@opengsd/gsd-pi 1.1.1-dev.616a1a1 → 1.1.1-dev.9bb7453

package/src/resources/extensions/gsd/tests/validate-milestone-write-order.test.ts

@@ -180,6 +180,35 @@ describe("handleValidateMilestone write ordering (#2725)", () => {
     assert.equal(row, undefined, "assessment row should not be written when verification classes are invalid");
   });
 
+  it("reports all missing planned verification class rows at once", async () => {
+    base = makeTmpBase();
+    const dbPath = join(base, ".gsd", "gsd.db");
+    openDatabase(dbPath);
+    insertMilestone({
+      id: "M001",
+      planning: {
+        verificationContract: "Contract command exits 0",
+        verificationOperational: "Process lifecycle proof",
+        verificationUat: "Browser-observable UAT proof",
+      },
+    });
+    insertSlice({ id: "S01", milestoneId: "M001" });
+
+    const result = await handleValidateMilestone(
+      { ...VALID_PARAMS, verificationClasses: "| Check | Result |\n| --- | --- |\n| Generic verification | PASS |" },
+      base,
+    );
+    assert.ok("error" in result, "expected validation to fail");
+    assert.match(result.error, /canonical rows "Contract", "Operational", "UAT"/);
+    assert.match(result.error, /planned contract, operational, uat verification/);
+
+    const adapter = _getAdapter()!;
+    const row = adapter.prepare(
+      `SELECT status FROM assessments WHERE milestone_id = 'M001' AND scope = 'milestone-validation'`,
+    ).get() as { status: string } | undefined;
+    assert.equal(row, undefined, "assessment row should not be written when verification classes are invalid");
+  });
+
   it("accepts verificationClasses when planned Operational class is present", async () => {
     base = makeTmpBase();
     const dbPath = join(base, ".gsd", "gsd.db");
@@ -406,6 +435,110 @@ describe("handleValidateMilestone write ordering (#2725)", () => {
     assert.equal(result.verdict, "pass");
   });
 
+  it("keeps pass when browser-like criteria are verified by runtime-executable UAT", async () => {
+    base = makeTmpBase();
+    const dbPath = join(base, ".gsd", "gsd.db");
+    openDatabase(dbPath);
+    insertMilestone({
+      id: "M001",
+      planning: {
+        successCriteria: [
+          "Clicking Mark All Complete sets all todos completed",
+          "Reload keeps completed state",
+        ],
+        verificationUat: "Run the Node.js DOM-state script against the static app source.",
+      },
+    });
+    insertSlice({
+      id: "S01",
+      milestoneId: "M001",
+      // Uses localhost so hasBrowserRequiredText returns true and the gate is
+      // actually triggered before the runtime evidence bypasses it.
+      demo: "Visit localhost:3000 to verify DOM state after clicking Mark All Complete.",
+    });
+    const sliceDir = join(base, ".gsd", "milestones", "M001", "slices", "S01");
+    mkdirSync(sliceDir, { recursive: true });
+    writeFileSync(
+      join(sliceDir, "S01-ASSESSMENT.md"),
+      [
+        "---",
+        "sliceId: S01",
+        "uatType: runtime-executable",
+        "verdict: PASS",
+        "attempt: 1",
+        "---",
+        "# UAT Result - S01",
+        "",
+        "## Checks",
+        "",
+        "| Check | Mode | Result | Evidence | Notes |",
+        "|-------|------|--------|----------|-------|",
+        "| DOM-state script | runtime | PASS | gsd_uat_exec:.gsd/evidence/uat/M001/S01/dom-state.json | Runtime assertion verified completed state and reload persistence. |",
+        "",
+      ].join("\n"),
+      "utf-8",
+    );
+
+    const result = await handleValidateMilestone(
+      {
+        ...VALID_PARAMS,
+        verificationClasses:
+          `${VALID_PARAMS.verificationClasses}\n| UAT | Runtime executable UAT verified static-app behavior. |`,
+      },
+      base,
+    );
+
+    assert.ok(!("error" in result), `unexpected error: ${"error" in result ? result.error : ""}`);
+    assert.equal(result.verdict, "pass");
+  });
+
+  it("downgrades to needs-attention when only one of two browser-requiring slices has runtime evidence", async () => {
+    base = makeTmpBase();
+    const dbPath = join(base, ".gsd", "gsd.db");
+    openDatabase(dbPath);
+    insertMilestone({ id: "M001" });
+    insertSlice({
+      id: "S01",
+      milestoneId: "M001",
+      demo: "Visit localhost:3000 to verify DOM state.",
+    });
+    insertSlice({
+      id: "S02",
+      milestoneId: "M001",
+      demo: "Visit localhost:3000 to confirm persistence after reload.",
+    });
+    // S01 has runtime-executable evidence; S02 has none.
+    mkdirSync(join(base, ".gsd", "milestones", "M001", "slices", "S01"), { recursive: true });
+    writeFileSync(
+      join(base, ".gsd", "milestones", "M001", "slices", "S01", "S01-ASSESSMENT.md"),
+      [
+        "---",
+        "sliceId: S01",
+        "uatType: runtime-executable",
+        "verdict: PASS",
+        "---",
+        "| DOM check | runtime | PASS | gsd_uat_exec:.gsd/evidence/uat/M001/S01/dom.json | Verified. |",
+        "",
+      ].join("\n"),
+      "utf-8",
+    );
+
+    const result = await handleValidateMilestone(
+      {
+        ...VALID_PARAMS,
+        verificationClasses: `${VALID_PARAMS.verificationClasses}\n| UAT | S01 runtime verified; S02 still needs evidence. |`,
+      },
+      base,
+    );
+
+    assert.ok(!("error" in result), `unexpected error: ${"error" in result ? result.error : ""}`);
+    assert.equal(
+      result.verdict,
+      "needs-attention",
+      "S01 runtime evidence must not bypass the gate for S02 which has browser requirements but no evidence",
+    );
+  });
+
   it("ignores slice full_uat_md planning text for browser requirement detection", async () => {
     base = makeTmpBase();
     const dbPath = join(base, ".gsd", "gsd.db");

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

@@ -54,6 +54,8 @@ test("validation block allows recovery, diagnostics, and unrelated commands", ()
     "status",
     "verdict pass --rationale ok",
     "validate-milestone",
+    "dispatch reassess",
+    "dispatch reassess-roadmap",
     "dispatch validate",
     "dispatch validate-milestone",
     "park M006",
@@ -122,3 +124,22 @@ test("validation block message includes attempted command and recovery options",
   assert.match(message, /\/gsd verdict pass --rationale/);
   assert.match(message, /\/gsd park M006/);
 });
+
+test("validation block message can guide remediation through dispatch reassess", () => {
+  const message = formatValidationBlockedMessage({
+    ...blockedState(),
+    blockers: [
+      [
+        "Milestone M006 is blocked because milestone validation returned needs-remediation, but all slices are complete.",
+        "Fix options:",
+        "1. Run `/gsd dispatch reassess` to add remediation slices, then run `/gsd auto`",
+        "2. If the finding is acceptable, override it: `/gsd verdict pass --rationale \"why this is okay\"`",
+        "3. If this should wait, defer it explicitly: `/gsd park M006`",
+      ].join("\n"),
+    ],
+  }, "auto");
+
+  assert.ok(message);
+  assert.match(message, /\/gsd dispatch reassess/);
+  assert.doesNotMatch(message, /gsd_reassess_roadmap/);
+});

package/src/resources/extensions/gsd/tests/verification-gate.test.ts

@@ -35,6 +35,20 @@ function makeTempDir(prefix: string): string {
   return dir;
 }
 
+function withRtkDisabled<T>(callback: () => T): T {
+  const previous = process.env.GSD_RTK_DISABLED;
+  process.env.GSD_RTK_DISABLED = "1";
+  try {
+    return callback();
+  } finally {
+    if (previous === undefined) {
+      delete process.env.GSD_RTK_DISABLED;
+    } else {
+      process.env.GSD_RTK_DISABLED = previous;
+    }
+  }
+}
+
 // ─── Discovery Tests ─────────────────────────────────────────────────────────
 
 describe("verification-gate: discovery", () => {
@@ -430,6 +444,38 @@ describe("verification-gate: execution", () => {
     assert.ok(result.checks[1].stderr.includes("err"));
   });
 
+  test("grep -c zero-match failure includes absence-check warning", () => {
+    writeFileSync(join(tmp, "sample.txt"), "present\n");
+
+    const result = withRtkDisabled(() => runVerificationGate({
+      cwd: tmp,
+      preferenceCommands: ["grep -c missing sample.txt"],
+    }));
+
+    assert.equal(result.passed, false);
+    assert.equal(result.checks.length, 1);
+    assert.equal(result.checks[0].exitCode, 1);
+    assert.equal(result.checks[0].stdout.trim(), "0");
+    assert.match(result.checks[0].stderr, /grep -c/);
+    assert.match(result.checks[0].stderr, /count=0/);
+    assert.match(result.checks[0].stderr, /! grep -q/);
+  });
+
+  test("grep -c matching count does not warn", () => {
+    writeFileSync(join(tmp, "sample.txt"), "present\n");
+
+    const result = withRtkDisabled(() => runVerificationGate({
+      cwd: tmp,
+      preferenceCommands: ["grep -c present sample.txt"],
+    }));
+
+    assert.equal(result.passed, true);
+    assert.equal(result.checks.length, 1);
+    assert.equal(result.checks[0].exitCode, 0);
+    assert.equal(result.checks[0].stdout.trim(), "1");
+    assert.equal(result.checks[0].stderr, "");
+  });
+
   test("no commands discovered → gate passes with 0 checks", () => {
     const result = runVerificationGate({
       cwd: tmp,
@@ -686,6 +732,11 @@ test("isLikelyCommand: bash negation with known command is accepted", () => {
   assert.equal(isLikelyCommand("! grep needle file.txt"), true);
 });
 
+test("validateVerificationCommand accepts negated quiet absence checks", () => {
+  assert.equal(validateVerificationCommand("! grep -q needle file.txt").ok, true);
+  assert.equal(validateVerificationCommand("! rg -q needle file.txt").ok, true);
+});
+
 test("validateVerificationCommand allows shell pipelines", () => {
   assert.deepEqual(validateVerificationCommand("python3 -m pytest tests/ -q --tb=short").ok, true);
   const result = validateVerificationCommand("python3 -m pytest tests/ -q --tb=short | tail -5");

package/src/resources/extensions/gsd/tests/workflow-mcp-auto-prep.test.ts

@@ -19,6 +19,14 @@ test("shouldAutoPrepareWorkflowMcp enables prep for externalCli local transport"
   assert.equal(result, true);
 });
 
+test("shouldAutoPrepareWorkflowMcp enables prep when Claude Code provider is known before auth mode settles", () => {
+  const result = shouldAutoPrepareWorkflowMcp({
+    model: { provider: "claude-code", baseUrl: "local://claude-code" },
+  });
+
+  assert.equal(result, true);
+});
+
 test("prepareWorkflowMcpForProject uses the selected unit model when session provider differs", (t) => {
   const projectRoot = mkdtempSync(join(tmpdir(), "gsd-mcp-unit-model-"));
   const notifications: Array<{ message: string; level: "info" | "warning" | "error" | "success" }> = [];
@@ -46,7 +54,7 @@ test("prepareWorkflowMcpForProject uses the selected unit model when session pro
 
   assert.equal(result?.status, "created");
   assert.equal(existsSync(join(projectRoot, ".mcp.json")), true);
-  assert.match(notifications.map((entry) => entry.message).join("\n"), /Claude Code MCP prepared/);
+  assert.match(notifications.map((entry) => entry.message).join("\n"), /GSD MCP Server Prepared/);
 });
 
 test("shouldAutoPrepareWorkflowMcp stays disabled for non-Claude active provider even when claude-code is ready", () => {
@@ -163,7 +171,14 @@ test("before_agent_start auto-prepares project workflow MCP for Claude Code CLI"
   };
   assert.ok(parsed.mcpServers?.[GSD_WORKFLOW_MCP_SERVER_NAME]);
   assert.ok(parsed.mcpServers?.[GSD_BROWSER_MCP_SERVER_NAME]);
-  assert.match(notifications.join("\n"), /Claude Code MCP prepared/);
+  const settings = JSON.parse(readFileSync(join(projectRoot, ".claude", "settings.local.json"), "utf-8")) as {
+    enabledMcpjsonServers?: string[];
+  };
+  assert.deepEqual(settings.enabledMcpjsonServers, [
+    GSD_WORKFLOW_MCP_SERVER_NAME,
+    GSD_BROWSER_MCP_SERVER_NAME,
+  ]);
+  assert.match(notifications.join("\n"), /GSD MCP Server Prepared/);
 });
 
 test("before_agent_start returns discovered skill fallback without project .gsd", async (t) => {

package/src/resources/extensions/gsd/tests/workflow-tool-executors.test.ts

@@ -30,6 +30,7 @@ import {
   executeSliceComplete,
   executeSliceReopen,
   executeValidateMilestone,
+  executeUatResultSave,
 } from "../tools/workflow-tool-executors.ts";
 
 function makeTmpBase(): string {
@@ -234,6 +235,62 @@ test("executeTaskComplete derives missing verification from evidence", async ()
   }
 });
 
+test("executeTaskComplete surfaces escalation questions and metadata", async () => {
+  const base = makeTmpBase();
+  try {
+    openTestDb(base);
+    writeFileSync(join(base, ".gsd", "PREFERENCES.md"), [
+      "---",
+      "version: 1",
+      "phases:",
+      "  mid_execution_escalation: true",
+      "---",
+    ].join("\n"));
+    const planDir = join(base, ".gsd", "milestones", "M001", "slices", "S01");
+    mkdirSync(planDir, { recursive: true });
+    writeFileSync(join(planDir, "S01-PLAN.md"), "# S01\n\n- [ ] **T01: Demo** `est:5m`\n");
+
+    const result = await inProjectDir(base, () => executeTaskComplete({
+      milestoneId: "M001",
+      sliceId: "S01",
+      taskId: "T01",
+      oneLiner: "Completed task",
+      narrative: "Did the work but found an ambiguity.",
+      verification: "npm test",
+      escalation: {
+        question: "Should the cache use write-through or write-back?",
+        options: [
+          { id: "A", label: "Write-through", tradeoffs: "Simpler reads; slower writes." },
+          { id: "B", label: "Write-back", tradeoffs: "Faster writes; more flush complexity." },
+        ],
+        recommendation: "A",
+        recommendationRationale: "Current usage favors correctness over write latency.",
+        continueWithDefault: true,
+      },
+    }, base));
+
+    assert.equal(result.details.operation, "complete_task");
+    assert.match(
+      String(result.content[0]?.text),
+      /Task completed with escalation decision required: Should the cache use write-through or write-back\?/,
+    );
+    assert.match(String(result.content[0]?.text), /Resolve with: \/gsd escalate resolve T01/);
+    assert.equal((result.details.escalation as { question?: string }).question, "Should the cache use write-through or write-back?");
+
+    const db = _getAdapter();
+    assert.ok(db, "DB should be open");
+    const row = db!.prepare(
+      "SELECT escalation_pending, escalation_awaiting_review, escalation_artifact_path FROM tasks WHERE milestone_id = ? AND slice_id = ? AND id = ?",
+    ).get("M001", "S01", "T01") as Record<string, unknown> | undefined;
+    assert.equal(row?.escalation_pending, 0);
+    assert.equal(row?.escalation_awaiting_review, 1);
+    assert.ok(String(row?.escalation_artifact_path ?? "").endsWith("T01-ESCALATION.json"));
+  } finally {
+    closeDatabase();
+    cleanup(base);
+  }
+});
+
 test("executeTaskComplete returns a tool error when verification cannot be derived", async () => {
   const base = makeTmpBase();
   try {
@@ -504,6 +561,162 @@ test("executePlanSlice marks validation failures with isError", async () => {
   }
 });
 
+test("executeUatResultSave accepts gsd_uat_exec evidence written in a milestone worktree", async () => {
+  const base = makeTmpBase();
+  const worktree = join(base, ".gsd", "worktrees", "M001");
+  const worktreeExecDir = join(worktree, ".gsd", "exec");
+  const browserTimelineDir = join(base, ".artifacts", "browser", "session");
+  const evidenceId = "worktree-uat-evidence";
+  const browserTimelinePath = join(browserTimelineDir, "s02-uat-browser-timeline.json");
+  try {
+    openTestDb(base);
+    seedMilestone("M001", "Milestone One");
+    seedSlice("M001", "S02", "complete");
+    mkdirSync(worktreeExecDir, { recursive: true });
+    mkdirSync(browserTimelineDir, { recursive: true });
+    writeFileSync(browserTimelinePath, JSON.stringify({ summary: "browser timeline evidence" }), "utf-8");
+    writeFileSync(
+      join(worktreeExecDir, `${evidenceId}.meta.json`),
+      JSON.stringify({
+        id: evidenceId,
+        metadata: {
+          kind: "uat_exec",
+          milestoneId: "M001",
+          sliceId: "S02",
+          checkId: "UAT-01",
+          intent: "uat-runtime-check",
+        },
+      }),
+      "utf-8",
+    );
+
+    const result = await inProjectDir(worktree, () => executeUatResultSave({
+      milestoneId: "M001",
+      sliceId: "S02",
+      uatType: "runtime-executable",
+      verdict: "PASS",
+      checks: [{
+        id: "UAT-01",
+        description: "Runtime path C:\\tmp|uat evidence was captured in the active worktree",
+        mode: "runtime",
+        result: "PASS",
+        evidence: [
+          { kind: "gsd_uat_exec", ref: evidenceId },
+          { kind: "browser", ref: browserTimelinePath },
+        ],
+        notes: "Worktree-local gsd_uat_exec metadata should resolve with backslash \\ and pipe |.",
+      }],
+      presentation: {
+        surface: "mcp",
+        presentedTools: [
+          "gsd_uat_exec",
+          "gsd_uat_result_save",
+          "gsd_resume",
+          "gsd_milestone_status",
+          "gsd_journal_query",
+        ],
+        blockedTools: [
+          { name: "gsd_exec", reason: "forbidden during run-uat" },
+          { name: "gsd_summary_save", reason: "forbidden during run-uat" },
+          { name: "gsd_save_gate_result", reason: "forbidden during run-uat" },
+        ],
+      },
+      notes: "UAT passed with worktree-local evidence.",
+    }, worktree));
+
+    assert.equal(result.isError, undefined);
+    assert.equal(result.details.operation, "save_uat_result");
+    assert.equal(result.details.verdict, "PASS");
+    assert.ok(
+      existsSync(join(base, ".gsd", "uat", "M001", "S02", "attempt-1.json")),
+      "attempt JSON should be persisted under the authoritative project .gsd",
+    );
+    const assessment = readFileSync(
+      join(base, ".gsd", "milestones", "M001", "slices", "S02", "S02-ASSESSMENT.md"),
+      "utf-8",
+    );
+    assert.match(assessment, /Runtime path C:\\\\tmp\\\|uat evidence/);
+    assert.match(assessment, /backslash \\\\ and pipe \\\|/);
+  } finally {
+    closeDatabase();
+    cleanup(base);
+  }
+});
+
+test("executeUatResultSave rejects artifact-driven PASS with human follow-up checks", async () => {
+  const base = makeTmpBase();
+  const worktree = join(base, ".gsd", "worktrees", "M001");
+  const evidenceId = "uat-artifact-nonautomatable";
+  const worktreeExecDir = join(worktree, ".gsd", "exec");
+  try {
+    openTestDb(base);
+    seedMilestone("M001", "Milestone One");
+    seedSlice("M001", "S01", "complete");
+    mkdirSync(worktreeExecDir, { recursive: true });
+    writeFileSync(
+      join(worktreeExecDir, `${evidenceId}.meta.json`),
+      JSON.stringify({
+        id: evidenceId,
+        metadata: {
+          kind: "uat_exec",
+          milestoneId: "M001",
+          sliceId: "S01",
+          checkId: "UAT-01",
+          intent: "uat-artifact-check",
+        },
+      }),
+      "utf-8",
+    );
+
+    const result = await inProjectDir(worktree, () => executeUatResultSave({
+      milestoneId: "M001",
+      sliceId: "S01",
+      uatType: "artifact-driven",
+      verdict: "PASS",
+      checks: [
+        {
+          id: "UAT-01",
+          description: "Static contract passes",
+          mode: "artifact",
+          result: "PASS",
+          evidence: [{ kind: "gsd_uat_exec", ref: evidenceId }],
+          notes: "Artifact check passed.",
+        },
+        {
+          id: "UAT-02",
+          description: "Browser polish is deferred to the next slice",
+          mode: "human-follow-up",
+          result: "NEEDS-HUMAN",
+          notes: "Out of scope for this artifact-driven UAT.",
+          nonAutomatable: true,
+        },
+      ],
+      presentation: {
+        surface: "mcp",
+        presentedTools: [
+          "gsd_uat_exec",
+          "gsd_uat_result_save",
+          "gsd_resume",
+          "gsd_milestone_status",
+          "gsd_journal_query",
+        ],
+        blockedTools: [
+          { name: "gsd_exec", reason: "forbidden during run-uat" },
+          { name: "gsd_summary_save", reason: "forbidden during run-uat" },
+          { name: "gsd_save_gate_result", reason: "forbidden during run-uat" },
+        ],
+      },
+      notes: "UAT passed; non-automatable browser polish is deferred.",
+    }, worktree));
+
+    assert.equal(result.isError, true);
+    assert.match(String(result.content[0]?.text), /artifact-driven UAT cannot PASS with human-only checks/);
+  } finally {
+    closeDatabase();
+    cleanup(base);
+  }
+});
+
 test("executeSliceComplete coerces string enrichment entries and writes summary/UAT artifacts", async () => {
   const base = makeTmpBase();
   try {

package/src/resources/extensions/gsd/tests/write-gate-planning-unit.test.ts

@@ -27,6 +27,10 @@ const PLANNING_DISPATCH_REVIEW: ToolsPolicy = {
 const READ_ONLY: ToolsPolicy = { mode: 'read-only' };
 const ALL: ToolsPolicy = { mode: 'all' };
 const VERIFICATION: ToolsPolicy = { mode: 'verification' };
+const VERIFICATION_UAT: ToolsPolicy = {
+  mode: 'verification',
+  allowedSubagents: ['mnemo', 'scout', 'reviewer', 'tester'],
+};
 const DOCS: ToolsPolicy = {
   mode: 'docs',
   allowedPathGlobs: ['docs/**', 'README.md', 'README.*.md', 'CHANGELOG.md', '*.md'],
@@ -469,6 +473,27 @@ test('verification-mode: run-uat still blocks subagent dispatch', () => {
   assert.match(r.reason!, /subagent dispatch is not permitted/);
 });
 
+test('verification-mode: run-uat allows explicit UAT specialist subagents', () => {
+  for (const agent of ['mnemo', 'scout', 'reviewer', 'tester']) {
+    const r = shouldBlockPlanningUnit('subagent', '', BASE, 'run-uat', VERIFICATION_UAT, [agent]);
+    assert.strictEqual(r.block, false, `expected ${agent} to be allowed: ${r.reason}`);
+  }
+});
+
+test('verification-mode: run-uat blocks implementation-tier subagents', () => {
+  const r = shouldBlockPlanningUnit('subagent', '', BASE, 'run-uat', VERIFICATION_UAT, ['worker']);
+  assert.strictEqual(r.block, true);
+  assert.match(r.reason!, /"worker"/);
+  assert.match(r.reason!, /read-only specialists/);
+});
+
+test('verification-mode: run-uat blocks read-only specialists not listed by policy', () => {
+  const r = shouldBlockPlanningUnit('subagent', '', BASE, 'run-uat', VERIFICATION_UAT, ['security']);
+  assert.strictEqual(r.block, true);
+  assert.match(r.reason!, /"security"/);
+  assert.match(r.reason!, /ToolsPolicy\.allowedSubagents|permitted agents for this unit/);
+});
+
 // ─── read-only mode ───────────────────────────────────────────────────────
 
 test('read-only: blocks any edit even to .gsd/', () => {