gsd-pi 2.64.0 → 2.65.0-dev.5c8557b

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

@@ -0,0 +1,457 @@
+/**
+ * pre-execution-pause-wiring.test.ts — Integration tests for pre-execution check → pauseAuto wiring.
+ *
+ * Tests that verify the control flow from pre-execution checks through to pauseAuto:
+ *   1. When runPreExecutionChecks returns status: "fail" with blocking: true, pauseAuto is called
+ *   2. When enhanced_verification_strict: true and status: "warn", pauseAuto is also called
+ *
+ * These are integration-level tests that exercise the actual postUnitPostVerification function
+ * with controlled mocks for external dependencies.
+ */
+
+import { describe, test, mock, beforeEach, afterEach } from "node:test";
+import assert from "node:assert/strict";
+import { tmpdir } from "node:os";
+import { mkdirSync, writeFileSync, rmSync } from "node:fs";
+import { join } from "node:path";
+
+import { postUnitPostVerification, type PostUnitContext } from "../auto-post-unit.ts";
+import { AutoSession } from "../auto/session.ts";
+import { openDatabase, closeDatabase, insertMilestone, insertSlice, insertTask } from "../gsd-db.ts";
+import { invalidateAllCaches } from "../cache.ts";
+import { _clearGsdRootCache } from "../paths.ts";
+
+// ─── Test Fixtures ───────────────────────────────────────────────────────────
+
+let tempDir: string;
+let dbPath: string;
+let originalCwd: string;
+
+/**
+ * Create a minimal mock ExtensionContext.
+ */
+function makeMockCtx() {
+  return {
+    ui: {
+      notify: mock.fn(),
+      setStatus: () => {},
+      setWidget: () => {},
+      setFooter: () => {},
+    },
+    model: { id: "test-model" },
+  } as any;
+}
+
+/**
+ * Create a minimal mock ExtensionAPI.
+ */
+function makeMockPi() {
+  return {
+    sendMessage: mock.fn(),
+    setModel: mock.fn(async () => true),
+  } as any;
+}
+
+/**
+ * Create a minimal AutoSession for testing.
+ */
+function makeMockSession(basePath: string, currentUnit?: { type: string; id: string }): AutoSession {
+  const s = new AutoSession();
+  s.basePath = basePath;
+  s.active = true;
+  if (currentUnit) {
+    s.currentUnit = {
+      type: currentUnit.type,
+      id: currentUnit.id,
+      startedAt: Date.now(),
+    };
+  }
+  return s;
+}
+
+/**
+ * Create a PostUnitContext with a mockable pauseAuto.
+ */
+function makePostUnitContext(
+  s: AutoSession,
+  ctx: ReturnType<typeof makeMockCtx>,
+  pi: ReturnType<typeof makeMockPi>,
+  pauseAutoMock: ReturnType<typeof mock.fn>,
+): PostUnitContext {
+  return {
+    s,
+    ctx,
+    pi,
+    buildSnapshotOpts: () => ({}),
+    lockBase: () => tempDir,
+    stopAuto: mock.fn(async () => {}) as unknown as PostUnitContext["stopAuto"],
+    pauseAuto: pauseAutoMock as unknown as PostUnitContext["pauseAuto"],
+    updateProgressWidget: () => {},
+  };
+}
+
+/**
+ * Set up a temp directory with GSD structure and DB.
+ * Also changes cwd so preferences loading finds the right PREFERENCES.md.
+ */
+function setupTestEnvironment(): void {
+  // Save original cwd so we can restore it
+  originalCwd = process.cwd();
+  
+  tempDir = join(tmpdir(), `pre-exec-pause-test-${Date.now()}-${Math.random().toString(36).slice(2)}`);
+  mkdirSync(tempDir, { recursive: true });
+  
+  // Create .gsd directory structure
+  const gsdDir = join(tempDir, ".gsd");
+  mkdirSync(gsdDir, { recursive: true });
+  
+  // Create milestones directory structure
+  const milestonesDir = join(gsdDir, "milestones", "M001", "slices", "S01", "tasks");
+  mkdirSync(milestonesDir, { recursive: true });
+  
+  // Change cwd so loadEffectiveGSDPreferences finds our PREFERENCES.md
+  process.chdir(tempDir);
+  
+  // Clear gsdRoot cache so it finds the new .gsd directory
+  _clearGsdRootCache();
+  
+  // Initialize DB
+  dbPath = join(gsdDir, "gsd.db");
+  openDatabase(dbPath);
+}
+
+/**
+ * Clean up test environment.
+ */
+function cleanupTestEnvironment(): void {
+  // Restore original cwd before cleanup
+  try {
+    process.chdir(originalCwd);
+  } catch {
+    // Ignore if original cwd doesn't exist
+  }
+  
+  try {
+    closeDatabase();
+  } catch {
+    // Ignore close errors
+  }
+  try {
+    rmSync(tempDir, { recursive: true, force: true });
+  } catch {
+    // Ignore cleanup errors
+  }
+}
+
+/**
+ * Create a PREFERENCES.md file with specified preferences.
+ * Uses YAML frontmatter format (---\nkey: value\n---).
+ * Also invalidates caches so the preferences are re-read.
+ */
+function writePreferences(prefs: Record<string, unknown>): void {
+  const yamlLines = Object.entries(prefs).map(([k, v]) => `${k}: ${JSON.stringify(v)}`);
+  const prefsContent = `---
+${yamlLines.join("\n")}
+---
+
+# GSD Preferences
+`;
+  writeFileSync(join(tempDir, ".gsd", "PREFERENCES.md"), prefsContent);
+  // Invalidate caches so the new preferences file is found
+  invalidateAllCaches();
+  _clearGsdRootCache();
+}
+
+/**
+ * Create tasks in DB that will cause pre-execution checks to fail.
+ * A task that references a non-existent file will produce a blocking failure.
+ */
+function createFailingTasks(): void {
+  // Insert milestone first
+  insertMilestone({ id: "M001" });
+
+  // Insert slice
+  insertSlice({
+    id: "S01",
+    milestoneId: "M001",
+    title: "Test Slice",
+    risk: "low",
+  });
+
+  // Create a task that references a file that doesn't exist
+  // This will cause checkFilePathConsistency to produce a blocking failure
+  insertTask({
+    id: "T01",
+    sliceId: "S01",
+    milestoneId: "M001",
+    title: "Task with missing file",
+    status: "pending",
+    planning: {
+      description: "This task references a non-existent file",
+      estimate: "1h",
+      files: ["nonexistent-file-that-does-not-exist.ts"],
+      verify: "npm test",
+      inputs: [],
+      expectedOutput: [],
+      observabilityImpact: "",
+    },
+    sequence: 0,
+  });
+}
+
+/**
+ * Create tasks in DB that will produce only warnings (non-blocking issues).
+ * Interface contract mismatches produce warnings, not blocking failures.
+ */
+function createWarningOnlyTasks(): void {
+  // Insert milestone first
+  insertMilestone({ id: "M001" });
+
+  // Insert slice
+  insertSlice({
+    id: "S01",
+    milestoneId: "M001",
+    title: "Test Slice",
+    risk: "low",
+  });
+
+  // Create tasks with interface contract mismatch (produces warn, not fail)
+  insertTask({
+    id: "T01",
+    sliceId: "S01",
+    milestoneId: "M001",
+    title: "Task 1 with function signature",
+    status: "pending",
+    planning: {
+      description: `
+\`\`\`typescript
+function processData(input: string): boolean
+\`\`\`
+      `.trim(),
+      estimate: "1h",
+      files: [],
+      verify: "npm test",
+      inputs: [],
+      expectedOutput: [],
+      observabilityImpact: "",
+    },
+    sequence: 0,
+  });
+
+  insertTask({
+    id: "T02",
+    sliceId: "S01",
+    milestoneId: "M001",
+    title: "Task 2 with mismatched signature",
+    status: "pending",
+    planning: {
+      description: `
+\`\`\`typescript
+function processData(input: number): string
+\`\`\`
+      `.trim(),
+      estimate: "1h",
+      files: [],
+      verify: "npm test",
+      inputs: [],
+      expectedOutput: [],
+      observabilityImpact: "",
+    },
+    sequence: 1,
+  });
+}
+
+// ─── Tests ───────────────────────────────────────────────────────────────────
+
+describe("Pre-execution checks → pauseAuto wiring", () => {
+  beforeEach(() => {
+    setupTestEnvironment();
+  });
+
+  afterEach(() => {
+    cleanupTestEnvironment();
+  });
+
+  test("pauseAuto is called when pre-execution checks return status: fail with blocking: true", async () => {
+    // Set up tasks that will cause a blocking failure
+    createFailingTasks();
+
+    // Create mocks
+    const ctx = makeMockCtx();
+    const pi = makeMockPi();
+    const pauseAutoMock = mock.fn(async () => {});
+    const s = makeMockSession(tempDir, { type: "plan-slice", id: "M001/S01" });
+    const pctx = makePostUnitContext(s, ctx, pi, pauseAutoMock);
+
+    // Call postUnitPostVerification
+    const result = await postUnitPostVerification(pctx);
+
+    // Verify pauseAuto was called
+    assert.equal(
+      pauseAutoMock.mock.callCount(),
+      1,
+      "pauseAuto should be called exactly once when pre-execution checks fail with blocking issues"
+    );
+
+    // Verify return value is "stopped"
+    assert.equal(
+      result,
+      "stopped",
+      "postUnitPostVerification should return 'stopped' when pre-execution checks fail"
+    );
+
+    // Verify UI was notified of the failure
+    const notifyCalls = ctx.ui.notify.mock.calls;
+    const errorNotify = notifyCalls.find(
+      (call: { arguments: unknown[] }) =>
+        call.arguments[1] === "error" &&
+        String(call.arguments[0]).includes("Pre-execution checks failed")
+    );
+    assert.ok(errorNotify, "Should show error notification about pre-execution check failure");
+  });
+
+  test("pauseAuto is called when enhanced_verification_strict: true and pre-execution returns warn", async () => {
+    // Write preferences with strict mode enabled
+    writePreferences({
+      enhanced_verification: true,
+      enhanced_verification_pre: true,
+      enhanced_verification_strict: true,
+    });
+
+    // Set up tasks that will produce only warnings (interface contract mismatch)
+    createWarningOnlyTasks();
+
+    // Create mocks
+    const ctx = makeMockCtx();
+    const pi = makeMockPi();
+    const pauseAutoMock = mock.fn(async () => {});
+    const s = makeMockSession(tempDir, { type: "plan-slice", id: "M001/S01" });
+    const pctx = makePostUnitContext(s, ctx, pi, pauseAutoMock);
+
+    // Call postUnitPostVerification
+    const result = await postUnitPostVerification(pctx);
+
+    // Verify pauseAuto was called (strict mode promotes warnings to blocking)
+    assert.equal(
+      pauseAutoMock.mock.callCount(),
+      1,
+      "pauseAuto should be called when strict mode is enabled and pre-execution returns warn"
+    );
+
+    // Verify return value is "stopped"
+    assert.equal(
+      result,
+      "stopped",
+      "postUnitPostVerification should return 'stopped' when strict mode treats warnings as blocking"
+    );
+
+    // Verify UI was notified of the warning
+    const notifyCalls = ctx.ui.notify.mock.calls;
+    const warnNotify = notifyCalls.find(
+      (call: { arguments: unknown[] }) =>
+        call.arguments[1] === "warning" &&
+        String(call.arguments[0]).includes("Pre-execution checks passed with warnings")
+    );
+    assert.ok(warnNotify, "Should show warning notification about pre-execution check warnings");
+  });
+
+  test("pauseAuto is NOT called when enhanced_verification_strict: false and pre-execution returns warn", async () => {
+    // Write preferences with strict mode disabled (default behavior)
+    writePreferences({
+      enhanced_verification: true,
+      enhanced_verification_pre: true,
+      enhanced_verification_strict: false,
+    });
+
+    // Set up tasks that will produce only warnings
+    createWarningOnlyTasks();
+
+    // Create mocks
+    const ctx = makeMockCtx();
+    const pi = makeMockPi();
+    const pauseAutoMock = mock.fn(async () => {});
+    const s = makeMockSession(tempDir, { type: "plan-slice", id: "M001/S01" });
+    const pctx = makePostUnitContext(s, ctx, pi, pauseAutoMock);
+
+    // Call postUnitPostVerification
+    const result = await postUnitPostVerification(pctx);
+
+    // Verify pauseAuto was NOT called (warnings don't block in non-strict mode)
+    assert.equal(
+      pauseAutoMock.mock.callCount(),
+      0,
+      "pauseAuto should NOT be called when strict mode is disabled and only warnings exist"
+    );
+
+    // Verify return value is "continue" (not "stopped")
+    assert.equal(
+      result,
+      "continue",
+      "postUnitPostVerification should return 'continue' when warnings don't block in non-strict mode"
+    );
+  });
+
+  test("pre-execution checks are skipped when unit type is not plan-slice", async () => {
+    // Set up tasks that would fail if checked
+    createFailingTasks();
+
+    // Create mocks with execute-task unit (not plan-slice)
+    const ctx = makeMockCtx();
+    const pi = makeMockPi();
+    const pauseAutoMock = mock.fn(async () => {});
+    const s = makeMockSession(tempDir, { type: "execute-task", id: "M001/S01/T01" });
+    const pctx = makePostUnitContext(s, ctx, pi, pauseAutoMock);
+
+    // Call postUnitPostVerification
+    const result = await postUnitPostVerification(pctx);
+
+    // Verify pauseAuto was NOT called (pre-execution checks only run for plan-slice)
+    assert.equal(
+      pauseAutoMock.mock.callCount(),
+      0,
+      "pauseAuto should NOT be called for non-plan-slice unit types"
+    );
+
+    // Verify return value is "continue"
+    assert.equal(
+      result,
+      "continue",
+      "postUnitPostVerification should return 'continue' for non-plan-slice unit types"
+    );
+  });
+
+  test("pre-execution checks are skipped when enhanced_verification_pre: false", async () => {
+    // Write preferences with pre-execution checks disabled
+    writePreferences({
+      enhanced_verification: true,
+      enhanced_verification_pre: false,
+    });
+
+    // Set up tasks that would fail if checked
+    createFailingTasks();
+
+    // Create mocks
+    const ctx = makeMockCtx();
+    const pi = makeMockPi();
+    const pauseAutoMock = mock.fn(async () => {});
+    const s = makeMockSession(tempDir, { type: "plan-slice", id: "M001/S01" });
+    const pctx = makePostUnitContext(s, ctx, pi, pauseAutoMock);
+
+    // Call postUnitPostVerification
+    const result = await postUnitPostVerification(pctx);
+
+    // Verify pauseAuto was NOT called (pre-execution checks disabled)
+    assert.equal(
+      pauseAutoMock.mock.callCount(),
+      0,
+      "pauseAuto should NOT be called when enhanced_verification_pre is disabled"
+    );
+
+    // Verify return value is "continue"
+    assert.equal(
+      result,
+      "continue",
+      "postUnitPostVerification should return 'continue' when pre-execution checks are disabled"
+    );
+  });
+});

package/src/resources/extensions/gsd/tests/unstructured-continue-context-injection.test.ts

@@ -0,0 +1,163 @@
+// GSD-2 — Regression test for #3615: unstructured "continue" must inject task context
+// Copyright (c) 2026 Jeremy McSpadden <jeremy@fluxlabs.net>
+
+/**
+ * Bug #3615: When a user types "continue" (or any bare text) to resume
+ * an in-progress session, buildGuidedExecuteContextInjection() only
+ * matched two hardcoded regex patterns (auto-dispatch and guided-resume).
+ * The function returned null for any other input, so no task context was
+ * injected — causing the agent to rebuild everything from scratch and
+ * burn ~86k tokens.
+ *
+ * This test verifies:
+ *   1. Structural: the fallback exists with phase + intent guards
+ *   2. Behavioral: RESUME_INTENT_PATTERNS matches expected prompts and
+ *      rejects non-resume prompts (control, help, diagnostic, etc.)
+ */
+
+import { describe, test } from "node:test";
+import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
+import { join, dirname } from "node:path";
+import { fileURLToPath } from "node:url";
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
+const systemContextSource = readFileSync(
+  join(__dirname, "..", "bootstrap", "system-context.ts"),
+  "utf-8",
+);
+
+// ── Structural tests ────────────────────────────────────────────────
+
+describe("#3615 — structural: fallback exists with correct guards", () => {
+  const fnStart = systemContextSource.indexOf("async function buildGuidedExecuteContextInjection(");
+  assert.ok(fnStart >= 0, "should find buildGuidedExecuteContextInjection");
+  const fnEnd = systemContextSource.indexOf("\nasync function ", fnStart + 1);
+  const fnBody = fnEnd >= 0
+    ? systemContextSource.slice(fnStart, fnEnd)
+    : systemContextSource.slice(fnStart);
+
+  test("has a deriveState fallback after the two regex branches", () => {
+    const deriveStateCalls = fnBody.match(/deriveState\(basePath\)/g);
+    assert.ok(
+      deriveStateCalls && deriveStateCalls.length >= 2,
+      `expected >=2 deriveState(basePath) calls, got ${deriveStateCalls?.length ?? 0}`,
+    );
+  });
+
+  test("fallback is phase-gated to executing only", () => {
+    const afterFallback = fnBody.indexOf("// Fallback:");
+    assert.ok(afterFallback >= 0, "should have a fallback comment");
+    const fallbackSection = fnBody.slice(afterFallback);
+    assert.ok(
+      fallbackSection.includes('state.phase === "executing"'),
+      'fallback must be gated on state.phase === "executing"',
+    );
+  });
+
+  test("fallback is intent-gated via RESUME_INTENT_PATTERNS", () => {
+    const afterFallback = fnBody.indexOf("// Fallback:");
+    const fallbackSection = fnBody.slice(afterFallback);
+    assert.ok(
+      fallbackSection.includes("RESUME_INTENT_PATTERNS"),
+      "fallback must check RESUME_INTENT_PATTERNS before deriveState",
+    );
+  });
+
+  test("fallback calls buildTaskExecutionContextInjection with derived state", () => {
+    const afterFallback = fnBody.indexOf("// Fallback:");
+    const fallbackSection = fnBody.slice(afterFallback);
+    assert.ok(
+      fallbackSection.includes("buildTaskExecutionContextInjection") &&
+      fallbackSection.includes("state.activeMilestone.id") &&
+      fallbackSection.includes("state.activeSlice.id") &&
+      fallbackSection.includes("state.activeTask.id"),
+      "fallback must call buildTaskExecutionContextInjection with state-derived IDs",
+    );
+  });
+
+  test("only one return null at the end", () => {
+    const returnNulls = fnBody.match(/return null;/g);
+    assert.ok(
+      returnNulls && returnNulls.length === 1,
+      `expected exactly 1 'return null' (at end after fallback), got ${returnNulls?.length ?? 0}`,
+    );
+  });
+});
+
+// ── Behavioral tests: RESUME_INTENT_PATTERNS ────────────────────────
+
+describe("#3615 — behavioral: RESUME_INTENT_PATTERNS matches resume prompts", () => {
+  // Extract the regex from source so the test stays in sync
+  const patternMatch = systemContextSource.match(/const RESUME_INTENT_PATTERNS\s*=\s*\/(.+)\/;/);
+  assert.ok(patternMatch, "should find RESUME_INTENT_PATTERNS definition");
+  const pattern = new RegExp(patternMatch[1]);
+
+  // Helper: normalize prompt the same way the production code does
+  const normalize = (s: string) => s.trim().toLowerCase().replace(/[.!?,]+$/g, "");
+
+  const shouldMatch = [
+    "continue",
+    "Continue",
+    "CONTINUE",
+    "continue.",
+    "continue!",
+    "resume",
+    "ok",
+    "OK",
+    "Ok!",
+    "go",
+    "go ahead",
+    "Go ahead.",
+    "proceed",
+    "keep going",
+    "carry on",
+    "next",
+    "yes",
+    "yeah",
+    "yep",
+    "sure",
+    "do it",
+    "let's go",
+    "pick up where you left off",
+    "  continue  ",  // whitespace padded
+  ];
+
+  const shouldNotMatch = [
+    "help",
+    "status",
+    "/gsd auto",
+    "/gsd stats",
+    "what's the plan?",
+    "show me the logs",
+    "abort",
+    "stop",
+    "cancel",
+    "replan this slice",
+    "I think we should change the approach",
+    "can you explain what you just did?",
+    "run the tests",
+    "check the build",
+    "Execute the next task: T01",
+    "what files were changed",
+    "",
+  ];
+
+  for (const prompt of shouldMatch) {
+    test(`matches resume prompt: "${prompt}"`, () => {
+      assert.ok(
+        pattern.test(normalize(prompt)),
+        `expected RESUME_INTENT_PATTERNS to match "${prompt}" (normalized: "${normalize(prompt)}")`,
+      );
+    });
+  }
+
+  for (const prompt of shouldNotMatch) {
+    test(`rejects non-resume prompt: "${prompt}"`, () => {
+      assert.ok(
+        !pattern.test(normalize(prompt)),
+        `expected RESUME_INTENT_PATTERNS to NOT match "${prompt}" (normalized: "${normalize(prompt)}")`,
+      );
+    });
+  }
+});

package/src/resources/extensions/gsd/verification-evidence.ts

@@ -52,6 +52,32 @@ export interface BrowserEvidenceJSON {
   duration: number;
 }
 
+export interface PreExecutionCheckJSON {
+  /** Check category: package, file, tool, endpoint, schema */
+  category: "package" | "file" | "tool" | "endpoint" | "schema";
+  /** What was checked (e.g., package name, file path) */
+  target: string;
+  /** Whether the check passed */
+  passed: boolean;
+  /** Human-readable message explaining the result */
+  message: string;
+  /** Whether this failure should block execution (only meaningful when passed=false) */
+  blocking?: boolean;
+}
+
+export interface PostExecutionCheckJSON {
+  /** Check category: import, signature, pattern */
+  category: "import" | "signature" | "pattern";
+  /** What was checked (e.g., file:line, function name) */
+  target: string;
+  /** Whether the check passed */
+  passed: boolean;
+  /** Human-readable message explaining the result */
+  message: string;
+  /** Whether this failure should block completion (only meaningful when passed=false) */
+  blocking?: boolean;
+}
+
 export interface EvidenceJSON {
   schemaVersion: 1;
   taskId: string;
@@ -65,6 +91,10 @@ export interface EvidenceJSON {
   runtimeErrors?: RuntimeErrorJSON[];
   auditWarnings?: AuditWarningJSON[];
   browser?: BrowserEvidenceJSON;
+  /** Pre-execution checks run before task execution (package existence, file refs, etc.) */
+  preExecutionChecks?: PreExecutionCheckJSON[];
+  /** Post-execution checks run after task completion (import resolution, signature drift, pattern consistency) */
+  postExecutionChecks?: PostExecutionCheckJSON[];
 }
 
 /**
@@ -124,6 +154,44 @@ export function writeVerificationJSON(
   writeFileSync(filePath, JSON.stringify(evidence, null, 2) + "\n", "utf-8");
 }
 
+// ─── Pre-Execution Evidence ──────────────────────────────────────────────────
+
+export interface PreExecutionEvidenceJSON {
+  schemaVersion: 1;
+  milestoneId: string;
+  sliceId: string;
+  timestamp: number;
+  status: "pass" | "warn" | "fail";
+  durationMs: number;
+  checks: PreExecutionCheckJSON[];
+}
+
+/**
+ * Write pre-execution check results to a PRE-EXEC-VERIFY.json artifact
+ * in the slice directory.
+ */
+export function writePreExecutionEvidence(
+  result: { status: "pass" | "warn" | "fail"; checks: PreExecutionCheckJSON[]; durationMs: number },
+  sliceDir: string,
+  milestoneId: string,
+  sliceId: string,
+): void {
+  mkdirSync(sliceDir, { recursive: true });
+
+  const evidence: PreExecutionEvidenceJSON = {
+    schemaVersion: 1,
+    milestoneId,
+    sliceId,
+    timestamp: Date.now(),
+    status: result.status,
+    durationMs: result.durationMs,
+    checks: result.checks,
+  };
+
+  const filePath = join(sliceDir, `${sliceId}-PRE-EXEC-VERIFY.json`);
+  writeFileSync(filePath, JSON.stringify(evidence, null, 2) + "\n", "utf-8");
+}
+
 // ─── Markdown Evidence Table ─────────────────────────────────────────────────
 
 /**