gsd-pi 2.76.0-dev.fe143342a → 2.77.0-dev.1d17f366c

package/src/resources/extensions/gsd/tests/cmux.test.ts

@@ -14,7 +14,7 @@ import {
   shouldPromptToEnableCmux,
 } from "../../cmux/index.ts";
 import { autoEnableCmuxPreferences } from "../commands-cmux.ts";
-import type { GSDState } from "../types.ts";
+import type { CmuxStateInput } from "../../shared/cmux-events.ts";
 
 test("detectCmuxEnvironment requires workspace, surface, and socket", () => {
   const detected = detectCmuxEnvironment(
@@ -146,15 +146,11 @@ describe("autoEnableCmuxPreferences", () => {
 });
 
 test("buildCmuxStatusLabel and progress prefer deepest active unit", () => {
-  const state: GSDState = {
-    activeMilestone: { id: "M001", title: "Milestone" },
-    activeSlice: { id: "S02", title: "Slice" },
-    activeTask: { id: "T03", title: "Task" },
+  const state: CmuxStateInput = {
+    activeMilestone: { id: "M001" },
+    activeSlice: { id: "S02" },
+    activeTask: { id: "T03" },
     phase: "executing",
-    recentDecisions: [],
-    blockers: [],
-    nextAction: "Keep going",
-    registry: [],
     progress: {
       milestones: { done: 0, total: 1 },
       slices: { done: 1, total: 3 },

package/src/resources/extensions/gsd/tests/complete-milestone-false-merge.test.ts

@@ -105,6 +105,21 @@ test("#4175: postUnitPreVerification pauses complete-milestone after retries exh
   );
 });
 
+test("#4658: postUnitPreVerification waits briefly for DB-close on complete-milestone before retrying", () => {
+  assert.ok(
+    /async function waitForMilestoneDbClose/.test(postUnitSrc),
+    "auto-post-unit should define a DB settle helper for complete-milestone",
+  );
+  assert.ok(
+    /setTimeout\(resolve,\s*COMPLETE_MILESTONE_DB_SETTLE_POLL_MS\)/.test(postUnitSrc),
+    "DB settle helper should poll with a bounded timeout window",
+  );
+  assert.ok(
+    /s\.currentUnit\.type\s*===\s*"complete-milestone"[\s\S]*waitForMilestoneDbClose/.test(postUnitSrc),
+    "complete-milestone path should invoke DB settle check before retry flow",
+  );
+});
+
 test("#4175: recoverTimedOutUnit pauses complete-milestone instead of writing a blocker placeholder", () => {
   // The complete-milestone pause branch must sit immediately above the
   // "retries exhausted" writeBlockerPlaceholder call so a failed

package/src/resources/extensions/gsd/tests/complete-milestone.test.ts

@@ -1,11 +1,14 @@
 import { describe, test, afterEach } from "node:test";
 import assert from "node:assert/strict";
-import { mkdtempSync, mkdirSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { mkdtempSync, mkdirSync, readFileSync, rmSync, writeFileSync, existsSync } from "node:fs";
 import { join, dirname } from "node:path";
 import { tmpdir } from "node:os";
 import { fileURLToPath } from "node:url";
 import { invalidateAllCaches } from '../cache.ts';
 import { parseUnitId } from "../unit-id.ts";
+import { openDatabase, closeDatabase, insertMilestone, insertSlice, insertTask } from "../gsd-db.ts";
+import { clearPathCache } from "../paths.ts";
+import { clearParseCache } from "../files.ts";
 
 // loadPrompt reads from ~/.gsd/agent/extensions/gsd/prompts/ (main checkout).
 // In a worktree the file may not exist there yet, so we resolve prompts
@@ -404,6 +407,63 @@ describe("complete-milestone", () => {
     }
   });
 
+  test("handleCompleteMilestone does not overwrite existing SUMMARY.md on re-dispatch (#4598)", async () => {
+    // This test verifies that when SUMMARY.md already exists (from a prior completion),
+    // re-calling handleCompleteMilestone does not overwrite it.
+    // Before the fix this test FAILS because the handler unconditionally writes SUMMARY.md.
+    const { handleCompleteMilestone } = await import("../tools/complete-milestone.ts");
+    const base = createFixtureBase();
+    const mid = "M001";
+    const dbPath = join(base, ".gsd", "gsd.db");
+    try {
+      // Set up DB with milestone and a complete slice + task
+      openDatabase(dbPath);
+      insertMilestone({ id: mid, title: "Test Milestone", status: "active" });
+      insertSlice({ id: "S01", milestoneId: mid, title: "Slice One", status: "complete" });
+      insertTask({ id: "T01", sliceId: "S01", milestoneId: mid, title: "Task One", status: "complete" });
+
+      // Pre-write an existing SUMMARY.md to simulate a prior completion
+      const milestoneDir = join(base, ".gsd", "milestones", mid);
+      mkdirSync(milestoneDir, { recursive: true });
+      const summaryPath = join(milestoneDir, `${mid}-SUMMARY.md`);
+      const originalContent = "original content — must not be overwritten";
+      writeFileSync(summaryPath, originalContent, "utf-8");
+
+      // Call handleCompleteMilestone — this is the re-dispatch scenario
+      const params = {
+        milestoneId: mid,
+        title: "Test Milestone",
+        oneLiner: "Re-dispatched",
+        narrative: "This is a re-dispatch",
+        successCriteriaResults: "Met",
+        definitionOfDoneResults: "Done",
+        requirementOutcomes: "Covered",
+        keyDecisions: [],
+        keyFiles: [],
+        lessonsLearned: [],
+        followUps: "",
+        deviations: "",
+        verificationPassed: true,
+      };
+
+      const result = await handleCompleteMilestone(params, base);
+
+      // The call may return an error (milestone already complete) or success
+      // but in either case the SUMMARY.md must NOT be overwritten.
+      const actualContent = readFileSync(summaryPath, "utf-8");
+      assert.strictEqual(
+        actualContent,
+        originalContent,
+        "existing SUMMARY.md must not be overwritten on re-dispatch (#4598)",
+      );
+    } finally {
+      try { closeDatabase(); } catch { /* */ }
+      clearPathCache();
+      clearParseCache();
+      cleanup(base);
+    }
+  });
+
   test("deriveState completing-milestone integration", async () => {
     const { deriveState, isMilestoneComplete } = await import("../state.ts");
     const { invalidateAllCaches: invalidateAllCachesDynamic } = await import("../cache.ts");

package/src/resources/extensions/gsd/tests/complete-slice.test.ts

@@ -125,9 +125,9 @@ console.log('\n=== complete-slice: schema v6 migration ===');
 
   const adapter = _getAdapter()!;
 
-  // Verify schema version is current (v21 — ADR-013 structured_fields column)
+  // Verify schema version is current (v22 — quality_gates DDL fix)
   const versionRow = adapter.prepare('SELECT MAX(version) as v FROM schema_version').get();
-  assertEq(versionRow?.['v'], 21, 'schema version should be 21');
+  assertEq(versionRow?.['v'], 22, 'schema version should be 22');
 
   // Verify slices table has full_summary_md and full_uat_md columns
   const cols = adapter.prepare("PRAGMA table_info(slices)").all();

package/src/resources/extensions/gsd/tests/complete-task.test.ts

@@ -109,9 +109,9 @@ console.log('\n=== complete-task: schema v5 migration ===');
 
   const adapter = _getAdapter()!;
 
-  // Verify schema version is current (v21 — ADR-013 structured_fields column)
+  // Verify schema version is current (v22 — quality_gates DDL fix)
   const versionRow = adapter.prepare('SELECT MAX(version) as v FROM schema_version').get();
-  assertEq(versionRow?.['v'], 21, 'schema version should be 21');
+  assertEq(versionRow?.['v'], 22, 'schema version should be 22');
 
   // Verify all 4 new tables exist
   const tables = adapter.prepare(

package/src/resources/extensions/gsd/tests/custom-engine-loop-integration.test.ts

@@ -179,6 +179,8 @@ function makeMockDeps(overrides?: Partial<LoopDeps>): LoopDeps & { callLog: stri
     autoWorktreeBranch: () => "auto/M001",
     resolveMilestoneFile: () => null,
     reconcileMergeState: () => "clean",
+    preflightCleanRoot: () => ({ stashPushed: false, summary: "" }),
+    postflightPopStash: () => {},
     getLedger: () => null,
     getProjectTotals: () => ({ cost: 0 }),
     formatCost: (c: number) => `$${c.toFixed(2)}`,
@@ -479,6 +481,165 @@ describe("Custom engine loop integration", () => {
     );
   });
 
+  it("stops custom workflow after repeated verification retries", async () => {
+    _resetPendingResolve();
+
+    const runDir = makeTmpDir();
+    const graph = makeGraph([makeStep({ id: "retry-step" })], "retry-exhaustion");
+    writeGraph(runDir, graph);
+    writeFileSync(join(runDir, "DEFINITION.yaml"), stringify({
+      version: 1,
+      name: "retry-exhaustion",
+      steps: [{
+        id: "retry-step",
+        name: "retry-step",
+        prompt: "Do retry-step",
+        produces: "retry-step/output.md",
+        verify: { policy: "shell-command", command: "exit 1" },
+      }],
+    }));
+
+    const ctx = makeMockCtx();
+    const pi = makeMockPi();
+    const s = makeLoopSession({
+      activeEngineId: "custom",
+      activeRunDir: runDir,
+      basePath: runDir,
+    });
+    const deps = makeMockDeps({
+      stopAuto: async (_ctx, _pi, reason) => {
+        deps.callLog.push(`stopAuto:${reason ?? "no-reason"}`);
+        s.active = false;
+      },
+    });
+
+    const resolver = setInterval(() => {
+      resolveAgentEnd({ messages: [{ role: "assistant" }] });
+    }, 25);
+    let timeout: NodeJS.Timeout | undefined;
+    try {
+      await Promise.race([
+        autoLoop(ctx, pi, s, deps),
+        new Promise((_, reject) =>
+          timeout = setTimeout(() => {
+            s.active = false;
+            resolveAgentEnd({ messages: [{ role: "assistant" }] });
+            reject(new Error(
+              `autoLoop did not stop after verification retry exhaustion; calls=${pi.calls.length}; log=${deps.callLog.join(",")}`,
+            ));
+          }, 3_000),
+        ),
+      ]);
+    } finally {
+      clearInterval(resolver);
+      if (timeout) clearTimeout(timeout);
+    }
+
+    assert.equal(pi.calls.length, 4, "verification retry should be capped after four dispatched attempts");
+    const stopEntry = deps.callLog.find((e: string) => e.startsWith("stopAuto:"));
+    assert.match(stopEntry ?? "", /requested retry 4 times without passing/);
+    const finalGraph = readGraph(runDir);
+    assert.equal(finalGraph.steps[0]?.status, "active", "failed verification must not reconcile the step complete");
+  });
+
+  it("persists custom verification retry budget across a session restart", async () => {
+    _resetPendingResolve();
+
+    const runDir = makeTmpDir();
+    const graph = makeGraph([makeStep({ id: "retry-step" })], "retry-restart");
+    writeGraph(runDir, graph);
+    writeFileSync(join(runDir, "DEFINITION.yaml"), stringify({
+      version: 1,
+      name: "retry-restart",
+      steps: [{
+        id: "retry-step",
+        name: "retry-step",
+        prompt: "Do retry-step",
+        produces: "retry-step/output.md",
+        verify: { policy: "shell-command", command: "exit 1" },
+      }],
+    }));
+
+    const ctx1 = makeMockCtx();
+    const pi1 = makeMockPi();
+    const s1 = makeLoopSession({
+      activeEngineId: "custom",
+      activeRunDir: runDir,
+      basePath: runDir,
+    });
+    const deps1 = makeMockDeps();
+    const resolver1 = setInterval(() => {
+      resolveAgentEnd({ messages: [{ role: "assistant" }] });
+      if (pi1.calls.length >= 2) {
+        s1.active = false;
+      }
+    }, 25);
+    let timeout1: NodeJS.Timeout | undefined;
+    try {
+      await Promise.race([
+        autoLoop(ctx1, pi1, s1, deps1),
+        new Promise((_, reject) =>
+          timeout1 = setTimeout(() => {
+            s1.active = false;
+            resolveAgentEnd({ messages: [{ role: "assistant" }] });
+            reject(new Error(
+              `first autoLoop did not pause after two retry attempts; calls=${pi1.calls.length}; log=${deps1.callLog.join(",")}`,
+            ));
+          }, 3_000),
+        ),
+      ]);
+    } finally {
+      clearInterval(resolver1);
+      if (timeout1) clearTimeout(timeout1);
+    }
+    assert.equal(pi1.calls.length, 2, "first session should consume two retry attempts");
+    assert.equal(
+      deps1.callLog.some((e: string) => e.startsWith("stopAuto:")),
+      false,
+      "first session should stop because the session deactivated, not because retry budget exhausted",
+    );
+
+    _resetPendingResolve();
+    const ctx2 = makeMockCtx();
+    const pi2 = makeMockPi();
+    const s2 = makeLoopSession({
+      activeEngineId: "custom",
+      activeRunDir: runDir,
+      basePath: runDir,
+    });
+    const deps2 = makeMockDeps({
+      stopAuto: async (_ctx, _pi, reason) => {
+        deps2.callLog.push(`stopAuto:${reason ?? "no-reason"}`);
+        s2.active = false;
+      },
+    });
+    const resolver2 = setInterval(() => {
+      resolveAgentEnd({ messages: [{ role: "assistant" }] });
+    }, 25);
+    let timeout2: NodeJS.Timeout | undefined;
+    try {
+      await Promise.race([
+        autoLoop(ctx2, pi2, s2, deps2),
+        new Promise((_, reject) =>
+          timeout2 = setTimeout(() => {
+            s2.active = false;
+            resolveAgentEnd({ messages: [{ role: "assistant" }] });
+            reject(new Error(
+              `second autoLoop did not stop after persisted retry exhaustion; calls=${pi2.calls.length}; log=${deps2.callLog.join(",")}`,
+            ));
+          }, 3_000),
+        ),
+      ]);
+    } finally {
+      clearInterval(resolver2);
+      if (timeout2) clearTimeout(timeout2);
+    }
+
+    assert.equal(pi2.calls.length, 2, "second session should exhaust after attempts 3 and 4");
+    const stopEntry = deps2.callLog.find((e: string) => e.startsWith("stopAuto:"));
+    assert.match(stopEntry ?? "", /requested retry 4 times without passing/);
+  });
+
   it("GRAPH.yaml step stays pending when session deactivates before reconcile", async () => {
     _resetPendingResolve();
 

package/src/resources/extensions/gsd/tests/db-access-guardrails.test.ts

@@ -50,6 +50,7 @@ test("doctor-heal.md contains DB access guardrail naming gsd_milestone_status",
   const prompt = readPrompt("doctor-heal");
   assert.match(prompt, /gsd_milestone_status/, "doctor-heal.md must name gsd_milestone_status as the DB inspection tool");
   assert.match(prompt, /Do NOT query.*\.gsd\/gsd\.db/i, "doctor-heal.md must prohibit direct DB queries");
+  assert.doesNotMatch(prompt, /\{\{milestoneId\}\}/, "doctor-heal.md must not declare unprovided milestoneId template variables");
 });
 
 test("forensics.md contains DB inspection guardrail", () => {

package/src/resources/extensions/gsd/tests/derive-state.test.ts

@@ -431,7 +431,7 @@ Continue from step 2.
       cleanup(base1);
     }
 
-    // Case B: S01 depends on nonexistent S99 → truly blocked
+    // Case B: S01 depends on nonexistent S99 -> fallback picks best available slice
     const base2 = createFixtureBase();
     try {
       writeRoadmap(base2, 'M001', `# M001: Test Milestone
@@ -446,7 +446,6 @@ Continue from step 2.
 
       const state2 = await deriveState(base2);
 
-      // With partial-dep fallback, S01 is picked despite unmet dep on S99
       assert.deepStrictEqual(state2.phase, 'planning', 'blocked-B: phase is planning (fallback picks S01)');
       assert.deepStrictEqual(state2.activeSlice?.id, 'S01', 'blocked-B: activeSlice is S01 via fallback');
     } finally {

package/src/resources/extensions/gsd/tests/dispatch-complete-milestone-guard.test.ts

@@ -47,21 +47,26 @@ describe("completing-milestone dispatch guard (#4324)", () => {
     assert.ok(dispatchIdx > -1, "complete-milestone dispatch should exist after the skip guard");
   });
 
-  test("imports updateMilestoneStatus for DB reconciliation", () => {
-    assert.match(source, /import\s*\{[^}]*updateMilestoneStatus[^}]*\}\s*from\s*["']\.\/gsd-db/);
-  });
-
-  test("reconciles DB when SUMMARY exists on disk but DB is out of sync", () => {
+  test("classifies SUMMARY outcome and conditionally reconciles DB (#4658)", () => {
     const phaseCheck = source.indexOf('phase !== "completing-milestone"');
     // The SUMMARY-exists reconciliation guard must appear in this rule
     const summaryGuard = source.indexOf('resolveMilestoneFile(basePath, mid, "SUMMARY")', phaseCheck);
     assert.ok(summaryGuard > -1, "SUMMARY file check should exist in the completing-milestone rule");
 
+    const classifyCall = source.indexOf("classifyMilestoneSummaryContent", summaryGuard);
+    assert.ok(classifyCall > -1, "SUMMARY mismatch handling should classify summary content");
+    const dbGateBeforeClassify = source.indexOf("existingSummary && isDbAvailable()", summaryGuard);
+    assert.ok(
+      dbGateBeforeClassify === -1 || dbGateBeforeClassify > classifyCall,
+      "SUMMARY classification must not be gated on DB availability",
+    );
+
     const reconcileCall = source.indexOf('updateMilestoneStatus(mid, "complete"', summaryGuard);
-    assert.ok(reconcileCall > -1, "updateMilestoneStatus reconciliation should follow the SUMMARY check");
+    assert.ok(reconcileCall > -1, "successful SUMMARY should reconcile DB to complete");
 
-    // Reconciliation must happen before the dispatch action
-    const dispatchIdx = source.indexOf('unitType: "complete-milestone"', reconcileCall);
-    assert.ok(dispatchIdx > -1, "reconciliation should appear before the dispatch action");
+    const stopAction = source.indexOf('action: "stop"', summaryGuard);
+    assert.ok(stopAction > -1, "SUMMARY mismatch should return stop action");
+    const warningLevel = source.indexOf('level: "warning"', summaryGuard);
+    assert.ok(warningLevel > -1, "SUMMARY mismatch should be warning-level stop (pauses auto-mode)");
   });
 });

package/src/resources/extensions/gsd/tests/dispatch-guard-summary-db-mismatch.test.ts

@@ -0,0 +1,77 @@
+// GSD-2 dispatch-guard regression test: SUMMARY/DB mismatch fail-closed behavior (#4663)
+//
+// Sibling bug to #4658 / PR #4660. A failure-path SUMMARY file on disk
+// must not let the cross-milestone dispatch guard treat an "active"
+// milestone as complete. DB status is authoritative when available.
+
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, mkdirSync, rmSync, writeFileSync } from "node:fs";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+import { getPriorSliceCompletionBlocker } from "../dispatch-guard.ts";
+import { openDatabase, closeDatabase, insertMilestone, insertSlice } from "../gsd-db.ts";
+
+function setupRepo(): string {
+  const repo = mkdtempSync(join(tmpdir(), "gsd-dispatch-guard-4663-"));
+  mkdirSync(join(repo, ".gsd"), { recursive: true });
+  openDatabase(join(repo, ".gsd", "gsd.db"));
+  return repo;
+}
+
+function teardownRepo(repo: string): void {
+  closeDatabase();
+  rmSync(repo, { recursive: true, force: true });
+}
+
+test("#4663: dispatch guard blocks when prior milestone has failure SUMMARY but DB is still active", (t) => {
+  const repo = setupRepo();
+  t.after(() => teardownRepo(repo));
+
+  mkdirSync(join(repo, ".gsd", "milestones", "M002"), { recursive: true });
+  mkdirSync(join(repo, ".gsd", "milestones", "M003"), { recursive: true });
+
+  // M002: DB says active with a pending slice, but a failure SUMMARY exists on disk.
+  insertMilestone({ id: "M002", title: "Previous", status: "active" });
+  insertSlice({ id: "S01", milestoneId: "M002", title: "Pending", status: "pending", depends: [] });
+
+  insertMilestone({ id: "M003", title: "Current", status: "active" });
+  insertSlice({ id: "S01", milestoneId: "M003", title: "First", status: "pending", depends: [] });
+
+  writeFileSync(join(repo, ".gsd", "milestones", "M002", "M002-ROADMAP.md"), "# M002\n");
+  writeFileSync(join(repo, ".gsd", "milestones", "M003", "M003-ROADMAP.md"), "# M003\n");
+  writeFileSync(
+    join(repo, ".gsd", "milestones", "M002", "M002-SUMMARY.md"),
+    "# M002 Summary\nverification FAILED — not complete.\n",
+  );
+
+  // Before #4663: SUMMARY presence short-circuited the loop and M002 was skipped,
+  // allowing M003/S01 to dispatch. After: DB status is consulted and M002 still blocks.
+  assert.equal(
+    getPriorSliceCompletionBlocker(repo, "main", "plan-slice", "M003/S01"),
+    "Cannot dispatch plan-slice M003/S01: earlier slice M002/S01 is not complete.",
+  );
+});
+
+test("#4663: dispatch guard allows dispatch when prior milestone has SUMMARY and DB is complete", (t) => {
+  const repo = setupRepo();
+  t.after(() => teardownRepo(repo));
+
+  mkdirSync(join(repo, ".gsd", "milestones", "M002"), { recursive: true });
+  mkdirSync(join(repo, ".gsd", "milestones", "M003"), { recursive: true });
+
+  insertMilestone({ id: "M002", title: "Previous", status: "complete" });
+  insertSlice({ id: "S01", milestoneId: "M002", title: "Done", status: "complete", depends: [] });
+
+  insertMilestone({ id: "M003", title: "Current", status: "active" });
+  insertSlice({ id: "S01", milestoneId: "M003", title: "First", status: "pending", depends: [] });
+
+  writeFileSync(join(repo, ".gsd", "milestones", "M002", "M002-ROADMAP.md"), "# M002\n");
+  writeFileSync(join(repo, ".gsd", "milestones", "M003", "M003-ROADMAP.md"), "# M003\n");
+  writeFileSync(join(repo, ".gsd", "milestones", "M002", "M002-SUMMARY.md"), "# M002 Summary\nDone.\n");
+
+  assert.equal(
+    getPriorSliceCompletionBlocker(repo, "main", "plan-slice", "M003/S01"),
+    null,
+  );
+});

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

@@ -255,6 +255,31 @@ test("dispatch guard skips completed milestone with SUMMARY even if it has unche
   );
 });
 
+test("dispatch guard does not skip failed milestone SUMMARY without blocker prose", (t) => {
+  const repo = setupRepo();
+  t.after(() => teardownRepo(repo));
+
+  mkdirSync(join(repo, ".gsd", "milestones", "M001"), { recursive: true });
+  mkdirSync(join(repo, ".gsd", "milestones", "M002"), { recursive: true });
+
+  insertMilestone({ id: "M001", title: "Previous" });
+  insertSlice({ id: "S01", milestoneId: "M001", title: "Core", status: "complete", depends: [], sequence: 1 });
+  insertSlice({ id: "S02", milestoneId: "M001", title: "Unfinished", status: "pending", depends: ["S01"], sequence: 2 });
+
+  insertMilestone({ id: "M002", title: "Current" });
+  insertSlice({ id: "S01", milestoneId: "M002", title: "Start", status: "pending", depends: [], sequence: 1 });
+
+  writeFileSync(join(repo, ".gsd", "milestones", "M001", "M001-ROADMAP.md"), "# M001\n");
+  writeFileSync(join(repo, ".gsd", "milestones", "M001", "M001-SUMMARY.md"),
+    "---\nstatus: failed\n---\n# M001 Summary\nRecovery stopped.\n");
+  writeFileSync(join(repo, ".gsd", "milestones", "M002", "M002-ROADMAP.md"), "# M002\n");
+
+  assert.equal(
+    getPriorSliceCompletionBlocker(repo, "main", "plan-slice", "M002/S01"),
+    "Cannot dispatch plan-slice M002/S01: earlier slice M001/S02 is not complete.",
+  );
+});
+
 test("dispatch guard works without git repo", (t) => {
   const repo = setupRepo();
   t.after(() => teardownRepo(repo));

package/src/resources/extensions/gsd/tests/dispatch-missing-task-plans.test.ts

@@ -57,6 +57,17 @@ function scaffoldSlicePlan(basePath: string, mid: string, sid: string): void {
   ].join("\n"));
 }
 
+function scaffoldMilestoneContext(basePath: string, mid: string): void {
+  const dir = join(basePath, ".gsd", "milestones", mid);
+  mkdirSync(dir, { recursive: true });
+  writeFileSync(join(dir, `${mid}-CONTEXT.md`), [
+    `# ${mid}: Test Milestone`,
+    "",
+    "Context for dispatch recovery tests.",
+    "",
+  ].join("\n"));
+}
+
 function scaffoldTaskPlan(basePath: string, mid: string, sid: string, tid: string): void {
   const dir = join(basePath, ".gsd", "milestones", mid, "slices", sid, "tasks");
   mkdirSync(dir, { recursive: true });
@@ -76,6 +87,7 @@ test("dispatch: missing task plan triggers plan-slice (not stop) — issue #909"
   t.after(() => rmSync(tmp, { recursive: true, force: true }));
 
   // Slice plan exists with tasks, but tasks/ directory is empty
+  scaffoldMilestoneContext(tmp, "M002");
   scaffoldSlicePlan(tmp, "M002", "S03");
 
   const ctx = makeContext(tmp);
@@ -92,6 +104,7 @@ test("dispatch: present task plan proceeds to execute-task normally", async (t)
   const tmp = mkdtempSync(join(tmpdir(), "gsd-909-ok-"));
   t.after(() => rmSync(tmp, { recursive: true, force: true }));
 
+  scaffoldMilestoneContext(tmp, "M002");
   scaffoldSlicePlan(tmp, "M002", "S03");
   scaffoldTaskPlan(tmp, "M002", "S03", "T01");
 
@@ -111,6 +124,7 @@ test("dispatch: plan-slice recovery loop — second call after plan-slice still
   const tmp = mkdtempSync(join(tmpdir(), "gsd-909-loop-"));
   t.after(() => rmSync(tmp, { recursive: true, force: true }));
 
+  scaffoldMilestoneContext(tmp, "M002");
   scaffoldSlicePlan(tmp, "M002", "S03");
 
   const ctx = makeContext(tmp);

package/src/resources/extensions/gsd/tests/doctor-providers.test.ts

@@ -768,6 +768,37 @@ test("runProviderChecks detects claude.cmd in PATH on Windows (#4503)", { skip:
   });
 });
 
+test("runProviderChecks detects claude.exe in PATH on Windows (#4548)", { skip: process.platform !== "win32" }, () => {
+  const tmpHome = realpathSync(mkdtempSync(join(tmpdir(), "gsd-providers-cc-exe-home-")));
+  const binDir = join(tmpHome, "bin");
+  mkdirSync(binDir, { recursive: true });
+
+  // Some Windows installs ship a direct claude.exe binary (not a .cmd shim).
+  const fakeClaudeExe = join(binDir, "claude.exe");
+  writeFileSync(fakeClaudeExe, "");
+
+  withEnv({
+    HOME: tmpHome,
+    ANTHROPIC_API_KEY: undefined,
+    ANTHROPIC_OAUTH_TOKEN: undefined,
+    COPILOT_GITHUB_TOKEN: undefined,
+    GH_TOKEN: undefined,
+    GITHUB_TOKEN: undefined,
+    PATH: `${binDir};${process.env.PATH ?? ""}`,
+    PATHEXT: ".COM;.EXE;.BAT;.CMD",
+  }, () => {
+    try {
+      const results = runProviderChecks();
+      const anthropic = results.find(r => r.name === "anthropic");
+      assert.ok(anthropic, "anthropic result should exist");
+      assert.equal(anthropic!.status, "ok", "should be ok when claude.exe is in PATH (#4548)");
+      assert.ok(anthropic!.message.toLowerCase().includes("claude"), "should mention claude-code as source");
+    } finally {
+      rmSync(tmpHome, { recursive: true, force: true });
+    }
+  });
+});
+
 test("PROVIDER_ROUTES includes google-gemini-cli as route for google (#2922)", async () => {
   const { readFileSync: readFS } = await import("node:fs");
   const { dirname: dirn, join: joinPath } = await import("node:path");

package/src/resources/extensions/gsd/tests/double-merge-guard.test.ts

@@ -42,7 +42,7 @@ describe("double mergeAndExit guard (#2645)", () => {
     const allCompleteIdx = phasesSrc.indexOf("incomplete.length === 0");
     assert.ok(allCompleteIdx > 0, "phases.ts should have an all-milestones-complete check");
 
-    const afterAllComplete = phasesSrc.slice(allCompleteIdx, allCompleteIdx + 600);
+    const afterAllComplete = phasesSrc.slice(allCompleteIdx, allCompleteIdx + 800);
     const mergeIdx = afterAllComplete.indexOf("deps.resolver.mergeAndExit");
     const flagIdx = afterAllComplete.indexOf("s.milestoneMergedInPhases = true");
 

package/src/resources/extensions/gsd/tests/ensure-db-open.test.ts

@@ -389,7 +389,7 @@ describe('ensure-db-open', () => {
       assert.ok(db, 'adapter should be available after ensureDbOpen');
       assert.equal(
         db.prepare('SELECT MAX(version) as version FROM schema_version').get()?.version,
-        21,
+        22,
         'legacy DB should migrate to current schema version',
       );
 

package/src/resources/extensions/gsd/tests/escalation.test.ts

@@ -348,7 +348,7 @@ test("ADR-011 P2: schema v20 fresh DB has all escalation columns on tasks + sour
   assert.ok(decCols.includes("source"), "decisions table must have source column");
 
   const version = adapter.prepare("SELECT MAX(version) as v FROM schema_version").get();
-  assert.equal(version?.["v"], 21);
+  assert.equal(version?.["v"], 22);
 });
 
 test("ADR-011 P2: findUnappliedEscalationOverride returns null when escalation_pending=1 (still pending)", (t) => {