInput("\u001b[B");
+            lastRender = component.render(100);
+            component.handleInput("\r");
+            return resolved;
+          },
+        },
+      } as any;
+
+      await showQueueReorder(ctx, [], pending);
+
+      const joined = lastRender.join("\n");
+      assert.ok(joined.includes("M016"), "selected item should stay visible after scrolling");
+      assert.ok(lastRender.length <= 16, `overlay should fit terminal max-height, got ${lastRender.length}`);
+    } finally {
+      if (originalRowsDescriptor) {
+        Object.defineProperty(process.stdout, "rows", originalRowsDescriptor);
+      } else {
+        delete (process.stdout as { rows?: number }).rows;
+      }
+    }
+  });
+});

package/src/resources/extensions/gsd/tests/remediation-completion-guard.test.ts

@@ -1,6 +1,6 @@
 /**
- * Regression test for #2675: completing-milestone dispatch rule must
- * block completion when VALIDATION verdict is "needs-remediation".
+ * Regression tests for non-passing VALIDATION verdicts: completing-milestone
+ * dispatch must block completion when VALIDATION needs remediation or attention.
  *
  * Without this guard, needs-remediation + allSlicesDone causes a loop:
  * complete-milestone dispatched → agent refuses (correct) → no SUMMARY
@@ -66,6 +66,93 @@ test("completing-milestone blocks when VALIDATION verdict is needs-remediation (
   }
 });
 
+test("completing-milestone blocks when VALIDATION verdict is needs-attention (#5747)", async () => {
+  const base = mkdtempSync(join(tmpdir(), "gsd-attention-"));
+  mkdirSync(join(base, ".gsd", "milestones", "M001"), { recursive: true });
+
+  try {
+    writeFileSync(
+      join(base, ".gsd", "milestones", "M001", "M001-VALIDATION.md"),
+      [
+        "---",
+        "verdict: needs-attention",
+        "remediation_round: 0",
+        "---",
+        "",
+        "# Validation Report",
+        "",
+        "Acceptance proof is incomplete and needs human attention.",
+      ].join("\n"),
+    );
+
+    const ctx = {
+      mid: "M001",
+      midTitle: "Test Milestone",
+      basePath: base,
+      state: { phase: "completing-milestone" } as any,
+      prefs: {} as any,
+      session: undefined,
+    };
+
+    const result = await completingRule!.match(ctx);
+
+    assert.ok(result !== null, "rule should match");
+    assert.equal(result!.action, "stop", "should return stop action");
+    if (result!.action === "stop") {
+      assert.equal(result!.level, "warning", "should be warning level (pausable)");
+      assert.ok(
+        result!.reason.includes("needs-attention"),
+        "reason should mention needs-attention",
+      );
+    }
+  } finally {
+    rmSync(base, { recursive: true, force: true });
+  }
+});
+
+test("completing-milestone blocks when VALIDATION verdict is fail (#5920)", async () => {
+  const base = mkdtempSync(join(tmpdir(), "gsd-fail-verdict-"));
+  mkdirSync(join(base, ".gsd", "milestones", "M001"), { recursive: true });
+
+  try {
+    writeFileSync(
+      join(base, ".gsd", "milestones", "M001", "M001-VALIDATION.md"),
+      [
+        "---",
+        "verdict: fail",
+        "---",
+        "",
+        "# Validation Report",
+        "",
+        "Blocking failures remain unresolved.",
+      ].join("\n"),
+    );
+
+    const ctx = {
+      mid: "M001",
+      midTitle: "Test Milestone",
+      basePath: base,
+      state: { phase: "completing-milestone" } as any,
+      prefs: {} as any,
+      session: undefined,
+    };
+
+    const result = await completingRule!.match(ctx);
+
+    assert.ok(result !== null, "rule should match");
+    assert.equal(result!.action, "stop", "should return stop action");
+    if (result!.action === "stop") {
+      assert.equal(result!.level, "warning", "should be warning level (pausable)");
+      assert.ok(
+        result!.reason.includes('verdict is "fail"'),
+        "reason should mention fail verdict",
+      );
+    }
+  } finally {
+    rmSync(base, { recursive: true, force: true });
+  }
+});
+
 test("completing-milestone proceeds normally when VALIDATION verdict is pass (#2675 guard)", async () => {
   const base = mkdtempSync(join(tmpdir(), "gsd-remediation-"));
   mkdirSync(join(base, ".gsd", "milestones", "M001"), { recursive: true });

package/src/resources/extensions/gsd/tests/run-uat-replay-cap.test.ts

@@ -36,7 +36,7 @@ function makeUatProject(): string {
   return base;
 }
 
-test("run-uat dispatch stops after three attempts without a verdict", async () => {
+test("run-uat dispatch skips after three attempts without a verdict", async () => {
   const basePath = makeUatProject();
   const rule = DISPATCH_RULES.find((r) => r.name === "run-uat (post-completion)");
   assert.ok(rule, "run-uat dispatch rule is registered");
@@ -58,8 +58,7 @@ test("run-uat dispatch stops after three attempts without a verdict", async () =
     }
 
     const capped = await rule.match(ctx as any);
-    assert.equal(capped?.action, "stop");
-    assert.match(capped?.reason ?? "", /dispatched 3 times/);
+    assert.equal(capped?.action, "skip");
     assert.equal(getUatCount(basePath, "M001", "S01"), 4);
   } finally {
     rmSync(basePath, { recursive: true, force: true });

package/src/resources/extensions/gsd/tests/session-switch-abort-misclassification.test.ts

@@ -5,6 +5,7 @@ import test from "node:test";
 import assert from "node:assert/strict";
 
 import {
+  _hasEmptyAgentEndContent,
   _handleSessionSwitchAgentEnd,
   isBareClaudeCodeStreamAbortPlaceholder,
   isClaudeCodeSessionSwitchAbortMessage,
@@ -188,6 +189,15 @@ test("empty-content aborted during session-switch is silently ignored", () => {
   assert.equal(cancelledWith, null);
 });
 
+test("missing agent_end content is classified as empty abort content", () => {
+  // Providers may omit content entirely for a late aborted agent_end. That is
+  // equivalent to empty content and must not pause/cancel the next unit.
+  assert.equal(_hasEmptyAgentEndContent(undefined), true);
+  assert.equal(_hasEmptyAgentEndContent(null), true);
+  assert.equal(_hasEmptyAgentEndContent([]), true);
+  assert.equal(_hasEmptyAgentEndContent([{ type: "text", text: "partial" }]), false);
+});
+
 test("completed assistant content with aborted stopReason during session-switch is ignored", () => {
   // newSession() can abort the just-finished provider stream while the last
   // assistant message still carries the completed unit summary. That is a

package/src/resources/extensions/gsd/tests/smart-entry-routing.test.ts

@@ -0,0 +1,113 @@
+// GSD-2 — Smart entry routing behavior tests.
+// Verifies guided wizard choices resolve to the correct execution path.
+
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  resolveActiveTaskChoiceRoute,
+  resolveGuidedExecuteLaunchMode,
+  type ActiveTaskChoice,
+  type ActiveTaskRoute,
+  type SmartEntryIsolationMode,
+} from "../smart-entry-routing.ts";
+
+test("guided execute route enters auto step bootstrap only for worktree isolation", () => {
+  const cases: Array<{
+    isolationMode: SmartEntryIsolationMode;
+    expectedRoute: ActiveTaskRoute;
+  }> = [
+    {
+      isolationMode: "worktree",
+      expectedRoute: {
+        kind: "auto-bootstrap",
+        verboseMode: false,
+        options: {
+          step: true,
+          milestoneLock: "M001",
+        },
+      },
+    },
+    {
+      isolationMode: "none",
+      expectedRoute: {
+        kind: "guided-dispatch",
+        unitType: "execute-task",
+      },
+    },
+    {
+      isolationMode: "branch",
+      expectedRoute: {
+        kind: "guided-dispatch",
+        unitType: "execute-task",
+      },
+    },
+  ];
+
+  for (const testCase of cases) {
+    assert.deepEqual(
+      resolveActiveTaskChoiceRoute({
+        choice: "execute",
+        isolationMode: testCase.isolationMode,
+        milestoneId: "M001",
+      }),
+      testCase.expectedRoute,
+    );
+  }
+});
+
+test("active task smart entry choices resolve to explicit routes", () => {
+  const cases: Array<{
+    choice: ActiveTaskChoice;
+    expectedRoute: ActiveTaskRoute;
+  }> = [
+    {
+      choice: "auto",
+      expectedRoute: {
+        kind: "auto-bootstrap",
+        verboseMode: false,
+      },
+    },
+    {
+      choice: "status",
+      expectedRoute: {
+        kind: "status",
+      },
+    },
+    {
+      choice: "milestone_actions",
+      expectedRoute: {
+        kind: "milestone-actions",
+      },
+    },
+  ];
+
+  for (const testCase of cases) {
+    assert.deepEqual(
+      resolveActiveTaskChoiceRoute({
+        choice: testCase.choice,
+        isolationMode: "worktree",
+        milestoneId: "M001",
+      }),
+      testCase.expectedRoute,
+    );
+  }
+});
+
+test("active task route rejects invalid choices from untyped callers", () => {
+  assert.throws(
+    () =>
+      resolveActiveTaskChoiceRoute({
+        choice: "not_yet" as ActiveTaskChoice,
+        isolationMode: "worktree",
+        milestoneId: "M001",
+      }),
+    /Invalid ActiveTaskChoice: not_yet/,
+  );
+});
+
+test("guided execute launch mode remains a small compatibility helper", () => {
+  assert.equal(resolveGuidedExecuteLaunchMode("worktree"), "auto-step");
+  assert.equal(resolveGuidedExecuteLaunchMode("none"), "guided-dispatch");
+  assert.equal(resolveGuidedExecuteLaunchMode("branch"), "guided-dispatch");
+});