fix(macp): fail-closed typed gate states — no placeholder/empty-command passes (#1275 )

RI-N2 / SDLC-D-035. Every GateResult now carries a typed status discriminator (passed|failed|simulated|waiting|capability_failure); passed:true remains true only for really-executed, really-green gates. - ci-pipeline without a CI provider → capability_failure (MACP_NO_CI_PIPELINE), never the placeholder pass - empty-command gate → capability_failure (MACP_NO_COMMAND / MACP_NO_REVIEWER), and runGates no longer silently skips it - manual gate type waits (MACP_AUTHORITY_REQUIRED) — neither pass nor fail - explicit simulation only (simulate option / --simulate): typed simulated results can never make the aggregate passed (RunGatesResult.state) - CLI stubs (tasks list, submit, events tail) exit nonzero with typed MACP_NOT_IMPLEMENTED; macp gate is now wired to runGates with --simulate - legacy __tests__/gate-runner.test.ts consolidated into src/gate-runner.spec.ts (root eslint project service does not cover packages/macp/__tests__)
Merge pull request 'fix(ci): unwire test-start-agent-session.sh, restore its signed exclusion — unblocks every PR on next' (#1270 ) from fix/1269-ci-chain-unblock into next
2026-08-17 16:40:16 -05:00 · 2026-08-17 20:44:59 +00:00 · 2026-08-16 18:03:02 -05:00 · 2026-08-16 17:58:49 -05:00
24 changed files with 1180 additions and 1695 deletions
@@ -34,7 +34,6 @@ export default tseslint.config(
            'packages/storage/vitest.config.ts',
            'packages/mosaic/vitest.config.ts',
            'packages/mosaic/__tests__/*.ts',
-            'packages/forge/__tests__/*.ts',
            'tools/federation-harness/*.ts',
          ],
        },
@@ -539,43 +539,3 @@ Not every brief needs full Board of Directors review. The classification system
 ### Backward compatibility

 Existing briefs without a `class` field are auto-classified. The default (no matching keywords) is `strategic`, so all existing runs get the full pipeline unless keywords trigger `technical`.
-
---
-
-## Fail-Closed Execution & Explicit Simulation (SDLC-D-035)
-
-**Added:** 2026-08-17
-
-Forge fails closed when a required capability is missing. It never runs a
-pipeline with a stub executor and reports success.
-
-### Normal mode (default)
-
- No task executor wired → the CLI exits nonzero with the typed capability
-  error `FORGE_NO_EXECUTOR`. No run is created.
- A stage whose gate is approval-based (board approval, planning approvals,
-  remediation re-review, discovery/analysis attestations) records a typed
-  `waiting-for-authority` stage result and raises `FORGE_AUTHORITY_REQUIRED`.
-  It never passes vacuously.
- A stage whose gate requires an unwired provider (AI reviewer, CI pipeline)
-  records a typed `blocked` stage result and raises `FORGE_NO_REVIEWER` /
-  `FORGE_NO_CI_PIPELINE`. The synthetic echo-review approval in `06-review`
-  and all vacuous `true` gates were removed.
-
-### Explicit simulation (`--simulate`)
-
-Opts into stub/synthetic execution. Every stage result, every gate result, and
-the run manifest carry the distinct typed status `simulated` (manifest also
-records `mode: "simulated"`). `simulated` is a non-satisfying outcome:
-`isSatisfyingOutcome()` and all completion/gate consumers treat only `passed`
-as satisfying. The CLI exits 0 for a simulated run only because the caller
-explicitly passed `--simulate`, and prints a loud SIMULATED banner.
-
-### Typed outcome model
-
-Every gate/task outcome is one of the closed set
-`passed | failed | blocked | error | waiting-for-authority | simulated |
-not-applicable`, with the reason recorded on the stage status and each gate
-result in `manifest.json`. Missing implementations, missing gate evidence,
-unknown stages, process errors, and timeouts map to fail-closed members —
-never to `passed`.
@@ -1,319 +0,0 @@
-import fs from 'node:fs';
-import os from 'node:os';
-import path from 'node:path';
-import { describe, it, expect, beforeEach, afterEach } from 'vitest';
-
-import { generateBoardTasks } from '../src/board-tasks.js';
-import { STAGE_SPECS } from '../src/constants.js';
-import { ForgeCapabilityError } from '../src/errors.js';
-import {
-  evaluateStageGates,
-  gateLabel,
-  isCommandGate,
-  isSatisfyingOutcome,
-} from '../src/outcomes.js';
-import { loadManifest, runPipeline } from '../src/pipeline-runner.js';
-import type { ForgeTask, ForgeTaskResult, TaskExecutor } from '../src/types.js';
-
-/**
- * Mock real executor that returns typed results.
- *
- * Command gates are "verified" by the mock so normal-mode runs can pass
- * mechanically gated stages; authority/provider gates are never reported
- * because they have no mechanical implementation.
- */
-function createTypedExecutor(options?: {
-  failStage?: string;
-  gateOutcomes?: Record<string, 'passed' | 'failed' | 'simulated' | 'error' | 'blocked'>;
-}): TaskExecutor & { submittedTasks: ForgeTask[] } {
-  const submittedTasks: ForgeTask[] = [];
-  return {
-    submittedTasks,
-    async submitTask(task: ForgeTask) {
-      submittedTasks.push(task);
-    },
-    async waitForCompletion(taskId: string): Promise<ForgeTaskResult> {
-      const task = submittedTasks.find((t) => t.id === taskId);
-      const stageName = task?.metadata?.['stageName'] as string | undefined;
-
-      if (options?.failStage && stageName === options.failStage) {
-        return {
-          task_id: taskId,
-          outcome: 'failed',
-          reason: 'mock task failure',
-          completed_at: new Date().toISOString(),
-          exit_code: 1,
-          gate_results: [],
-        };
-      }
-
-      const gateResults = (task?.qualityGates ?? [])
-        .filter((gate) => isCommandGate(gate))
-        .map((gate) => {
-          const label = gateLabel(gate);
-          const outcome = options?.gateOutcomes?.[label] ?? 'passed';
-          return {
-            gate: label,
-            outcome,
-            reason: outcome === 'passed' ? 'mock verified' : `mock gate outcome: ${outcome}`,
-          };
-        });
-
-      return {
-        task_id: taskId,
-        outcome: 'passed',
-        reason: 'mock verified',
-        completed_at: new Date().toISOString(),
-        exit_code: 0,
-        gate_results: gateResults,
-      };
-    },
-    async getTaskStatus() {
-      return 'completed' as const;
-    },
-  };
-}
-
-describe('fail-closed: no executor wired', () => {
-  let tmpDir: string;
-  let briefPath: string;
-
-  beforeEach(() => {
-    tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'forge-failclosed-'));
-    briefPath = path.join(tmpDir, 'brief.md');
-    fs.writeFileSync(briefPath, '# Fix bug\n\nA bugfix for lint cleanup.');
-  });
-
-  afterEach(() => {
-    fs.rmSync(tmpDir, { recursive: true, force: true });
-  });
-
-  it('throws a typed FORGE_NO_EXECUTOR capability error without --simulate', async () => {
-    await expect(
-      runPipeline(briefPath, tmpDir, {
-        // no executor, no simulate — must fail closed, never run with a stub
-        stages: ['00-intake'],
-      }),
-    ).rejects.toMatchObject({
-      name: 'ForgeCapabilityError',
-      code: 'FORGE_NO_EXECUTOR',
-      capability: 'task-executor',
-    });
-  });
-
-  it('does not create a run directory when failing closed on a missing executor', async () => {
-    try {
-      await runPipeline(briefPath, tmpDir, { stages: ['00-intake'] });
-    } catch {
-      // expected
-    }
-    expect(fs.existsSync(path.join(tmpDir, '.forge', 'runs'))).toBe(false);
-  });
-
-  it('completes with every result typed simulated when simulate is set', async () => {
-    const result = await runPipeline(briefPath, tmpDir, {
-      simulate: true,
-      stages: ['00-intake', '00b-discovery', '02-planning-1', '06-review'],
-    });
-
-    expect(result.manifest.mode).toBe('simulated');
-    expect(result.manifest.status).toBe('simulated');
-
-    for (const stage of result.stages) {
-      const stageStatus = result.manifest.stages[stage];
-      expect(stageStatus?.status, `stage ${stage}`).toBe('simulated');
-      expect(stageStatus?.status, `stage ${stage}`).not.toBe('passed');
-      expect(stageStatus?.reason, `stage ${stage}`).toBeTruthy();
-      for (const gateResult of stageStatus?.gateResults ?? []) {
-        expect(gateResult.outcome, `gate ${gateResult.gate} of ${stage}`).toBe('simulated');
-        expect(gateResult.outcome, `gate ${gateResult.gate} of ${stage}`).not.toBe('passed');
-      }
-    }
-
-    // The persisted manifest agrees.
-    const persisted = loadManifest(result.runDir);
-    expect(persisted.mode).toBe('simulated');
-    expect(persisted.status).toBe('simulated');
-    expect(persisted.stages['02-planning-1']?.status).toBe('simulated');
-  });
-});
-
-describe('fail-closed: typed outcome model', () => {
-  it('only passed satisfies the gate/dependency predicate', () => {
-    expect(isSatisfyingOutcome('passed')).toBe(true);
-    expect(isSatisfyingOutcome('failed')).toBe(false);
-    expect(isSatisfyingOutcome('blocked')).toBe(false);
-    expect(isSatisfyingOutcome('error')).toBe(false);
-    expect(isSatisfyingOutcome('waiting-for-authority')).toBe(false);
-    expect(isSatisfyingOutcome('simulated')).toBe(false);
-    expect(isSatisfyingOutcome('not-applicable')).toBe(false);
-  });
-
-  it('a simulated gate result cannot satisfy the stage gate evaluation', () => {
-    const evaluation = evaluateStageGates('05-coding', STAGE_SPECS['05-coding']!.qualityGates, {
-      task_id: 'FORGE-x-05',
-      outcome: 'passed',
-      reason: 'executor claims success',
-      completed_at: new Date().toISOString(),
-      exit_code: 0,
-      gate_results: [{ gate: 'pnpm lint', outcome: 'simulated', reason: 'simulated gate' }],
-    });
-    expect(isSatisfyingOutcome(evaluation.outcome)).toBe(false);
-    expect(evaluation.outcome).toBe('error');
-  });
-
-  it('a simulated task outcome cannot satisfy evaluation in normal mode', () => {
-    const evaluation = evaluateStageGates('00-intake', [], {
-      task_id: 'FORGE-x-00',
-      outcome: 'simulated',
-      reason: 'executor reported simulated',
-      completed_at: new Date().toISOString(),
-      exit_code: 0,
-      gate_results: [],
-    });
-    expect(isSatisfyingOutcome(evaluation.outcome)).toBe(false);
-  });
-
-  it('a missing gate result blocks the stage instead of passing vacuously', () => {
-    const evaluation = evaluateStageGates('05-coding', STAGE_SPECS['05-coding']!.qualityGates, {
-      task_id: 'FORGE-x-05',
-      outcome: 'passed',
-      reason: 'executor claims success',
-      completed_at: new Date().toISOString(),
-      exit_code: 0,
-      gate_results: [],
-    });
-    expect(evaluation.outcome).toBe('blocked');
-  });
-});
-
-describe('fail-closed: authority and provider gates', () => {
-  let tmpDir: string;
-  let briefPath: string;
-
-  beforeEach(() => {
-    tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'forge-authority-'));
-    briefPath = path.join(tmpDir, 'brief.md');
-    fs.writeFileSync(briefPath, '# Fix bug\n\nA bugfix for lint cleanup.');
-  });
-
-  afterEach(() => {
-    fs.rmSync(tmpDir, { recursive: true, force: true });
-  });
-
-  it.each(['02-planning-1', '03-planning-2', '04-planning-3', '07-remediate'])(
-    'planning/remediation stage %s yields waiting-for-authority (not passed) in normal mode',
-    async (stage) => {
-      const executor = createTypedExecutor();
-      let runDir: string | undefined;
-
-      try {
-        await runPipeline(briefPath, tmpDir, {
-          executor,
-          stages: [stage as string],
-        });
-        expect.unreachable('runPipeline should have failed closed');
-      } catch (err) {
-        expect(err).toBeInstanceOf(ForgeCapabilityError);
-        expect((err as ForgeCapabilityError).code).toBe('FORGE_AUTHORITY_REQUIRED');
-        runDir = path.join(tmpDir, '.forge', 'runs');
-      }
-
-      const runIds = fs.readdirSync(runDir!);
-      expect(runIds).toHaveLength(1);
-      const manifest = loadManifest(path.join(runDir!, runIds[0]!));
-      expect(manifest.stages[stage]?.status).toBe('waiting-for-authority');
-      expect(manifest.stages[stage]?.status).not.toBe('passed');
-      expect(manifest.status).toBe('waiting-for-authority');
-    },
-  );
-
-  it('review stage fails closed with a typed FORGE_NO_REVIEWER error in normal mode', async () => {
-    const executor = createTypedExecutor();
-
-    try {
-      await runPipeline(briefPath, tmpDir, {
-        executor,
-        stages: ['06-review'],
-      });
-      expect.unreachable('runPipeline should have failed closed');
-    } catch (err) {
-      expect(err).toBeInstanceOf(ForgeCapabilityError);
-      expect((err as ForgeCapabilityError).code).toBe('FORGE_NO_REVIEWER');
-      expect((err as ForgeCapabilityError).capability).toBe('reviewer');
-    }
-
-    const runsDir = path.join(tmpDir, '.forge', 'runs');
-    const runIds = fs.readdirSync(runsDir);
-    const manifest = loadManifest(path.join(runsDir, runIds[0]!));
-    expect(manifest.stages['06-review']?.status).toBe('blocked');
-    expect(manifest.stages['06-review']?.status).not.toBe('passed');
-    expect(manifest.status).toBe('failed');
-  });
-
-  it('review stage produces simulated results under --simulate', async () => {
-    const result = await runPipeline(briefPath, tmpDir, {
-      simulate: true,
-      stages: ['06-review'],
-    });
-
-    expect(result.manifest.mode).toBe('simulated');
-    expect(result.manifest.stages['06-review']?.status).toBe('simulated');
-    for (const gateResult of result.manifest.stages['06-review']?.gateResults ?? []) {
-      expect(gateResult.outcome).toBe('simulated');
-    }
-  });
-
-  it('deploy stage fails closed without a wired ci-pipeline provider in normal mode', async () => {
-    const executor = createTypedExecutor();
-
-    await expect(
-      runPipeline(briefPath, tmpDir, {
-        executor,
-        stages: ['09-deploy'],
-      }),
-    ).rejects.toMatchObject({
-      name: 'ForgeCapabilityError',
-      code: 'FORGE_NO_CI_PIPELINE',
-    });
-  });
-});
-
-describe('fail-closed: no vacuous gate commands remain', () => {
-  it('stage constants contain no echo/synthetic-approval, vacuous true, or empty gate commands', () => {
-    for (const [stageName, spec] of Object.entries(STAGE_SPECS)) {
-      for (const gate of spec.qualityGates) {
-        const serialized = JSON.stringify(gate);
-        // The echo-review synthetic approval must be gone.
-        expect(serialized, `stage ${stageName} gate ${serialized}`).not.toContain('echo');
-        expect(serialized, `stage ${stageName} gate ${serialized}`).not.toMatch(/"verdict"\s*:/);
-        expect(serialized, `stage ${stageName} gate ${serialized}`).not.toMatch(
-          /"summary"\s*:\s*"review-pass"/,
-        );
-        // No vacuous literal `true` gate.
-        expect(gate, `stage ${stageName}`).not.toBe('true');
-        // Command gates must carry a real, non-empty command.
-        if (isCommandGate(gate)) {
-          const command = typeof gate === 'string' ? gate : gate.command;
-          expect(command.trim().length, `stage ${stageName} gate ${serialized}`).toBeGreaterThan(0);
-        }
-      }
-    }
-  });
-
-  it('board tasks contain no vacuous true gates', () => {
-    const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'forge-board-gates-'));
-    try {
-      const tasks = generateBoardTasks('# Brief', [], tmpDir, 'BOARD-TEST');
-      for (const task of tasks) {
-        for (const gate of task.qualityGates) {
-          expect(gate, `task ${task.id}`).not.toBe('true');
-          const serialized = JSON.stringify(gate);
-          expect(serialized, `task ${task.id} gate ${serialized}`).not.toContain('echo');
-        }
-      }
-    } finally {
-      fs.rmSync(tmpDir, { recursive: true, force: true });
-    }
-  });
-});
@@ -12,10 +12,10 @@ import {
  resumePipeline,
  getPipelineStatus,
 } from '../src/pipeline-runner.js';
-import type { ForgeTask, ForgeTaskResult, RunManifest, TaskExecutor } from '../src/types.js';
-import { gateLabel, isCommandGate } from '../src/outcomes.js';
+import type { ForgeTask, RunManifest, TaskExecutor } from '../src/types.js';
+import type { TaskResult } from '@mosaicstack/macp';

-/** Mock TaskExecutor that records submitted tasks and returns typed results. */
+/** Mock TaskExecutor that records submitted tasks and returns success. */
 function createMockExecutor(options?: {
  failStage?: string;
 }): TaskExecutor & { submittedTasks: ForgeTask[] } {
@@ -25,7 +25,7 @@ function createMockExecutor(options?: {
    async submitTask(task: ForgeTask) {
      submittedTasks.push(task);
    },
-    async waitForCompletion(taskId: string): Promise<ForgeTaskResult> {
+    async waitForCompletion(taskId: string): Promise<TaskResult> {
      const failStage = options?.failStage;
      const task = submittedTasks.find((t) => t.id === taskId);
      const stageName = task?.metadata?.['stageName'] as string | undefined;
@@ -33,8 +33,7 @@ function createMockExecutor(options?: {
      if (failStage && stageName === failStage) {
        return {
          task_id: taskId,
-          outcome: 'failed',
-          reason: 'mock task failure',
+          status: 'failed',
          completed_at: new Date().toISOString(),
          exit_code: 1,
          gate_results: [],
@@ -42,17 +41,10 @@ function createMockExecutor(options?: {
      }
      return {
        task_id: taskId,
-        outcome: 'passed',
-        reason: 'mock verified',
+        status: 'completed',
        completed_at: new Date().toISOString(),
        exit_code: 0,
-        gate_results: (task?.qualityGates ?? [])
-          .filter((gate) => isCommandGate(gate))
-          .map((gate) => ({
-            gate: gateLabel(gate),
-            outcome: 'passed' as const,
-            reason: 'mock verified',
-          })),
+        gate_results: [],
      };
    },
    async getTaskStatus() {
@@ -164,13 +156,12 @@ describe('runPipeline', () => {
    const executor = createMockExecutor();
    const result = await runPipeline(briefPath, tmpDir, {
      executor,
-      stages: ['00-intake', '05-coding'],
+      stages: ['00-intake', '00b-discovery'],
    });

    expect(result.runId).toMatch(/^\d{8}-\d{6}$/);
-    expect(result.stages).toEqual(['00-intake', '05-coding']);
+    expect(result.stages).toEqual(['00-intake', '00b-discovery']);
    expect(result.manifest.status).toBe('completed');
-    expect(result.manifest.mode).toBe('normal');
    expect(executor.submittedTasks).toHaveLength(2);
  });

@@ -189,17 +180,12 @@ describe('runPipeline', () => {
    const executor = createMockExecutor();
    const result = await runPipeline(briefPath, tmpDir, {
      executor,
-      stages: ['00-intake', '05-coding'],
+      stages: ['00-intake', '00b-discovery'],
    });

    const manifest = loadManifest(result.runDir);
    expect(manifest.stages['00-intake']?.status).toBe('passed');
-    expect(manifest.stages['05-coding']?.status).toBe('passed');
-    expect(manifest.stages['05-coding']?.gateResults?.map((g) => g.outcome)).toEqual([
-      'passed',
-      'passed',
-      'passed',
-    ]);
+    expect(manifest.stages['00b-discovery']?.status).toBe('passed');
  });

  it('respects CLI class override', async () => {
@@ -229,7 +215,7 @@ describe('runPipeline', () => {
    const executor = createMockExecutor();
    await runPipeline(briefPath, tmpDir, {
      executor,
-      stages: ['00-intake', '05-coding', '08-test'],
+      stages: ['00-intake', '00b-discovery', '02-planning-1'],
    });

    expect(executor.submittedTasks[0]!.dependsOn).toBeUndefined();
@@ -238,14 +224,14 @@ describe('runPipeline', () => {
  });

  it('handles stage failure', async () => {
-    const executor = createMockExecutor({ failStage: '05-coding' });
+    const executor = createMockExecutor({ failStage: '00b-discovery' });

    await expect(
      runPipeline(briefPath, tmpDir, {
        executor,
-        stages: ['00-intake', '05-coding'],
+        stages: ['00-intake', '00b-discovery'],
      }),
-    ).rejects.toThrow('Stage 05-coding failed');
+    ).rejects.toThrow('Stage 00b-discovery failed');
  });

  it('marks manifest as failed on stage failure', async () => {
@@ -284,143 +270,30 @@ describe('resumePipeline', () => {
    fs.rmSync(tmpDir, { recursive: true, force: true });
  });

-  it('resumes from first incomplete stage and fails closed at the next provider gate', async () => {
-    // Simulate a run whose authority stages were approved out-of-band
-    // (recorded as passed) and whose coding stage failed mechanically.
-    const runId = '20260101-000000';
-    const runDir = path.join(tmpDir, '.forge', 'runs', runId);
-    fs.mkdirSync(runDir, { recursive: true });
-    const passed = { status: 'passed' as const, startedAt: '2026-01-01T00:00:00Z' };
-    saveManifest(runDir, {
-      runId,
-      brief: briefPath,
-      codebase: tmpDir,
-      briefClass: 'hotfix',
-      classSource: 'frontmatter',
-      forceBoard: false,
-      mode: 'normal',
-      createdAt: '2026-01-01T00:00:00Z',
-      updatedAt: '2026-01-01T00:00:00Z',
-      currentStage: '05-coding',
-      status: 'failed',
-      stages: {
-        '00-intake': passed,
-        '00b-discovery': passed,
-        '02-planning-1': passed,
-        '03-planning-2': passed,
-        '04-planning-3': passed,
-        '05-coding': { status: 'failed', reason: 'gate failed' },
-      },
-    });
+  it('resumes from first incomplete stage', async () => {
+    // First run fails on discovery
+    const executor1 = createMockExecutor({ failStage: '00b-discovery' });
+    let runDir: string;

-    // Resume re-runs 05-coding (the first non-passed stage), then fails
-    // closed at 06-review because no reviewer provider is wired.
-    const executor = createMockExecutor();
-    await expect(resumePipeline(runDir, executor)).rejects.toMatchObject({
-      name: 'ForgeCapabilityError',
-      code: 'FORGE_NO_REVIEWER',
-    });
-
-    const manifest = loadManifest(runDir);
-    expect(manifest.stages['05-coding']?.status).toBe('passed');
-    expect(manifest.stages['06-review']?.status).toBe('blocked');
-    expect(manifest.status).toBe('failed');
-  });
-
-  it('resumes to completion as simulated under explicit simulate', async () => {
-    const runId = '20260101-000003';
-    const runDir = path.join(tmpDir, '.forge', 'runs', runId);
-    fs.mkdirSync(runDir, { recursive: true });
-    const passed = { status: 'passed' as const, startedAt: '2026-01-01T00:00:00Z' };
-    saveManifest(runDir, {
-      runId,
-      brief: briefPath,
-      codebase: tmpDir,
-      briefClass: 'hotfix',
-      classSource: 'frontmatter',
-      forceBoard: false,
-      mode: 'normal',
-      createdAt: '2026-01-01T00:00:00Z',
-      updatedAt: '2026-01-01T00:00:00Z',
-      currentStage: '05-coding',
-      status: 'failed',
-      stages: {
-        '00-intake': passed,
-        '00b-discovery': passed,
-        '02-planning-1': passed,
-        '03-planning-2': passed,
-        '04-planning-3': passed,
-        '05-coding': { status: 'failed', reason: 'gate failed' },
-      },
-    });
-
-    const result = await resumePipeline(runDir, undefined, { simulate: true });
-
-    expect(result.manifest.status).toBe('simulated');
-    expect(result.manifest.mode).toBe('simulated');
-    expect(result.stages[0]).toBe('05-coding');
-    for (const stage of result.stages) {
-      expect(result.manifest.stages[stage]?.status).toBe('simulated');
+    try {
+      await runPipeline(briefPath, tmpDir, {
+        executor: executor1,
+        stages: ['00-intake', '00b-discovery', '02-planning-1'],
+      });
+    } catch {
+      // expected
    }
-  });

-  it('fails closed on resume when the next stage needs authority sign-off', async () => {
-    const runId = '20260101-000001';
-    const runDir = path.join(tmpDir, '.forge', 'runs', runId);
-    fs.mkdirSync(runDir, { recursive: true });
-    saveManifest(runDir, {
-      runId,
-      brief: briefPath,
-      codebase: tmpDir,
-      briefClass: 'hotfix',
-      classSource: 'frontmatter',
-      forceBoard: false,
-      mode: 'normal',
-      createdAt: '2026-01-01T00:00:00Z',
-      updatedAt: '2026-01-01T00:00:00Z',
-      currentStage: '00-intake',
-      status: 'in_progress',
-      stages: {
-        '00-intake': { status: 'passed' },
-      },
-    });
+    const runsDir = path.join(tmpDir, '.forge', 'runs');
+    runDir = path.join(runsDir, fs.readdirSync(runsDir)[0]!);

-    const executor = createMockExecutor();
-    await expect(resumePipeline(runDir, executor)).rejects.toMatchObject({
-      name: 'ForgeCapabilityError',
-      code: 'FORGE_AUTHORITY_REQUIRED',
-    });
+    // Resume should pick up from 00b-discovery
+    const executor2 = createMockExecutor();
+    const result = await resumePipeline(runDir, executor2);

-    const manifest = loadManifest(runDir);
-    expect(manifest.stages['00b-discovery']?.status).toBe('waiting-for-authority');
-    expect(manifest.status).toBe('waiting-for-authority');
-  });
-
-  it('fails closed on resume without an executor or --simulate', async () => {
-    const runId = '20260101-000002';
-    const runDir = path.join(tmpDir, '.forge', 'runs', runId);
-    fs.mkdirSync(runDir, { recursive: true });
-    saveManifest(runDir, {
-      runId,
-      brief: briefPath,
-      codebase: tmpDir,
-      briefClass: 'hotfix',
-      classSource: 'frontmatter',
-      forceBoard: false,
-      mode: 'normal',
-      createdAt: '2026-01-01T00:00:00Z',
-      updatedAt: '2026-01-01T00:00:00Z',
-      currentStage: '00-intake',
-      status: 'in_progress',
-      stages: {
-        '00-intake': { status: 'passed' },
-      },
-    });
-
-    await expect(resumePipeline(runDir)).rejects.toMatchObject({
-      name: 'ForgeCapabilityError',
-      code: 'FORGE_NO_EXECUTOR',
-    });
+    expect(result.manifest.status).toBe('completed');
+    // Should have re-run from 00b-discovery onward
+    expect(result.stages[0]).toBe('00b-discovery');
  });
 });

@@ -95,14 +95,7 @@ export function generateBoardTasks(
      briefPath,
      resultPath: resultRelPath,
      timeoutSeconds: 120,
-      qualityGates: [
-        {
-          kind: 'authority',
-          capability: 'board-approval',
-          reason:
-            'persona evaluation is judged by board synthesis (authority review); no mechanical gate exists',
-        },
-      ],
+      qualityGates: ['true'],
      metadata: {
        personaName: persona.name,
        personaSlug: persona.slug,
@@ -128,13 +121,7 @@ export function generateBoardTasks(
    timeoutSeconds: 120,
    dependsOn: personaTaskIds,
    dependsOnPolicy: 'all_terminal',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'board-approval',
-        reason: 'board synthesis is an authority decision; no mechanical gate exists',
-      },
-    ],
+    qualityGates: ['true'],
    metadata: {
      resultOutputPath: synthesisResult,
      inputResultPaths: personaResultPaths,
@@ -1,11 +1,7 @@
-import fs from 'node:fs';
-import os from 'node:os';
-import path from 'node:path';
 import { Command } from 'commander';
-import { describe, expect, it, vi, beforeEach, afterEach } from 'vitest';
+import { describe, expect, it } from 'vitest';

 import { registerForgeCommand } from './cli.js';
-import { loadManifest } from './pipeline-runner.js';

 describe('registerForgeCommand', () => {
  it('registers a "forge" command on the parent program', () => {
@@ -59,94 +55,3 @@ describe('registerForgeCommand', () => {
    }).not.toThrow();
  });
 });
-
-describe('forge run fail-closed behavior (SDLC-D-035)', () => {
-  let tmpDir: string;
-  let briefPath: string;
-  let errSpy: ReturnType<typeof vi.spyOn>;
-  let logSpy: ReturnType<typeof vi.spyOn>;
-  let prevExitCode: string | number | null | undefined;
-
-  const parse = (args: string[]) => {
-    const program = new Command();
-    registerForgeCommand(program);
-    return program.parseAsync(['forge', ...args], { from: 'user' });
-  };
-
-  beforeEach(() => {
-    tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'forge-cli-failclosed-'));
-    briefPath = path.join(tmpDir, 'brief.md');
-    fs.writeFileSync(briefPath, '# Fix bug\n\nA bugfix for lint cleanup.');
-    errSpy = vi.spyOn(console, 'error').mockImplementation(() => {});
-    logSpy = vi.spyOn(console, 'log').mockImplementation(() => {});
-    prevExitCode = process.exitCode;
-  });
-
-  afterEach(() => {
-    errSpy.mockRestore();
-    logSpy.mockRestore();
-    process.exitCode = prevExitCode;
-    fs.rmSync(tmpDir, { recursive: true, force: true });
-  });
-
-  it('exits nonzero with a typed FORGE_NO_EXECUTOR error when no executor is wired and --simulate is absent', async () => {
-    await parse(['run', '--brief', briefPath, '--codebase', tmpDir]);
-
-    expect(process.exitCode).toBe(1);
-    const errText = errSpy.mock.calls.map((c) => c.join(' ')).join('\n');
-    expect(errText).toContain('FORGE_NO_EXECUTOR');
-    // It must never run the pipeline with a stub and report success.
-    expect(fs.existsSync(path.join(tmpDir, '.forge', 'runs'))).toBe(false);
-  });
-
-  it('completes with typed simulated results and exit 0 under explicit --simulate', async () => {
-    await parse(['run', '--brief', briefPath, '--codebase', tmpDir, '--simulate']);
-
-    expect(process.exitCode).toBeUndefined();
-
-    // Loud simulated-mode summary.
-    const logText = logSpy.mock.calls.map((c) => c.join(' ')).join('\n');
-    expect(logText).toContain('SIMULATED');
-
-    // Manifest records the mode and simulated per-result statuses.
-    const runsDir = path.join(tmpDir, '.forge', 'runs');
-    const runIds = fs.readdirSync(runsDir);
-    expect(runIds).toHaveLength(1);
-    const manifest = loadManifest(path.join(runsDir, runIds[0]!));
-    expect(manifest.mode).toBe('simulated');
-    expect(manifest.status).toBe('simulated');
-    for (const stageStatus of Object.values(manifest.stages)) {
-      expect(stageStatus?.status).toBe('simulated');
-      for (const gateResult of stageStatus?.gateResults ?? []) {
-        expect(gateResult.outcome).toBe('simulated');
-      }
-    }
-  });
-
-  it('resume exits nonzero with a typed FORGE_NO_EXECUTOR error without --simulate', async () => {
-    const runDir = path.join(tmpDir, '.forge', 'runs', '20260101-000000');
-    fs.mkdirSync(runDir, { recursive: true });
-    fs.writeFileSync(
-      path.join(runDir, 'manifest.json'),
-      JSON.stringify({
-        runId: '20260101-000000',
-        brief: briefPath,
-        codebase: tmpDir,
-        briefClass: 'hotfix',
-        classSource: 'frontmatter',
-        forceBoard: false,
-        createdAt: '2026-01-01T00:00:00Z',
-        updatedAt: '2026-01-01T00:00:00Z',
-        currentStage: '00-intake',
-        status: 'in_progress',
-        stages: { '00-intake': { status: 'passed' } },
-      }),
-    );
-
-    await parse(['resume', '20260101-000000', '--project', tmpDir]);
-
-    expect(process.exitCode).toBe(1);
-    const errText = errSpy.mock.calls.map((c) => c.join(' ')).join('\n');
-    expect(errText).toContain('FORGE_NO_EXECUTOR');
-  });
-});
@@ -5,47 +5,37 @@ import type { Command } from 'commander';

 import { classifyBrief } from './brief-classifier.js';
 import { STAGE_LABELS, STAGE_SEQUENCE } from './constants.js';
-import { ForgeCapabilityError } from './errors.js';
 import { getEffectivePersonas, loadBoardPersonas } from './persona-loader.js';
 import { generateRunId, getPipelineStatus, loadManifest, runPipeline } from './pipeline-runner.js';
-import { createSimulatedExecutor } from './simulated-executor.js';
-import type { PipelineOptions, RunManifest, RunMode } from './types.js';
+import type { PipelineOptions, RunManifest, TaskExecutor } from './types.js';
+
+// ---------------------------------------------------------------------------
+// Stub executor — used when no real executor is wired at CLI invocation time.
+// ---------------------------------------------------------------------------
+
+const stubExecutor: TaskExecutor = {
+  async submitTask(task) {
+    console.log(`  [forge] stage submitted: ${task.id} (${task.title})`);
+  },
+  async waitForCompletion(taskId, _timeoutMs) {
+    console.log(`  [forge] stage complete:  ${taskId}`);
+    return {
+      task_id: taskId,
+      status: 'completed' as const,
+      completed_at: new Date().toISOString(),
+      exit_code: 0,
+      gate_results: [],
+    };
+  },
+  async getTaskStatus(_taskId) {
+    return 'completed' as const;
+  },
+};

 // ---------------------------------------------------------------------------
 // Helpers
 // ---------------------------------------------------------------------------

-/** Resolve a run's effective mode, defaulting legacy manifests to normal. */
-function runModeOf(manifest: RunManifest): RunMode {
-  return manifest.mode ?? 'normal';
-}
-
-/** Print a loud banner so a simulated run can never be misread as verified. */
-function printSimulatedBanner(): void {
-  console.log('');
-  console.log('[forge] ===============================================================');
-  console.log('[forge] MODE: SIMULATED — no stage or gate was really executed.');
-  console.log('[forge] All results are synthetic and MUST NOT be read as verified');
-  console.log('[forge] success. Wire a real executor/providers and re-run to verify.');
-  console.log('[forge] ===============================================================');
-}
-
-/** Print a typed error line for fail-closed capability errors. */
-function printCapabilityError(err: ForgeCapabilityError): void {
-  console.error(`[forge] error ${err.code}: ${err.message}`);
-  console.error(`[forge] missing capability: ${err.capability}`);
-}
-
-/** Handle a pipeline error uniformly: typed capability errors get their code. */
-function handlePipelineError(err: unknown): void {
-  if (err instanceof ForgeCapabilityError) {
-    printCapabilityError(err);
-  } else {
-    console.error(`[forge] pipeline failed: ${err instanceof Error ? err.message : String(err)}`);
-  }
-  process.exitCode = 1;
-}
-
 function formatDuration(startedAt?: string, completedAt?: string): string {
  if (!startedAt || !completedAt) return '-';
  const ms = new Date(completedAt).getTime() - new Date(startedAt).getTime();
@@ -54,24 +44,19 @@ function formatDuration(startedAt?: string, completedAt?: string): string {
 }

 function printManifestTable(manifest: RunManifest): void {
-  const mode = runModeOf(manifest);
  console.log(`\nRun ID : ${manifest.runId}`);
  console.log(`Status : ${manifest.status}`);
-  console.log(`Mode   : ${mode}`);
-  if (mode === 'simulated') {
-    console.log('WARNING: SIMULATED RUN — results are synthetic, not verified success.');
-  }
  console.log(`Brief  : ${manifest.brief}`);
  console.log(`Class  : ${manifest.briefClass} (${manifest.classSource})`);
  console.log(`Updated: ${manifest.updatedAt}`);
  console.log('');
-  console.log('Stage'.padEnd(22) + 'Status'.padEnd(24) + 'Duration');
-  console.log('-'.repeat(60));
+  console.log('Stage'.padEnd(22) + 'Status'.padEnd(14) + 'Duration');
+  console.log('-'.repeat(50));
  for (const stage of STAGE_SEQUENCE) {
    const s = manifest.stages[stage];
    if (!s) continue;
    const label = (STAGE_LABELS[stage] ?? stage).padEnd(22);
-    const status = s.status.padEnd(24);
+    const status = s.status.padEnd(14);
    const dur = formatDuration(s.startedAt, s.completedAt);
    console.log(`${label}${status}${dur}`);
  }
@@ -105,58 +90,23 @@ function listRecentRuns(projectRoot?: string): void {
  }

  console.log('\nRecent runs:');
-  console.log('Run ID'.padEnd(22) + 'Status'.padEnd(24) + 'Mode'.padEnd(12) + 'Brief');
-  console.log('-'.repeat(80));
+  console.log('Run ID'.padEnd(22) + 'Status'.padEnd(14) + 'Brief');
+  console.log('-'.repeat(70));

  for (const runId of entries) {
    const runDir = path.join(runsDir, runId);
    try {
      const manifest = loadManifest(runDir);
-      const status = manifest.status.padEnd(24);
-      const mode = runModeOf(manifest).padEnd(12);
+      const status = manifest.status.padEnd(14);
      const brief = path.basename(manifest.brief);
-      console.log(`${runId.padEnd(22)}${status}${mode}${brief}`);
+      console.log(`${runId.padEnd(22)}${status}${brief}`);
    } catch {
-      console.log(`${runId.padEnd(22)}${'(unreadable)'.padEnd(24)}`);
+      console.log(`${runId.padEnd(22)}${'(unreadable)'.padEnd(14)}`);
    }
  }
  console.log('');
 }

-/**
- * Apply the exit-code policy for a finished pipeline run (SDLC-D-035):
- *
- * - exit 0 only for a verified `completed` normal run, or for an overall
- *   `simulated` run when the caller explicitly passed --simulate;
- * - anything else exits nonzero so it can never be read as success.
- */
-function applyRunExitPolicy(result: { manifest: RunManifest; runDir: string }, simulate: boolean) {
-  const { manifest } = result;
-
-  if (runModeOf(manifest) === 'simulated') {
-    if (!simulate || manifest.status !== 'simulated') {
-      console.error(
-        '[forge] error FORGE_MODE_MISMATCH: run reports simulated results without an explicit, ' +
-          'consistent --simulate request; refusing to report success.',
-      );
-      process.exitCode = 1;
-      return;
-    }
-    printSimulatedBanner();
-    console.log(`[forge] run directory: ${result.runDir}`);
-    return; // exit 0 — the caller explicitly opted into simulation
-  }
-
-  if (manifest.status !== 'completed') {
-    console.error(`[forge] run did not complete: terminal status '${manifest.status}'`);
-    process.exitCode = 1;
-    return;
-  }
-
-  console.log(`[forge] pipeline complete (mode: normal): ${manifest.runId}`);
-  console.log(`[forge] run directory: ${result.runDir}`);
-}
-
 // ---------------------------------------------------------------------------
 // Register function
 // ---------------------------------------------------------------------------
@@ -179,11 +129,6 @@ export function registerForgeCommand(parent: Command): void {
    .option('--config <path>', 'Path to forge config file (.forge/config.yaml)')
    .option('--codebase <path>', 'Codebase root to pass to the pipeline', process.cwd())
    .option('--dry-run', 'Print planned stages without executing', false)
-    .option(
-      '--simulate',
-      'Simulate execution without real providers (every result is typed simulated, never verified)',
-      false,
-    )
    .action(
      async (opts: {
        brief: string;
@@ -192,7 +137,6 @@ export function registerForgeCommand(parent: Command): void {
        config?: string;
        codebase: string;
        dryRun: boolean;
-        simulate: boolean;
      }) => {
        const briefPath = path.resolve(opts.brief);

@@ -205,22 +149,14 @@ export function registerForgeCommand(parent: Command): void {
        const briefContent = fs.readFileSync(briefPath, 'utf-8');
        const briefClass = classifyBrief(briefContent);
        const projectRoot = opts.codebase;
-        // A real executor is never wired at CLI invocation time today, so the
-        // only executor we may construct is the explicitly-requested simulated
-        // one. Normal mode fails closed with FORGE_NO_EXECUTOR.
-        const executor = opts.simulate ? createSimulatedExecutor() : undefined;

        if (opts.resume) {
          const runId = opts.runId ?? generateRunId();
          const runDir = resolveRunDir(runId, projectRoot);
          console.log(`[forge] resuming run: ${runId}`);
-          try {
-            const { resumePipeline } = await import('./pipeline-runner.js');
-            const result = await resumePipeline(runDir, executor, { simulate: opts.simulate });
-            applyRunExitPolicy(result, opts.simulate);
-          } catch (err) {
-            handlePipelineError(err);
-          }
+          const { resumePipeline } = await import('./pipeline-runner.js');
+          const result = await resumePipeline(runDir, stubExecutor);
+          console.log(`[forge] pipeline complete: ${result.runId}`);
          return;
        }

@@ -228,8 +164,7 @@ export function registerForgeCommand(parent: Command): void {
          briefClass,
          codebase: projectRoot,
          dryRun: opts.dryRun,
-          executor,
-          simulate: opts.simulate,
+          executor: stubExecutor,
        };

        if (opts.dryRun) {
@@ -245,15 +180,16 @@ export function registerForgeCommand(parent: Command): void {

        console.log(`[forge] starting pipeline for brief: ${briefPath}`);
        console.log(`[forge] classified as: ${briefClass}`);
-        if (opts.simulate) {
-          console.log('[forge] mode: SIMULATED (explicit --simulate)');
-        }

        try {
          const result = await runPipeline(briefPath, projectRoot, pipelineOptions);
-          applyRunExitPolicy(result, opts.simulate);
+          console.log(`[forge] pipeline complete: ${result.runId}`);
+          console.log(`[forge] run directory: ${result.runDir}`);
        } catch (err) {
-          handlePipelineError(err);
+          console.error(
+            `[forge] pipeline failed: ${err instanceof Error ? err.message : String(err)}`,
+          );
+          process.exitCode = 1;
        }
      },
    );
@@ -288,12 +224,7 @@ export function registerForgeCommand(parent: Command): void {
    .command('resume <runId>')
    .description('Resume a stopped or failed pipeline run')
    .option('--project <path>', 'Project root (defaults to cwd)', process.cwd())
-    .option(
-      '--simulate',
-      'Simulate execution without real providers (every result is typed simulated, never verified)',
-      false,
-    )
-    .action(async (runId: string, opts: { project: string; simulate: boolean }) => {
+    .action(async (runId: string, opts: { project: string }) => {
      const runDir = resolveRunDir(runId, opts.project);

      if (!fs.existsSync(runDir)) {
@@ -303,20 +234,15 @@ export function registerForgeCommand(parent: Command): void {
      }

      console.log(`[forge] resuming run: ${runId}`);
-      if (opts.simulate) {
-        console.log('[forge] mode: SIMULATED (explicit --simulate)');
-      }
-
-      // No real executor is wired at CLI invocation time; only the explicitly
-      // requested simulated executor may be constructed (fail closed otherwise).
-      const executor = opts.simulate ? createSimulatedExecutor() : undefined;

      try {
        const { resumePipeline } = await import('./pipeline-runner.js');
-        const result = await resumePipeline(runDir, executor, { simulate: opts.simulate });
-        applyRunExitPolicy(result, opts.simulate);
+        const result = await resumePipeline(runDir, stubExecutor);
+        console.log(`[forge] pipeline complete: ${result.runId}`);
+        console.log(`[forge] run directory: ${result.runDir}`);
      } catch (err) {
-        handlePipelineError(err);
+        console.error(`[forge] resume failed: ${err instanceof Error ? err.message : String(err)}`);
+        process.exitCode = 1;
      }
    });

@@ -9,16 +9,7 @@ export const PACKAGE_ROOT = path.resolve(path.dirname(fileURLToPath(import.meta.
 /** Pipeline asset directory (stages, agents, rails, gates, templates). */
 export const PIPELINE_DIR = path.join(PACKAGE_ROOT, 'pipeline');

-/** Stage specifications — defines every pipeline stage.
- *\n * Gate semantics (SDLC-D-035): every gate is one of
- * - a real command string / GateEntry a mechanical runner can execute,
- * - an `authority` gate (human/board sign-off; produces waiting-for-authority),
- * - a `provider` gate (requires a wired provider such as a reviewer or CI pipeline).
- *
- * Vacuous gates (`true`, echo'd synthetic approvals, placeholder ci-pipeline
- * commands) are forbidden: a stage whose gate has no real implementation
- * fails closed instead of passing.
- */
+/** Stage specifications — defines every pipeline stage. */
 export const STAGE_SPECS: Record<string, StageSpec> = {
  '00-intake': {
    number: '00',
@@ -36,13 +27,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'discovery-complete',
    promptFile: '00b-discovery.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'discovery-complete',
-        reason: 'discovery completion is attested by an authority; no mechanical check exists',
-      },
-    ],
+    qualityGates: ['true'],
  },
  '01-board': {
    number: '01',
@@ -51,13 +36,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'review',
    gate: 'board-approval',
    promptFile: '01-board.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'board-approval',
-        reason: 'board approval is a board/human decision; no mechanical gate exists',
-      },
-    ],
+    qualityGates: [{ type: 'ci-pipeline', command: 'board-approval (via board-tasks)' }],
  },
  '01b-brief-analyzer': {
    number: '01b',
@@ -66,13 +45,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'brief-analysis-complete',
    promptFile: '01-board.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'brief-analysis-complete',
-        reason: 'brief analysis completion is attested by an authority; no mechanical check exists',
-      },
-    ],
+    qualityGates: ['true'],
  },
  '02-planning-1': {
    number: '02',
@@ -81,13 +54,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'architecture-approval',
    promptFile: '02-planning-1-architecture.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'architecture-approval',
-        reason: 'ADR approval requires authority sign-off; no mechanical check exists',
-      },
-    ],
+    qualityGates: ['true'],
  },
  '03-planning-2': {
    number: '03',
@@ -96,14 +63,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'implementation-approval',
    promptFile: '03-planning-2-implementation.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'implementation-approval',
-        reason:
-          'implementation spec approval requires authority sign-off; no mechanical check exists',
-      },
-    ],
+    qualityGates: ['true'],
  },
  '04-planning-3': {
    number: '04',
@@ -112,14 +72,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'decomposition-approval',
    promptFile: '04-planning-3-decomposition.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 'decomposition-approval',
-        reason:
-          'task decomposition approval requires authority sign-off; no mechanical check exists',
-      },
-    ],
+    qualityGates: ['true'],
  },
  '05-coding': {
    number: '05',
@@ -139,10 +92,9 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    promptFile: '06-review.md',
    qualityGates: [
      {
-        kind: 'provider',
-        capability: 'reviewer',
-        reason:
-          'review verdicts require a wired reviewer provider; synthetic approvals are not permitted',
+        type: 'ai-review',
+        command:
+          'echo \'{"summary":"review-pass","verdict":"approve","findings":[],"stats":{"blockers":0,"should_fix":0,"suggestions":0}}\'',
      },
    ],
  },
@@ -153,13 +105,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'coding',
    gate: 're-review',
    promptFile: '07-remediate.md',
-    qualityGates: [
-      {
-        kind: 'authority',
-        capability: 're-review',
-        reason: 'remediation re-review is an approval-based gate; no mechanical check exists',
-      },
-    ],
+    qualityGates: ['true'],
  },
  '08-test': {
    number: '08',
@@ -177,13 +123,7 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'deploy',
    gate: 'deploy-verification',
    promptFile: '09-deploy.md',
-    qualityGates: [
-      {
-        kind: 'provider',
-        capability: 'ci-pipeline',
-        reason: 'deploy verification requires a wired CI pipeline provider',
-      },
-    ],
+    qualityGates: [{ type: 'ci-pipeline', command: 'deploy-verification' }],
  },
 };

@@ -1,46 +0,0 @@
-/**
- * Typed fail-closed capability errors (SDLC-D-035).
- *
- * A Forge run must fail closed when a required capability (executor, reviewer
- * provider, CI pipeline, authority sign-off) is missing. These typed errors
- * name the missing capability so callers can distinguish "not wired" from
- * ordinary execution failures.
- */
-
-/** Closed set of typed Forge capability error codes. */
-export const FORGE_ERROR_CODES = [
-  'FORGE_NO_EXECUTOR',
-  'FORGE_NO_REVIEWER',
-  'FORGE_NO_CI_PIPELINE',
-  'FORGE_NO_PROVIDER',
-  'FORGE_AUTHORITY_REQUIRED',
-] as const;
-
-export type ForgeErrorCode = (typeof FORGE_ERROR_CODES)[number];
-
-/** Raised when a required capability is missing and the pipeline must fail closed. */
-export class ForgeCapabilityError extends Error {
-  /** Typed error code from the closed FORGE_ERROR_CODES set. */
-  readonly code: ForgeErrorCode;
-  /** The missing capability, e.g. `task-executor`, `reviewer`, `board-approval`. */
-  readonly capability: string;
-
-  constructor(code: ForgeErrorCode, capability: string, message: string) {
-    super(message);
-    this.name = 'ForgeCapabilityError';
-    this.code = code;
-    this.capability = capability;
-  }
-}
-
-/** Map a provider gate capability to its typed error code. */
-export function providerErrorCode(capability: string): ForgeErrorCode {
-  switch (capability) {
-    case 'reviewer':
-      return 'FORGE_NO_REVIEWER';
-    case 'ci-pipeline':
-      return 'FORGE_NO_CI_PIPELINE';
-    default:
-      return 'FORGE_NO_PROVIDER';
-  }
-}
@@ -5,13 +5,6 @@ export type {
  StageSpec,
  BriefClass,
  ClassSource,
-  ForgeOutcome,
-  AuthorityGate,
-  ProviderGate,
-  ForgeGate,
-  ForgeGateResult,
-  ForgeTaskResult,
-  RunMode,
  StageStatus,
  RunManifest,
  ForgeTaskStatus,
@@ -88,24 +81,5 @@ export {
  getPipelineStatus,
 } from './pipeline-runner.js';

-// Fail-closed errors and typed outcome model (SDLC-D-035)
-export { FORGE_ERROR_CODES, ForgeCapabilityError, providerErrorCode } from './errors.js';
-export type { ForgeErrorCode } from './errors.js';
-export {
-  isSatisfyingOutcome,
-  isCapabilityGate,
-  isCommandGate,
-  gateLabel,
-  uniformGateResults,
-  simulatedGateResults,
-  waitingGateResults,
-  blockedGateResults,
-  evaluateStageGates,
-} from './outcomes.js';
-export type { StageEvaluation } from './outcomes.js';
-
-// Simulated executor (explicit --simulate only)
-export { createSimulatedExecutor } from './simulated-executor.js';
-
 // CLI
 export { registerForgeCommand } from './cli.js';
@@ -1,147 +0,0 @@
-import type { GateEntry } from '@mosaicstack/macp';
-
-import type {
-  AuthorityGate,
-  ForgeGate,
-  ForgeGateResult,
-  ForgeOutcome,
-  ForgeTaskResult,
-  ProviderGate,
-} from './types.js';
-
-/**
- * Gate and dependency satisfaction predicate (SDLC-D-035).
- *
- * ONLY a verified `passed` outcome satisfies. Every other member of the closed
- * outcome set — including `simulated` — is non-satisfying, so a simulated or
- * authority-blocked result can never be read as success-by-verification.
- */
-export function isSatisfyingOutcome(outcome: ForgeOutcome): boolean {
-  return outcome === 'passed';
-}
-
-/** Whether a gate is an authority or provider gate (capability-based, command-less). */
-export function isCapabilityGate(gate: ForgeGate): gate is AuthorityGate | ProviderGate {
-  if (typeof gate !== 'object' || gate === null) return false;
-  const kind = (gate as Record<string, unknown>)['kind'];
-  return kind === 'authority' || kind === 'provider';
-}
-
-/** Whether a gate definition carries a real command a mechanical runner can execute. */
-export function isCommandGate(gate: ForgeGate): gate is string | GateEntry {
-  if (typeof gate === 'string') {
-    return gate.trim().length > 0;
-  }
-  if (isCapabilityGate(gate)) {
-    // Authority and provider gates are satisfied by a capability, not a command.
-    return false;
-  }
-  return typeof gate.command === 'string' && gate.command.trim().length > 0;
-}
-
-/** Typed label identifying a gate in results and logs. */
-export function gateLabel(gate: ForgeGate): string {
-  if (typeof gate === 'string') return gate;
-  if (isCapabilityGate(gate)) return `${gate.kind}:${gate.capability}`;
-  return gate.command || gate.type || 'unnamed-gate';
-}
-
-/** Reason string stamped on every simulated gate result. */
-export const SIMULATED_GATE_REASON =
-  'simulated execution (--simulate): gate was not evaluated by a real implementation';
-
-/** Build typed gate results with a uniform outcome for a stage's declared gates. */
-export function uniformGateResults(
-  gates: ForgeGate[],
-  outcome: ForgeOutcome,
-  reason: string,
-): ForgeGateResult[] {
-  return gates.map((gate) => ({ gate: gateLabel(gate), outcome, reason }));
-}
-
-/** Typed simulated gate results — used exclusively in `--simulate` runs. */
-export function simulatedGateResults(gates: ForgeGate[]): ForgeGateResult[] {
-  return uniformGateResults(gates, 'simulated', SIMULATED_GATE_REASON);
-}
-
-/** Typed waiting-for-authority gate results for approval-based stages. */
-export function waitingGateResults(gates: ForgeGate[], reason: string): ForgeGateResult[] {
-  return uniformGateResults(gates, 'waiting-for-authority', reason);
-}
-
-/** Typed blocked gate results for stages whose provider capability is not wired. */
-export function blockedGateResults(gates: ForgeGate[], reason: string): ForgeGateResult[] {
-  return uniformGateResults(gates, 'blocked', reason);
-}
-
-/** Outcome of evaluating a completed stage in normal mode. */
-export interface StageEvaluation {
-  outcome: ForgeOutcome;
-  reason: string;
-  gateResults: ForgeGateResult[];
-}
-
-/**
- * Evaluate a stage's declared gates against the executor's typed result.
- *
- * Fail-closed mapping:
- * - a `simulated` task or gate outcome in normal mode maps to `error`
- * - a missing gate result for a required command gate maps to `blocked`
- * - a non-passing task outcome propagates as the stage outcome
- * - only verified `passed` task and gate outcomes yield a `passed` stage
- */
-export function evaluateStageGates(
-  stageName: string,
-  gates: ForgeGate[],
-  result: ForgeTaskResult,
-): StageEvaluation {
-  const gateResults = result.gate_results ?? [];
-
-  if (result.outcome === 'simulated') {
-    return {
-      outcome: 'error',
-      reason: `executor reported a simulated outcome for stage '${stageName}' in normal mode — refusing to treat simulated results as verified`,
-      gateResults,
-    };
-  }
-
-  if (!isSatisfyingOutcome(result.outcome)) {
-    return {
-      outcome: result.outcome,
-      reason: `task outcome is '${result.outcome}': ${result.reason}`,
-      gateResults,
-    };
-  }
-
-  for (const gate of gates) {
-    // Authority and provider gates are pre-flighted before execution; they have
-    // no mechanical result to verify here.
-    if (!isCommandGate(gate)) continue;
-
-    const label = gateLabel(gate);
-    const gateResult = gateResults.find((r) => r.gate === label);
-    if (!gateResult) {
-      return {
-        outcome: 'blocked',
-        reason: `no gate result was reported for required gate '${label}' (stage '${stageName}')`,
-        gateResults,
-      };
-    }
-    if (!isSatisfyingOutcome(gateResult.outcome)) {
-      return {
-        outcome: gateResult.outcome === 'simulated' ? 'error' : gateResult.outcome,
-        reason: `gate '${label}' outcome is '${gateResult.outcome}': ${gateResult.reason}`,
-        gateResults,
-      };
-    }
-  }
-
-  return {
-    outcome: 'passed',
-    reason:
-      gates.length === 0
-        ? "stage declares no gates; task outcome 'passed' accepted"
-        : 'all declared gates verified passed',
-    gateResults,
-  };
-}
@@ -1,33 +1,18 @@
 import fs from 'node:fs';
 import path from 'node:path';

-import { STAGE_SEQUENCE, STAGE_SPECS } from './constants.js';
+import { STAGE_SEQUENCE } from './constants.js';
 import { determineBriefClass, stagesForClass } from './brief-classifier.js';
-import { ForgeCapabilityError, providerErrorCode } from './errors.js';
-import {
-  blockedGateResults,
-  evaluateStageGates,
-  isCapabilityGate,
-  simulatedGateResults,
-  waitingGateResults,
-} from './outcomes.js';
 import { mapStageToTask } from './stage-adapter.js';
-import { createSimulatedExecutor } from './simulated-executor.js';
 import type {
  ForgeTask,
-  ForgeTaskResult,
  PipelineOptions,
  PipelineResult,
  RunManifest,
-  RunMode,
  StageStatus,
  TaskExecutor,
 } from './types.js';

-/** Reason stamped on stages that complete under explicit simulation. */
-const SIMULATED_STAGE_REASON =
-  'simulated execution (--simulate): stage was not executed by a real executor';
-
 /**
 * Generate a timestamp-based run ID.
 */
@@ -62,7 +47,6 @@ function createManifest(opts: {
  briefClass: RunManifest['briefClass'];
  classSource: RunManifest['classSource'];
  forceBoard: boolean;
-  mode: RunMode;
  runDir: string;
 }): RunManifest {
  const ts = nowISO();
@@ -73,7 +57,6 @@ function createManifest(opts: {
    briefClass: opts.briefClass,
    classSource: opts.classSource,
    forceBoard: opts.forceBoard,
-    mode: opts.mode,
    createdAt: ts,
    updatedAt: ts,
    currentStage: '',
@@ -125,199 +108,20 @@ export function selectStages(stages?: string[], skipTo?: string): string[] {
  return selected.slice(skipIndex);
 }

-/**
- * Fail closed when the required executor capability is missing (SDLC-D-035).
- */
-function requireExecutor(executor: TaskExecutor | undefined, simulate: boolean): TaskExecutor {
-  if (executor) return executor;
-  if (simulate) return createSimulatedExecutor({ log: false });
-  throw new ForgeCapabilityError(
-    'FORGE_NO_EXECUTOR',
-    'task-executor',
-    'no task executor is wired; refusing to run the pipeline with a stub executor (fail closed). ' +
-      'Pass --simulate to opt into explicitly simulated execution.',
-  );
-}
-
-/**
- * Pre-flight a stage's gates in normal mode (fail closed, SDLC-D-035).
- *
- * - authority gates: record a typed `waiting-for-authority` stage result and
- *   raise FORGE_AUTHORITY_REQUIRED — approval-based gates never pass vacuously.
- * - provider gates: record a typed `blocked` stage result and raise the typed
- *   capability error for the missing provider.
- *
- * Returns the stage status to record when the pre-flight blocks, or undefined
- * when the stage may proceed.
- */
-function preflightStageGates(
-  stageName: string,
-  manifest: RunManifest,
-): { status: StageStatus; error: ForgeCapabilityError } | undefined {
-  const spec = STAGE_SPECS[stageName];
-  if (!spec) throw new Error(`Unknown Forge stage: ${stageName}`);
-
-  for (const gate of spec.qualityGates) {
-    if (!isCapabilityGate(gate)) continue;
-
-    const startedAt = manifest.stages[stageName]?.startedAt;
-    const completedAt = nowISO();
-
-    if (gate.kind === 'authority') {
-      const reason = `gate '${gate.capability}' requires authority sign-off; no mechanical implementation exists (${gate.reason})`;
-      return {
-        status: {
-          status: 'waiting-for-authority',
-          reason,
-          startedAt,
-          completedAt,
-          gateResults: waitingGateResults(spec.qualityGates, reason),
-        },
-        error: new ForgeCapabilityError(
-          'FORGE_AUTHORITY_REQUIRED',
-          gate.capability,
-          `stage '${stageName}' is blocked on authority gate '${gate.capability}': ${gate.reason}. ` +
-            'The pipeline fails closed instead of passing vacuously. Record the approval out-of-band ' +
-            'or run with --simulate for explicitly simulated execution.',
-        ),
-      };
-    }
-
-    const reason = `gate '${gate.capability}' requires provider '${gate.capability}' and none is wired (${gate.reason})`;
-    return {
-      status: {
-        status: 'blocked',
-        reason,
-        startedAt,
-        completedAt,
-        gateResults: blockedGateResults(spec.qualityGates, reason),
-      },
-      error: new ForgeCapabilityError(
-        providerErrorCode(gate.capability),
-        gate.capability,
-        `stage '${stageName}' requires provider '${gate.capability}' which is not wired: ${gate.reason}. ` +
-          'The pipeline fails closed instead of passing vacuously.',
-      ),
-    };
-  }
-
-  return undefined;
-}
-
-/**
- * Execute the given stage tasks sequentially, updating the manifest.
- *
- * Normal mode requires a real executor and evaluates every declared command
- * gate through the typed outcome model; any non-verified result fails closed.
- * Simulate mode types every stage and gate result as `simulated`.
- */
-async function executeStages(opts: {
-  manifest: RunManifest;
-  runDir: string;
-  tasks: ForgeTask[];
-  stageNames: string[];
-  executor: TaskExecutor;
-  simulate: boolean;
-}): Promise<void> {
-  const { manifest, runDir, tasks, stageNames, executor, simulate } = opts;
-
-  for (let i = 0; i < tasks.length; i++) {
-    const task = tasks[i]!;
-    const stageName = stageNames[i]!;
-    const spec = STAGE_SPECS[stageName];
-    if (!spec) throw new Error(`Unknown Forge stage: ${stageName}`);
-
-    // Update manifest: stage in progress
-    manifest.currentStage = stageName;
-    manifest.stages[stageName] = {
-      status: 'in_progress',
-      startedAt: nowISO(),
-    };
-    saveManifest(runDir, manifest);
-
-    // Fail-closed pre-flight (normal mode only): authority/provider gates have
-    // no mechanical implementation and must never pass vacuously.
-    if (!simulate) {
-      const blocked = preflightStageGates(stageName, manifest);
-      if (blocked) {
-        manifest.stages[stageName] = blocked.status;
-        manifest.status =
-          blocked.status.status === 'waiting-for-authority' ? 'waiting-for-authority' : 'failed';
-        saveManifest(runDir, manifest);
-        throw blocked.error;
-      }
-    }
-
-    let result: ForgeTaskResult;
-    try {
-      await executor.submitTask(task);
-      result = await executor.waitForCompletion(task.id, task.timeoutSeconds * 1000);
-    } catch (error) {
-      // Process errors (including timeouts) map to the fail-closed `error` outcome.
-      const reason = error instanceof Error ? error.message : String(error);
-      manifest.stages[stageName] = {
-        status: 'error',
-        reason: `executor error: ${reason}`,
-        startedAt: manifest.stages[stageName]?.startedAt,
-        completedAt: nowISO(),
-        gateResults: [],
-      };
-      manifest.status = 'failed';
-      saveManifest(runDir, manifest);
-      throw error instanceof Error ? error : new Error(reason);
-    }
-
-    if (simulate) {
-      manifest.stages[stageName] = {
-        status: 'simulated',
-        reason: SIMULATED_STAGE_REASON,
-        startedAt: manifest.stages[stageName]?.startedAt,
-        completedAt: nowISO(),
-        gateResults: simulatedGateResults(spec.qualityGates),
-      };
-      saveManifest(runDir, manifest);
-      continue;
-    }
-
-    const evaluation = evaluateStageGates(stageName, spec.qualityGates, result);
-    manifest.stages[stageName] = {
-      status: evaluation.outcome,
-      reason: evaluation.reason,
-      startedAt: manifest.stages[stageName]?.startedAt,
-      completedAt: nowISO(),
-      gateResults: evaluation.gateResults,
-    };
-
-    if (evaluation.outcome !== 'passed') {
-      manifest.status =
-        evaluation.outcome === 'waiting-for-authority' ? 'waiting-for-authority' : 'failed';
-      saveManifest(runDir, manifest);
-      throw new Error(`Stage ${stageName} ${evaluation.outcome}: ${evaluation.reason}`);
-    }
-
-    saveManifest(runDir, manifest);
-  }
-}
-
 /**
 * Run the Forge pipeline.
 *
- * 1. Fail closed unless a real executor is wired or simulation is explicit
- * 2. Classify the brief
- * 3. Generate a run ID and create run directory
- * 4. Map stages to tasks and submit to TaskExecutor
- * 5. Track manifest with typed stage outcomes
- * 6. Return pipeline result
+ * 1. Classify the brief
+ * 2. Generate a run ID and create run directory
+ * 3. Map stages to tasks and submit to TaskExecutor
+ * 4. Track manifest with stage statuses
+ * 5. Return pipeline result
 */
 export async function runPipeline(
  briefPath: string,
  projectRoot: string,
  options: PipelineOptions,
 ): Promise<PipelineResult> {
-  const simulate = options.simulate ?? false;
-  const executor = requireExecutor(options.executor, simulate);
-  const mode: RunMode = simulate ? 'simulated' : 'normal';
-
  const resolvedRoot = path.resolve(projectRoot);
  const resolvedBrief = path.resolve(briefPath);
  const briefContent = fs.readFileSync(resolvedBrief, 'utf-8');
@@ -342,7 +146,6 @@ export async function runPipeline(
    briefClass,
    classSource,
    forceBoard: options.forceBoard ?? false,
-    mode,
    runDir,
  });

@@ -369,10 +172,54 @@ export async function runPipeline(
  }

  // Execute stages
-  await executeStages({ manifest, runDir, tasks, stageNames: selectedStages, executor, simulate });
+  const { executor } = options;
+  for (let i = 0; i < tasks.length; i++) {
+    const task = tasks[i]!;
+    const stageName = selectedStages[i]!;

-  // All stages reached a terminal state for this mode
-  manifest.status = simulate ? 'simulated' : 'completed';
+    // Update manifest: stage in progress
+    manifest.currentStage = stageName;
+    manifest.stages[stageName] = {
+      status: 'in_progress',
+      startedAt: nowISO(),
+    };
+    saveManifest(runDir, manifest);
+
+    try {
+      await executor.submitTask(task);
+      const result = await executor.waitForCompletion(task.id, task.timeoutSeconds * 1000);
+
+      // Update manifest: stage completed or failed
+      const stageStatus: StageStatus = {
+        status: result.status === 'completed' ? 'passed' : 'failed',
+        startedAt: manifest.stages[stageName]!.startedAt,
+        completedAt: nowISO(),
+      };
+      manifest.stages[stageName] = stageStatus;
+
+      if (result.status !== 'completed') {
+        manifest.status = 'failed';
+        saveManifest(runDir, manifest);
+        throw new Error(`Stage ${stageName} failed with status: ${result.status}`);
+      }
+
+      saveManifest(runDir, manifest);
+    } catch (error) {
+      if (!manifest.stages[stageName]?.completedAt) {
+        manifest.stages[stageName] = {
+          status: 'failed',
+          startedAt: manifest.stages[stageName]?.startedAt,
+          completedAt: nowISO(),
+        };
+      }
+      manifest.status = 'failed';
+      saveManifest(runDir, manifest);
+      throw error;
+    }
+  }
+
+  // All stages passed
+  manifest.status = 'completed';
  saveManifest(runDir, manifest);

  return {
@@ -387,30 +234,22 @@ export async function runPipeline(
 }

 /**
- * Resume a pipeline from the last non-passed stage.
+ * Resume a pipeline from the last incomplete stage.
 */
 export async function resumePipeline(
  runDir: string,
-  executor?: TaskExecutor,
-  options?: { simulate?: boolean },
+  executor: TaskExecutor,
 ): Promise<PipelineResult> {
-  const simulate = options?.simulate ?? false;
-  const wiredExecutor = requireExecutor(executor, simulate);
-  const mode: RunMode = simulate ? 'simulated' : 'normal';
-
  const manifest = loadManifest(runDir);
  const resolvedRoot = path.dirname(path.dirname(path.dirname(runDir))); // .forge/runs/{id} → project root

  const briefContent = fs.readFileSync(manifest.brief, 'utf-8');
  const allStages = stagesForClass(manifest.briefClass, manifest.forceBoard);

-  manifest.mode = mode;
-
-  // Find first non-satisfying stage (only a verified `passed` counts as done;
-  // simulated and waiting-for-authority stages are re-run).
+  // Find first non-passed stage
  const resumeFrom = allStages.find((s) => manifest.stages[s]?.status !== 'passed');
  if (!resumeFrom) {
-    manifest.status = mode === 'simulated' ? 'simulated' : 'completed';
+    manifest.status = 'completed';
    saveManifest(runDir, manifest);
    return {
      runId: manifest.runId,
@@ -445,16 +284,49 @@ export async function resumePipeline(
    tasks.push(task);
  }

-  await executeStages({
-    manifest,
-    runDir,
-    tasks,
-    stageNames: remainingStages,
-    executor: wiredExecutor,
-    simulate,
-  });
+  for (let i = 0; i < tasks.length; i++) {
+    const task = tasks[i]!;
+    const stageName = remainingStages[i]!;

-  manifest.status = simulate ? 'simulated' : 'completed';
+    manifest.currentStage = stageName;
+    manifest.stages[stageName] = {
+      status: 'in_progress',
+      startedAt: nowISO(),
+    };
+    saveManifest(runDir, manifest);
+
+    try {
+      await executor.submitTask(task);
+      const result = await executor.waitForCompletion(task.id, task.timeoutSeconds * 1000);
+
+      manifest.stages[stageName] = {
+        status: result.status === 'completed' ? 'passed' : 'failed',
+        startedAt: manifest.stages[stageName]!.startedAt,
+        completedAt: nowISO(),
+      };
+
+      if (result.status !== 'completed') {
+        manifest.status = 'failed';
+        saveManifest(runDir, manifest);
+        throw new Error(`Stage ${stageName} failed with status: ${result.status}`);
+      }
+
+      saveManifest(runDir, manifest);
+    } catch (error) {
+      if (!manifest.stages[stageName]?.completedAt) {
+        manifest.stages[stageName] = {
+          status: 'failed',
+          startedAt: manifest.stages[stageName]?.startedAt,
+          completedAt: nowISO(),
+        };
+      }
+      manifest.status = 'failed';
+      saveManifest(runDir, manifest);
+      throw error;
+    }
+  }
+
+  manifest.status = 'completed';
  saveManifest(runDir, manifest);

  return {
@@ -1,32 +0,0 @@
-import type { ForgeTask, ForgeTaskResult, TaskExecutor } from './types.js';
-
-/**
- * Simulated executor — used ONLY when the caller explicitly passes --simulate.
- *
- * It submits no real work and returns typed `simulated` results so a simulated
- * run can never be confused with a verified one. In normal mode (no --simulate)
- * the CLI refuses to run at all with FORGE_NO_EXECUTOR instead of wiring this
- * stub (fail closed, SDLC-D-035).
- */
-export function createSimulatedExecutor(options?: { log?: boolean }): TaskExecutor {
-  const log = options?.log ?? true;
-  return {
-    async submitTask(task: ForgeTask) {
-      if (log) console.log(`  [forge:simulated] stage submitted: ${task.id} (${task.title})`);
-    },
-    async waitForCompletion(taskId: string): Promise<ForgeTaskResult> {
-      if (log) console.log(`  [forge:simulated] stage complete:  ${taskId}`);
-      return {
-        task_id: taskId,
-        outcome: 'simulated',
-        reason: 'no executor wired; simulated execution requested via --simulate',
-        completed_at: new Date().toISOString(),
-        exit_code: 0,
-        gate_results: [],
-      };
-    },
-    async getTaskStatus() {
-      return 'completed' as const;
-    },
-  };
-}
@@ -1,4 +1,4 @@
-import type { GateEntry } from '@mosaicstack/macp';
+import type { GateEntry, TaskResult } from '@mosaicstack/macp';

 /** Stage dispatch mode. */
 export type StageDispatch = 'exec' | 'yolo' | 'pi';
@@ -6,58 +6,6 @@ export type StageDispatch = 'exec' | 'yolo' | 'pi';
 /** Stage type — determines agent selection and gate requirements. */
 export type StageType = 'research' | 'review' | 'coding' | 'deploy';

-/**
- * Typed outcome for every gate and stage evaluation — closed set (SDLC-D-035).
- *
- * Only `passed` means "verified by a real implementation". `simulated` is
- * produced exclusively in explicit `--simulate` runs and is never satisfying.
- */
-export type ForgeOutcome =
-  | 'passed'
-  | 'failed'
-  | 'blocked'
-  | 'error'
-  | 'waiting-for-authority'
-  | 'simulated'
-  | 'not-applicable';
-
-/** A gate that requires authority (human/board) sign-off; no mechanical command can satisfy it. */
-export interface AuthorityGate {
-  kind: 'authority';
-  capability: string;
-  reason: string;
-}
-
-/** A gate that requires a wired provider (e.g. an AI reviewer, CI pipeline) to evaluate. */
-export interface ProviderGate {
-  kind: 'provider';
-  capability: string;
-  reason: string;
-}
-
-/** Forge quality gate: a real command, an authority sign-off, or a provider-backed check. */
-export type ForgeGate = string | GateEntry | AuthorityGate | ProviderGate;
-
-/** Typed result of evaluating a single quality gate. */
-export interface ForgeGateResult {
-  gate: string;
-  outcome: ForgeOutcome;
-  reason: string;
-  exitCode?: number;
-  output?: string;
-  timedOut?: boolean;
-}
-
-/** Typed result of a task/stage execution returned by a TaskExecutor. */
-export interface ForgeTaskResult {
-  task_id: string;
-  outcome: ForgeOutcome;
-  reason: string;
-  completed_at: string;
-  exit_code: number;
-  gate_results: ForgeGateResult[];
-}
-
 /** Stage specification — defines a single pipeline stage. */
 export interface StageSpec {
  number: string;
@@ -66,7 +14,7 @@ export interface StageSpec {
  type: StageType;
  gate: string;
  promptFile: string;
-  qualityGates: ForgeGate[];
+  qualityGates: (string | GateEntry)[];
 }

 /** Brief classification. */
@@ -77,18 +25,11 @@ export type ClassSource = 'cli' | 'frontmatter' | 'auto';

 /** Per-stage status within a run manifest. */
 export interface StageStatus {
-  status: 'pending' | 'in_progress' | ForgeOutcome;
-  /** Why the stage reached its current (terminal) outcome, when applicable. */
-  reason?: string;
+  status: 'pending' | 'in_progress' | 'passed' | 'failed';
  startedAt?: string;
  completedAt?: string;
-  /** Typed per-gate results recorded alongside the stage outcome. */
-  gateResults?: ForgeGateResult[];
 }

-/** Execution mode of a run. */
-export type RunMode = 'normal' | 'simulated';
-
 /** Run manifest — persisted to disk as manifest.json. */
 export interface RunManifest {
  runId: string;
@@ -97,23 +38,10 @@ export interface RunManifest {
  briefClass: BriefClass;
  classSource: ClassSource;
  forceBoard: boolean;
-  /**
-   * Execution mode. `simulated` runs stub execution; their results are typed
-   * `simulated` and must never be read as verified success. Optional because
-   * manifests written before this field existed default to `normal`.
-   */
-  mode?: RunMode;
  createdAt: string;
  updatedAt: string;
  currentStage: string;
-  status:
-    | 'in_progress'
-    | 'completed'
-    | 'failed'
-    | 'interrupted'
-    | 'rejected'
-    | 'simulated'
-    | 'waiting-for-authority';
+  status: 'in_progress' | 'completed' | 'failed' | 'interrupted' | 'rejected';
  stages: Record<string, StageStatus>;
 }

@@ -137,7 +65,7 @@ export interface ForgeTask {
  briefPath: string;
  resultPath: string;
  timeoutSeconds: number;
-  qualityGates: ForgeGate[];
+  qualityGates: (string | GateEntry)[];
  worktree?: string;
  command?: string;
  dependsOn?: string[];
@@ -148,7 +76,7 @@ export interface ForgeTask {
 /** Abstract task executor — decouples from packages/coord. */
 export interface TaskExecutor {
  submitTask(task: ForgeTask): Promise<void>;
-  waitForCompletion(taskId: string, timeoutMs: number): Promise<ForgeTaskResult>;
+  waitForCompletion(taskId: string, timeoutMs: number): Promise<TaskResult>;
  getTaskStatus(taskId: string): Promise<ForgeTaskStatus>;
 }

@@ -194,16 +122,7 @@ export interface PipelineOptions {
  stages?: string[];
  skipTo?: string;
  dryRun?: boolean;
-  /**
-   * Real task executor. Required in normal mode: the pipeline fails closed
-   * with FORGE_NO_EXECUTOR when it is absent.
-   */
-  executor?: TaskExecutor;
-  /**
-   * Explicit opt-in to simulated execution. Every stage and gate result is
-   * typed `simulated` and is never satisfying.
-   */
-  simulate?: boolean;
+  executor: TaskExecutor;
 }

 /** Pipeline run result. */
@@ -1,253 +0,0 @@
-import { mkdirSync, readFileSync, rmSync } from 'node:fs';
-import { join } from 'node:path';
-import { tmpdir } from 'node:os';
-import { randomUUID } from 'node:crypto';
-import { describe, it, expect, beforeEach, afterEach } from 'vitest';
-import { normalizeGate, countAIFindings, runGate, runGates } from '../src/gate-runner.js';
-
-function makeTmpDir(): string {
-  const dir = join(tmpdir(), `macp-gate-${randomUUID()}`);
-  mkdirSync(dir, { recursive: true });
-  return dir;
-}
-
-describe('normalizeGate', () => {
-  it('normalizes a string to mechanical gate', () => {
-    expect(normalizeGate('echo test')).toEqual({
-      command: 'echo test',
-      type: 'mechanical',
-      fail_on: 'blocker',
-    });
-  });
-
-  it('normalizes an object gate with defaults', () => {
-    expect(normalizeGate({ command: 'lint' })).toEqual({
-      command: 'lint',
-      type: 'mechanical',
-      fail_on: 'blocker',
-    });
-  });
-
-  it('preserves explicit type and fail_on', () => {
-    expect(normalizeGate({ command: 'review', type: 'ai-review', fail_on: 'any' })).toEqual({
-      command: 'review',
-      type: 'ai-review',
-      fail_on: 'any',
-    });
-  });
-
-  it('handles non-string/non-object input', () => {
-    expect(normalizeGate(42)).toEqual({ command: '', type: 'mechanical', fail_on: 'blocker' });
-    expect(normalizeGate(null)).toEqual({ command: '', type: 'mechanical', fail_on: 'blocker' });
-  });
-});
-
-describe('countAIFindings', () => {
-  it('returns zeros for non-object', () => {
-    expect(countAIFindings(null)).toEqual({ blockers: 0, total: 0 });
-    expect(countAIFindings('string')).toEqual({ blockers: 0, total: 0 });
-    expect(countAIFindings([])).toEqual({ blockers: 0, total: 0 });
-  });
-
-  it('counts from stats block', () => {
-    const output = { stats: { blockers: 2, should_fix: 3, suggestions: 1 } };
-    expect(countAIFindings(output)).toEqual({ blockers: 2, total: 6 });
-  });
-
-  it('counts from findings array when stats has no blockers', () => {
-    const output = {
-      stats: { blockers: 0 },
-      findings: [{ severity: 'blocker' }, { severity: 'warning' }, { severity: 'blocker' }],
-    };
-    expect(countAIFindings(output)).toEqual({ blockers: 2, total: 3 });
-  });
-
-  it('uses stats blockers over findings array when stats has blockers', () => {
-    const output = {
-      stats: { blockers: 5 },
-      findings: [{ severity: 'blocker' }, { severity: 'warning' }],
-    };
-    // stats.blockers = 5, total from stats = 5+0+0 = 5, findings not used for total since stats total is non-zero
-    expect(countAIFindings(output)).toEqual({ blockers: 5, total: 5 });
-  });
-
-  it('counts findings length as total when stats has zero total', () => {
-    const output = {
-      findings: [{ severity: 'warning' }, { severity: 'info' }],
-    };
-    expect(countAIFindings(output)).toEqual({ blockers: 0, total: 2 });
-  });
-});
-
-describe('runGate', () => {
-  let tmp: string;
-  let logPath: string;
-
-  beforeEach(() => {
-    tmp = makeTmpDir();
-    logPath = join(tmp, 'gate.log');
-  });
-
-  afterEach(() => {
-    rmSync(tmp, { recursive: true, force: true });
-  });
-
-  it('passes mechanical gate on exit 0', () => {
-    const result = runGate('echo hello', tmp, logPath, 30);
-    expect(result.passed).toBe(true);
-    expect(result.exit_code).toBe(0);
-    expect(result.type).toBe('mechanical');
-    expect(result.output).toContain('hello');
-  });
-
-  it('fails mechanical gate on non-zero exit', () => {
-    const result = runGate('exit 1', tmp, logPath, 30);
-    expect(result.passed).toBe(false);
-    expect(result.exit_code).toBe(1);
-  });
-
-  it('ci-pipeline always passes', () => {
-    const result = runGate({ command: 'anything', type: 'ci-pipeline' }, tmp, logPath, 30);
-    expect(result.passed).toBe(true);
-    expect(result.type).toBe('ci-pipeline');
-    expect(result.output).toBe('CI pipeline gate placeholder');
-  });
-
-  it('empty command passes', () => {
-    const result = runGate({ command: '' }, tmp, logPath, 30);
-    expect(result.passed).toBe(true);
-  });
-
-  it('ai-review gate parses JSON output', () => {
-    const json = JSON.stringify({ stats: { blockers: 0, should_fix: 1 } });
-    const result = runGate({ command: `echo '${json}'`, type: 'ai-review' }, tmp, logPath, 30);
-    expect(result.passed).toBe(true);
-    expect(result.blockers).toBe(0);
-    expect(result.findings).toBe(1);
-  });
-
-  it('ai-review gate fails on blockers', () => {
-    const json = JSON.stringify({ stats: { blockers: 2 } });
-    const result = runGate({ command: `echo '${json}'`, type: 'ai-review' }, tmp, logPath, 30);
-    expect(result.passed).toBe(false);
-    expect(result.blockers).toBe(2);
-  });
-
-  it('ai-review gate with fail_on=any fails on any findings', () => {
-    const json = JSON.stringify({ stats: { blockers: 0, should_fix: 1 } });
-    const result = runGate(
-      { command: `echo '${json}'`, type: 'ai-review', fail_on: 'any' },
-      tmp,
-      logPath,
-      30,
-    );
-    expect(result.passed).toBe(false);
-    expect(result.fail_on).toBe('any');
-  });
-
-  it('ai-review gate fails on invalid JSON output', () => {
-    const result = runGate({ command: 'echo "not json"', type: 'ai-review' }, tmp, logPath, 30);
-    expect(result.passed).toBe(false);
-    expect(result.parse_error).toBeDefined();
-  });
-
-  it('writes to log file', () => {
-    runGate('echo logged', tmp, logPath, 30);
-    const log = readFileSync(logPath, 'utf-8');
-    expect(log).toContain('COMMAND: echo logged');
-    expect(log).toContain('logged');
-    expect(log).toContain('EXIT:');
-  });
-});
-
-describe('runGates', () => {
-  let tmp: string;
-  let logPath: string;
-  let eventsPath: string;
-
-  beforeEach(() => {
-    tmp = makeTmpDir();
-    logPath = join(tmp, 'gates.log');
-    eventsPath = join(tmp, 'events.ndjson');
-  });
-
-  afterEach(() => {
-    rmSync(tmp, { recursive: true, force: true });
-  });
-
-  it('runs multiple gates and returns results', () => {
-    const { allPassed, gateResults } = runGates(
-      ['echo one', 'echo two'],
-      tmp,
-      logPath,
-      30,
-      eventsPath,
-      'task-1',
-    );
-    expect(allPassed).toBe(true);
-    expect(gateResults).toHaveLength(2);
-  });
-
-  it('reports failure when any gate fails', () => {
-    const { allPassed, gateResults } = runGates(
-      ['echo ok', 'exit 1'],
-      tmp,
-      logPath,
-      30,
-      eventsPath,
-      'task-2',
-    );
-    expect(allPassed).toBe(false);
-    expect(gateResults[0]!.passed).toBe(true);
-    expect(gateResults[1]!.passed).toBe(false);
-  });
-
-  it('emits events for each gate', () => {
-    runGates(['echo test'], tmp, logPath, 30, eventsPath, 'task-3');
-    const events = readFileSync(eventsPath, 'utf-8')
-      .trim()
-      .split('\n')
-      .map((l) => JSON.parse(l));
-    expect(events).toHaveLength(2); // started + passed
-    expect(events[0].event_type).toBe('rail.check.started');
-    expect(events[1].event_type).toBe('rail.check.passed');
-  });
-
-  it('skips gates with empty command (non ci-pipeline)', () => {
-    const { gateResults } = runGates(
-      [{ command: '', type: 'mechanical' }, 'echo real'],
-      tmp,
-      logPath,
-      30,
-      eventsPath,
-      'task-4',
-    );
-    expect(gateResults).toHaveLength(1);
-  });
-
-  it('does not skip ci-pipeline even with empty command', () => {
-    const { gateResults } = runGates(
-      [{ command: '', type: 'ci-pipeline' }],
-      tmp,
-      logPath,
-      30,
-      eventsPath,
-      'task-5',
-    );
-    expect(gateResults).toHaveLength(1);
-    expect(gateResults[0]!.passed).toBe(true);
-  });
-
-  it('emits failed event with correct message', () => {
-    runGates(['exit 42'], tmp, logPath, 30, eventsPath, 'task-6');
-    const events = readFileSync(eventsPath, 'utf-8')
-      .trim()
-      .split('\n')
-      .map((l) => JSON.parse(l));
-    const failEvent = events.find(
-      (e: Record<string, unknown>) => e.event_type === 'rail.check.failed',
-    );
-    expect(failEvent).toBeDefined();
-    expect(failEvent.message).toContain('Gate failed (');
-  });
-});
@@ -1,5 +1,8 @@
-import { describe, it, expect } from 'vitest';
+import { describe, it, expect, afterEach, beforeEach, vi } from 'vitest';
 import { Command } from 'commander';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
 import { registerMacpCommand } from './cli.js';

 describe('registerMacpCommand', () => {
@@ -75,3 +78,162 @@ describe('registerMacpCommand', () => {
    expect(topLevel).toContain('events');
  });
 });
+
+/**
+ * RI-N2 fail-closed CLI behavior: an unimplemented capability is a failure,
+ * never a success. Every stub exits nonzero with a typed message, and the
+ * implemented `macp gate` mirrors the typed gate-runner states.
+ */
+describe('registerMacpCommand fail-closed (RI-N2)', () => {
+  let tmpDir: string;
+
+  function buildProgram(): Command {
+    const program = new Command();
+    program.exitOverride();
+    program.configureOutput({ writeErr: () => {} });
+    registerMacpCommand(program);
+    return program;
+  }
+
+  beforeEach(() => {
+    tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'macp-cli-failclosed-'));
+    process.exitCode = 0;
+  });
+
+  afterEach(() => {
+    process.exitCode = 0;
+    fs.rmSync(tmpDir, { recursive: true, force: true });
+  });
+
+  it('macp tasks list exits nonzero (unimplemented capability)', async () => {
+    const program = buildProgram();
+    await program.parseAsync(['macp', 'tasks', 'list'], { from: 'user' });
+    expect(process.exitCode).not.toBe(0);
+  });
+
+  it('macp submit exits nonzero with a typed MACP_NOT_IMPLEMENTED message', async () => {
+    const program = buildProgram();
+    const errSpy = vi.spyOn(console, 'error').mockImplementation(() => {});
+    try {
+      await program.parseAsync(['macp', 'submit', 'spec.json'], { from: 'user' });
+      expect(process.exitCode).not.toBe(0);
+      const errText = errSpy.mock.calls.map((c) => String(c[0])).join('\n');
+      expect(errText).toContain('MACP_NOT_IMPLEMENTED');
+    } finally {
+      errSpy.mockRestore();
+    }
+  });
+
+  it('macp events tail exits nonzero (unimplemented capability)', async () => {
+    const program = buildProgram();
+    await program.parseAsync(['macp', 'events', 'tail'], { from: 'user' });
+    expect(process.exitCode).not.toBe(0);
+  });
+
+  it('macp gate runs a green inline command and exits 0', async () => {
+    const program = buildProgram();
+    await program.parseAsync(
+      [
+        'macp',
+        'gate',
+        'exit 0',
+        '--cwd',
+        tmpDir,
+        '--log',
+        path.join(tmpDir, 'g.log'),
+        '--timeout',
+        '10',
+      ],
+      { from: 'user' },
+    );
+    expect(process.exitCode).toBe(0);
+  });
+
+  it('macp gate exits nonzero on a failing command', async () => {
+    const program = buildProgram();
+    await program.parseAsync(
+      [
+        'macp',
+        'gate',
+        'exit 9',
+        '--cwd',
+        tmpDir,
+        '--log',
+        path.join(tmpDir, 'g.log'),
+        '--timeout',
+        '10',
+      ],
+      { from: 'user' },
+    );
+    expect(process.exitCode).not.toBe(0);
+  });
+
+  it('macp gate with an unimplemented ci-pipeline capability exits nonzero', async () => {
+    const program = buildProgram();
+    const specPath = path.join(tmpDir, 'gates.json');
+    fs.writeFileSync(specPath, JSON.stringify([{ type: 'ci-pipeline' }]));
+    await program.parseAsync(
+      [
+        'macp',
+        'gate',
+        specPath,
+        '--cwd',
+        tmpDir,
+        '--log',
+        path.join(tmpDir, 'g.log'),
+        '--timeout',
+        '10',
+      ],
+      { from: 'user' },
+    );
+    expect(process.exitCode).not.toBe(0);
+  });
+
+  it('macp gate --simulate completes (exit 0) but reports simulated results', async () => {
+    const program = buildProgram();
+    const logSpy = vi.spyOn(console, 'log').mockImplementation(() => {});
+    try {
+      await program.parseAsync(
+        [
+          'macp',
+          'gate',
+          'exit 0',
+          '--simulate',
+          '--cwd',
+          tmpDir,
+          '--log',
+          path.join(tmpDir, 'g.log'),
+          '--timeout',
+          '10',
+        ],
+        { from: 'user' },
+      );
+      // completes only because the caller explicitly asked to simulate
+      expect(process.exitCode).toBe(0);
+      const outText = logSpy.mock.calls.map((c) => String(c[0])).join('\n');
+      expect(outText).toContain('simulated');
+      expect(outText).toContain('SIMULATED');
+    } finally {
+      logSpy.mockRestore();
+    }
+  });
+
+  it('macp gate with an empty spec exits nonzero with a typed error', async () => {
+    const program = buildProgram();
+    await program.parseAsync(
+      [
+        'macp',
+        'gate',
+        '   ',
+        '--cwd',
+        tmpDir,
+        '--log',
+        path.join(tmpDir, 'g.log'),
+        '--timeout',
+        '10',
+      ],
+      { from: 'user' },
+    );
+    expect(process.exitCode).not.toBe(0);
+  });
+});
@@ -1,5 +1,73 @@
+import { existsSync, readFileSync } from 'node:fs';
+
 import type { Command } from 'commander';

+import { runGates } from './gate-runner.js';
+import { MACPCapabilityError, type MacpErrorCode } from './errors.js';
+
+/**
+ * Load gates from a spec: an existing file (JSON gates array, a JSON object
+ * with `quality_gates`, a JSON gate object, or one command per line) or an
+ * inline command string. Fails closed with a typed capability error when the
+ * spec contains no executable gate definition.
+ */
+function loadGateSpec(spec: string): unknown[] {
+  if (existsSync(spec)) {
+    const raw = readFileSync(spec, 'utf-8');
+    try {
+      const parsed = JSON.parse(raw) as unknown;
+      if (Array.isArray(parsed)) {
+        if (parsed.length === 0) {
+          throw new MACPCapabilityError(
+            'MACP_NO_COMMAND',
+            'gate-spec',
+            `gate spec file '${spec}' contains an empty gates array`,
+          );
+        }
+        return parsed;
+      }
+      if (typeof parsed === 'object' && parsed !== null) {
+        const obj = parsed as Record<string, unknown>;
+        if (Array.isArray(obj['quality_gates'])) {
+          return obj['quality_gates'];
+        }
+        return [parsed];
+      }
+      throw new MACPCapabilityError(
+        'MACP_NO_COMMAND',
+        'gate-spec',
+        `gate spec file '${spec}' parsed to ${typeof parsed} — expected a gates array, a task with quality_gates, or a gate object`,
+      );
+    } catch (exc) {
+      if (exc instanceof MACPCapabilityError) throw exc;
+      // Not JSON — treat each non-empty line as a command gate.
+      const lines = raw
+        .split('\n')
+        .map((l) => l.trim())
+        .filter((l) => l.length > 0);
+      if (lines.length > 0) return lines;
+      throw new MACPCapabilityError(
+        'MACP_NO_COMMAND',
+        'gate-spec',
+        `gate spec file '${spec}' contains no gates`,
+      );
+    }
+  }
+  if (spec.trim().length > 0) return [spec];
+  throw new MACPCapabilityError('MACP_NO_COMMAND', 'gate-spec', 'gate spec is empty');
+}
+
+/** Print a typed not-implemented failure and exit nonzero (RI-N2 fail-closed). */
+function notImplemented(subcommand: string, capability: string, hint: string): void {
+  const err = new MACPCapabilityError(
+    'MACP_NOT_IMPLEMENTED',
+    capability,
+    `${subcommand} is not implemented in @mosaicstack/macp yet (${capability} capability absent) — ${hint}`,
+  );
+  console.error(`[macp] ${subcommand}: ${err.message} [${err.code}]`);
+  process.exitCode = 1;
+}
+
 /**
 * Register macp subcommands on an existing Commander program.
 * This avoids cross-package Commander version mismatches by using the
@@ -24,15 +92,14 @@ export function registerMacpCommand(parent: Command): void {
      'Filter by task type (coding|deploy|research|review|documentation|infrastructure)',
    )
    .action((opts: { status?: string; type?: string }) => {
-      // not yet wired — task persistence layer is not present in @mosaicstack/macp
-      console.log('[macp] tasks list: not yet wired — use macp package programmatically');
+      // unimplemented capability — a failure, never a success (RI-N2)
      if (opts.status) {
        console.log(`  status filter: ${opts.status}`);
      }
      if (opts.type) {
        console.log(`  type filter: ${opts.type}`);
      }
-      process.exitCode = 0;
+      notImplemented('tasks list', 'task-persistence', 'use the macp package programmatically');
    });

  // ─── submit ──────────────────────────────────────────────────────────────
@@ -41,12 +108,11 @@ export function registerMacpCommand(parent: Command): void {
    .command('submit <path>')
    .description('Submit a task from a JSON/YAML spec file')
    .action((specPath: string) => {
-      // not yet wired — task submission requires a running MACP server
-      console.log('[macp] submit: not yet wired — use macp package programmatically');
+      // unimplemented capability — a failure, never a success (RI-N2)
      console.log(`  spec path: ${specPath}`);
      console.log('  task id:   (unavailable — no MACP server connected)');
      console.log('  status:    (unavailable — no MACP server connected)');
-      process.exitCode = 0;
+      notImplemented('submit', 'macp-server', 'use the macp package programmatically');
    });

  // ─── gate ────────────────────────────────────────────────────────────────
@@ -58,16 +124,58 @@ export function registerMacpCommand(parent: Command): void {
    .option('--cwd <path>', 'Working directory for gate execution', process.cwd())
    .option('--log <path>', 'Path to write gate log output', '/tmp/macp-gate.log')
    .option('--timeout <seconds>', 'Gate timeout in seconds', '60')
-    .action((spec: string, opts: { failOn: string; cwd: string; log: string; timeout: string }) => {
-      // not yet wired — gate execution requires a task context and event sink
-      console.log('[macp] gate: not yet wired — use macp package programmatically');
-      console.log(`  spec:     ${spec}`);
-      console.log(`  fail-on:  ${opts.failOn}`);
-      console.log(`  cwd:      ${opts.cwd}`);
-      console.log(`  log:      ${opts.log}`);
-      console.log(`  timeout:  ${opts.timeout}s`);
-      process.exitCode = 0;
-    });
+    .option(
+      '--simulate',
+      'Simulate gates instead of executing them; results are typed simulated and never satisfy a check',
+    )
+    .action(
+      (
+        spec: string,
+        opts: { failOn: string; cwd: string; log: string; timeout: string; simulate?: boolean },
+      ) => {
+        let gates: unknown[];
+        try {
+          gates = loadGateSpec(spec);
+        } catch (exc) {
+          if (exc instanceof MACPCapabilityError) {
+            console.error(`[macp] gate: ${exc.message} [${exc.code}]`);
+          } else {
+            console.error(`[macp] gate: ${String(exc)}`);
+          }
+          process.exitCode = 1;
+          return;
+        }
+
+        const timeoutSec = Number.parseInt(opts.timeout, 10) || 60;
+        const eventsPath = `${opts.log}.events.ndjson`;
+        const { state, gateResults } = runGates(
+          gates,
+          opts.cwd,
+          opts.log,
+          timeoutSec,
+          eventsPath,
+          'macp-cli-gate',
+          {
+            simulate: opts.simulate,
+          },
+        );
+
+        for (const r of gateResults) {
+          const label = r.command || r.type;
+          const reason = r.reason ? ` — ${r.reason}` : '';
+          console.log(`[macp] gate ${r.status}: ${label}${reason}`);
+        }
+        if (opts.simulate) {
+          console.log(
+            '[macp] SIMULATED run — every result is typed simulated and can never satisfy a gate, dependency, or release check',
+          );
+        }
+
+        // Simulated runs may complete (exit 0) only because the caller
+        // explicitly passed --simulate; the typed state stays 'simulated'.
+        process.exitCode = state === 'passed' || state === 'simulated' ? 0 : 1;
+      },
+    );

  // ─── events ──────────────────────────────────────────────────────────────

@@ -79,14 +187,16 @@ export function registerMacpCommand(parent: Command): void {
    .option('--file <path>', 'Path to the MACP events NDJSON file')
    .option('--follow', 'Follow the file for new events (like tail -f)')
    .action((opts: { file?: string; follow?: boolean }) => {
-      // not yet wired — event streaming requires a live event source
-      console.log('[macp] events tail: not yet wired — use macp package programmatically');
+      // unimplemented capability — a failure, never a success (RI-N2)
      if (opts.file) {
        console.log(`  file:   ${opts.file}`);
      }
      if (opts.follow) {
        console.log('  mode:   follow');
      }
-      process.exitCode = 0;
+      notImplemented('events tail', 'event-source', 'use the macp package programmatically');
    });
 }
+
+// Re-export so CLI consumers can surface typed capability codes.
+export type { MacpErrorCode };
@@ -0,0 +1,35 @@
+/** Typed error code from the closed MACP_ERROR_CODES set. */
+export type MacpErrorCode = (typeof MACP_ERROR_CODES)[number];
+/**
+ * Typed fail-closed capability errors (RI-N2, SDLC-D-035).
+ *
+ * MACP must fail closed when a required capability (executor, reviewer,
+ * command, CI provider, human authority) is absent. These typed codes mirror
+ * the Forge failure vocabulary (FORGE_NO_*) so both packages speak the same
+ * language: an unimplemented capability is a failure, never a stub success.
+ */
+
+/** Closed set of typed MACP capability error codes. */
+export const MACP_ERROR_CODES = [
+  'MACP_NOT_IMPLEMENTED',
+  'MACP_NO_COMMAND',
+  'MACP_NO_REVIEWER',
+  'MACP_NO_CI_PIPELINE',
+  'MACP_NO_PROVIDER',
+  'MACP_AUTHORITY_REQUIRED',
+] as const;
+
+/** Raised when a required capability is missing and execution must fail closed. */
+export class MACPCapabilityError extends Error {
+  /** Typed error code from the closed MACP_ERROR_CODES set. */
+  readonly code: MacpErrorCode;
+  /** The missing capability, e.g. `ci-provider`, `task-persistence`, `command`. */
+  readonly capability: string;
+
+  constructor(code: MacpErrorCode, capability: string, message: string) {
+    super(message);
+    this.name = 'MACPCapabilityError';
+    this.code = code;
+    this.capability = capability;
+  }
+}
@@ -0,0 +1,429 @@
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import { afterEach, beforeEach, describe, expect, it } from 'vitest';
+
+import { countAIFindings, normalizeGate, runGate, runGates } from './gate-runner.js';
+
+function makeTmpDir(): string {
+  return fs.mkdtempSync(path.join(os.tmpdir(), 'macp-gate-'));
+}
+
+describe('normalizeGate', () => {
+  it('normalizes a string to mechanical gate', () => {
+    expect(normalizeGate('echo test')).toEqual({
+      command: 'echo test',
+      type: 'mechanical',
+      fail_on: 'blocker',
+    });
+  });
+
+  it('normalizes an object gate with defaults', () => {
+    expect(normalizeGate({ command: 'lint' })).toEqual({
+      command: 'lint',
+      type: 'mechanical',
+      fail_on: 'blocker',
+    });
+  });
+
+  it('preserves explicit type and fail_on', () => {
+    expect(normalizeGate({ command: 'review', type: 'ai-review', fail_on: 'any' })).toEqual({
+      command: 'review',
+      type: 'ai-review',
+      fail_on: 'any',
+    });
+  });
+
+  it('handles non-string/non-object input', () => {
+    expect(normalizeGate(42)).toEqual({ command: '', type: 'mechanical', fail_on: 'blocker' });
+    expect(normalizeGate(null)).toEqual({ command: '', type: 'mechanical', fail_on: 'blocker' });
+  });
+});
+
+describe('countAIFindings', () => {
+  it('returns zeros for non-object', () => {
+    expect(countAIFindings(null)).toEqual({ blockers: 0, total: 0 });
+    expect(countAIFindings('string')).toEqual({ blockers: 0, total: 0 });
+    expect(countAIFindings([])).toEqual({ blockers: 0, total: 0 });
+  });
+
+  it('counts from stats block', () => {
+    const output = { stats: { blockers: 2, should_fix: 3, suggestions: 1 } };
+    expect(countAIFindings(output)).toEqual({ blockers: 2, total: 6 });
+  });
+
+  it('counts from findings array when stats has no blockers', () => {
+    const output = {
+      stats: { blockers: 0 },
+      findings: [{ severity: 'blocker' }, { severity: 'warning' }, { severity: 'blocker' }],
+    };
+    expect(countAIFindings(output)).toEqual({ blockers: 2, total: 3 });
+  });
+
+  it('uses stats blockers over findings array when stats has blockers', () => {
+    const output = {
+      stats: { blockers: 5 },
+      findings: [{ severity: 'blocker' }, { severity: 'warning' }],
+    };
+    // stats.blockers = 5, total from stats = 5+0+0 = 5, findings not used for total since stats total is non-zero
+    expect(countAIFindings(output)).toEqual({ blockers: 5, total: 5 });
+  });
+
+  it('counts findings length as total when stats has zero total', () => {
+    const output = {
+      findings: [{ severity: 'warning' }, { severity: 'info' }],
+    };
+    expect(countAIFindings(output)).toEqual({ blockers: 0, total: 2 });
+  });
+});
+
+describe('runGate', () => {
+  let tmp: string;
+  let logPath: string;
+
+  beforeEach(() => {
+    tmp = makeTmpDir();
+    logPath = path.join(tmp, 'gate.log');
+  });
+
+  afterEach(() => {
+    fs.rmSync(tmp, { recursive: true, force: true });
+  });
+
+  it('passes mechanical gate on exit 0', () => {
+    const result = runGate('echo hello', tmp, logPath, 30);
+    expect(result.passed).toBe(true);
+    expect(result.exit_code).toBe(0);
+    expect(result.type).toBe('mechanical');
+    expect(result.output).toContain('hello');
+  });
+
+  it('fails mechanical gate on non-zero exit', () => {
+    const result = runGate('exit 1', tmp, logPath, 30);
+    expect(result.passed).toBe(false);
+    expect(result.exit_code).toBe(1);
+  });
+
+  it('ci-pipeline fails closed without a CI provider (no placeholder pass)', () => {
+    const result = runGate({ command: 'anything', type: 'ci-pipeline' }, tmp, logPath, 30);
+    expect(result.passed).toBe(false);
+    expect(result.status).toBe('capability_failure');
+    expect(result.capability_code).toBe('MACP_NO_CI_PIPELINE');
+    expect(result.type).toBe('ci-pipeline');
+    expect(result.output).not.toBe('CI pipeline gate placeholder');
+  });
+
+  it('empty command is a typed capability failure, never a pass', () => {
+    const result = runGate({ command: '' }, tmp, logPath, 30);
+    expect(result.passed).toBe(false);
+    expect(result.status).toBe('capability_failure');
+    expect(result.capability_code).toBe('MACP_NO_COMMAND');
+  });
+
+  it('ai-review gate parses JSON output', () => {
+    const json = JSON.stringify({ stats: { blockers: 0, should_fix: 1 } });
+    const result = runGate({ command: `echo '${json}'`, type: 'ai-review' }, tmp, logPath, 30);
+    expect(result.passed).toBe(true);
+    expect(result.blockers).toBe(0);
+    expect(result.findings).toBe(1);
+  });
+
+  it('ai-review gate fails on blockers', () => {
+    const json = JSON.stringify({ stats: { blockers: 2 } });
+    const result = runGate({ command: `echo '${json}'`, type: 'ai-review' }, tmp, logPath, 30);
+    expect(result.passed).toBe(false);
+    expect(result.blockers).toBe(2);
+  });
+
+  it('ai-review gate with fail_on=any fails on any findings', () => {
+    const json = JSON.stringify({ stats: { blockers: 0, should_fix: 1 } });
+    const result = runGate(
+      { command: `echo '${json}'`, type: 'ai-review', fail_on: 'any' },
+      tmp,
+      logPath,
+      30,
+    );
+    expect(result.passed).toBe(false);
+    expect(result.fail_on).toBe('any');
+  });
+
+  it('ai-review gate fails on invalid JSON output', () => {
+    const result = runGate({ command: 'echo "not json"', type: 'ai-review' }, tmp, logPath, 30);
+    expect(result.passed).toBe(false);
+    expect(result.parse_error).toBeDefined();
+  });
+
+  it('writes to log file', () => {
+    runGate('echo logged', tmp, logPath, 30);
+    const log = fs.readFileSync(logPath, 'utf-8');
+    expect(log).toContain('COMMAND: echo logged');
+    expect(log).toContain('logged');
+    expect(log).toContain('EXIT:');
+  });
+});
+
+describe('runGates', () => {
+  let tmp: string;
+  let logPath: string;
+  let eventsPath: string;
+
+  beforeEach(() => {
+    tmp = makeTmpDir();
+    logPath = path.join(tmp, 'gates.log');
+    eventsPath = path.join(tmp, 'events.ndjson');
+  });
+
+  afterEach(() => {
+    fs.rmSync(tmp, { recursive: true, force: true });
+  });
+
+  it('runs multiple gates and returns results', () => {
+    const { allPassed, gateResults } = runGates(
+      ['echo one', 'echo two'],
+      tmp,
+      logPath,
+      30,
+      eventsPath,
+      'task-1',
+    );
+    expect(allPassed).toBe(true);
+    expect(gateResults).toHaveLength(2);
+  });
+
+  it('reports failure when any gate fails', () => {
+    const { allPassed, gateResults } = runGates(
+      ['echo ok', 'exit 1'],
+      tmp,
+      logPath,
+      30,
+      eventsPath,
+      'task-2',
+    );
+    expect(allPassed).toBe(false);
+    expect(gateResults[0]!.passed).toBe(true);
+    expect(gateResults[1]!.passed).toBe(false);
+  });
+
+  it('emits events for each gate', () => {
+    runGates(['echo test'], tmp, logPath, 30, eventsPath, 'task-3');
+    const events = fs
+      .readFileSync(eventsPath, 'utf-8')
+      .trim()
+      .split('\n')
+      .map((l) => JSON.parse(l));
+    expect(events).toHaveLength(2); // started + passed
+    expect(events[0].event_type).toBe('rail.check.started');
+    expect(events[1].event_type).toBe('rail.check.passed');
+  });
+
+  it('does not silently skip gates with empty command — they become capability failures', () => {
+    const { gateResults, allPassed, state } = runGates(
+      [{ command: '', type: 'mechanical' }, 'echo real'],
+      tmp,
+      logPath,
+      30,
+      eventsPath,
+      'task-4',
+    );
+    expect(gateResults).toHaveLength(2);
+    expect(gateResults[0]!.status).toBe('capability_failure');
+    expect(gateResults[1]!.status).toBe('passed');
+    expect(allPassed).toBe(false);
+    expect(state).toBe('capability_failure');
+  });
+
+  it('does not skip ci-pipeline even with empty command — typed capability failure', () => {
+    const { gateResults, allPassed, state } = runGates(
+      [{ command: '', type: 'ci-pipeline' }],
+      tmp,
+      logPath,
+      30,
+      eventsPath,
+      'task-5',
+    );
+    expect(gateResults).toHaveLength(1);
+    expect(gateResults[0]!.passed).toBe(false);
+    expect(gateResults[0]!.status).toBe('capability_failure');
+    expect(allPassed).toBe(false);
+    expect(state).toBe('capability_failure');
+  });
+
+  it('emits failed event with correct message', () => {
+    runGates(['exit 42'], tmp, logPath, 30, eventsPath, 'task-6');
+    const events = fs
+      .readFileSync(eventsPath, 'utf-8')
+      .trim()
+      .split('\n')
+      .map((l) => JSON.parse(l));
+    const failEvent = events.find(
+      (e: Record<string, unknown>) => e.event_type === 'rail.check.failed',
+    );
+    expect(failEvent).toBeDefined();
+    expect(failEvent.message).toContain('Gate failed (');
+  });
+});
+
+/**
+ * RI-N2 / SDLC-D-035 fail-closed controls for the MACP gate runner.
+ *
+ * Invariant under test: `passed: true` occurs ONLY when a gate really executed
+ * and really exited green (`status === 'passed'`). Absent capabilities,
+ * manual sign-offs, and simulated runs are typed distinctly and can never
+ * make the aggregate `passed`.
+ */
+describe('gate-runner fail-closed (RI-N2)', () => {
+  let tmpDir: string;
+  let logPath: string;
+  let eventsPath: string;
+
+  beforeEach(() => {
+    tmpDir = makeTmpDir();
+    logPath = path.join(tmpDir, 'gate.log');
+    eventsPath = path.join(tmpDir, 'events.ndjson');
+  });
+
+  afterEach(() => {
+    fs.rmSync(tmpDir, { recursive: true, force: true });
+  });
+
+  function run(gates: unknown[], options?: { simulate?: boolean }) {
+    return runGates(gates, tmpDir, logPath, 10, eventsPath, 'spec-task', options);
+  }
+
+  // ─── positive controls ───────────────────────────────────────────────────
+
+  it('a really-executed green command gate still passes', () => {
+    const result = run([{ command: 'exit 0', type: 'mechanical' }]);
+    expect(result.gateResults[0]!.status).toBe('passed');
+    expect(result.gateResults[0]!.passed).toBe(true);
+    expect(result.allPassed).toBe(true);
+    expect(result.state).toBe('passed');
+  });
+
+  it('explicit simulate completes and types every result simulated', () => {
+    const result = run([{ command: 'exit 0', type: 'mechanical' }, 'echo hello'], {
+      simulate: true,
+    });
+    expect(result.gateResults).toHaveLength(2);
+    for (const gate of result.gateResults) {
+      expect(gate.status).toBe('simulated');
+      expect(gate.passed).toBe(false);
+    }
+    expect(result.state).toBe('simulated');
+  });
+
+  it('a really-executed red command gate fails with typed status failed', () => {
+    const result = run([{ command: 'exit 3', type: 'mechanical' }]);
+    expect(result.gateResults[0]!.status).toBe('failed');
+    expect(result.gateResults[0]!.passed).toBe(false);
+    expect(result.allPassed).toBe(false);
+    expect(result.state).toBe('failed');
+  });
+
+  // ─── negative controls — each asserts typed status AND aggregate not passed ──
+
+  it('an empty-command gate is a capability_failure, not skipped and not passed', () => {
+    const result = run([{ command: '', type: 'mechanical' }]);
+    // runGates must not silently skip it — it produces a typed result
+    expect(result.gateResults).toHaveLength(1);
+    const gate = result.gateResults[0]!;
+    expect(gate.status).toBe('capability_failure');
+    expect(gate.capability_code).toBe('MACP_NO_COMMAND');
+    expect(gate.passed).toBe(false);
+    // aggregate is not passed
+    expect(result.allPassed).toBe(false);
+    expect(result.state).toBe('capability_failure');
+    expect(result.state).not.toBe('passed');
+  });
+
+  it('a commandless ai-review gate is a typed MACP_NO_REVIEWER capability_failure', () => {
+    const result = run([{ command: '', type: 'ai-review' }]);
+    expect(result.gateResults[0]!.status).toBe('capability_failure');
+    expect(result.gateResults[0]!.capability_code).toBe('MACP_NO_REVIEWER');
+    expect(result.allPassed).toBe(false);
+    expect(result.state).not.toBe('passed');
+  });
+
+  it('a ci-pipeline gate without a provider implementation is a capability_failure, never a placeholder pass', () => {
+    const result = run([{ command: '', type: 'ci-pipeline' }]);
+    const gate = result.gateResults[0]!;
+    expect(gate.status).toBe('capability_failure');
+    expect(gate.capability_code).toBe('MACP_NO_CI_PIPELINE');
+    expect(gate.passed).toBe(false);
+    // the old false-success placeholder must be gone
+    expect(gate.output).not.toBe('CI pipeline gate placeholder');
+    expect(result.allPassed).toBe(false);
+    expect(result.state).not.toBe('passed');
+  });
+
+  it('a ci-pipeline gate fails closed even alongside an otherwise green run', () => {
+    const result = run(['exit 0', { type: 'ci-pipeline', command: 'fake-ci' }]);
+    expect(result.gateResults[1]!.status).toBe('capability_failure');
+    expect(result.gateResults[0]!.status).toBe('passed');
+    expect(result.allPassed).toBe(false);
+    expect(result.state).toBe('capability_failure');
+  });
+
+  it('a manual gate with no automation enters typed waiting — neither pass nor fail', () => {
+    const result = run([{ type: 'manual' }]);
+    const gate = result.gateResults[0]!;
+    expect(gate.status).toBe('waiting');
+    expect(gate.passed).toBe(false);
+    expect(gate.exit_code).toBe(0);
+    // aggregate is not passed while any gate is waiting
+    expect(result.allPassed).toBe(false);
+    expect(result.state).toBe('waiting');
+    expect(result.state).not.toBe('passed');
+  });
+
+  it('a simulated result can never make the aggregate passed', () => {
+    const result = run(['exit 0', 'exit 0'], { simulate: true });
+    expect(result.gateResults.every((g) => g.status === 'simulated')).toBe(true);
+    expect(result.allPassed).toBe(false);
+    expect(result.state).toBe('simulated');
+    expect(result.state).not.toBe('passed');
+  });
+
+  it('waiting dominates an otherwise green aggregate', () => {
+    const result = run(['exit 0', { type: 'manual' }]);
+    expect(result.allPassed).toBe(false);
+    expect(result.state).toBe('waiting');
+  });
+});
+
+describe('runGate fail-closed (RI-N2)', () => {
+  let tmpDir: string;
+  let logPath: string;
+
+  beforeEach(() => {
+    tmpDir = makeTmpDir();
+    logPath = path.join(tmpDir, 'gate.log');
+  });
+
+  afterEach(() => {
+    fs.rmSync(tmpDir, { recursive: true, force: true });
+  });
+
+  it('simulate: true returns a typed simulated result without executing', () => {
+    const result = runGate('this-command-does-not-exist-xyz', tmpDir, logPath, 10, {
+      simulate: true,
+    });
+    expect(result.status).toBe('simulated');
+    expect(result.passed).toBe(false);
+    expect(result.exit_code).toBe(0);
+  });
+
+  it('normal mode executes for real and types a green gate passed', () => {
+    const result = runGate('echo ok', tmpDir, logPath, 10);
+    expect(result.status).toBe('passed');
+    expect(result.passed).toBe(true);
+    expect(result.output).toContain('ok');
+  });
+
+  it('a bare string gate normalizes to mechanical and executes', () => {
+    const result = runGate('exit 7', tmpDir, logPath, 10);
+    expect(result.type).toBe('mechanical');
+    expect(result.status).toBe('failed');
+    expect(result.passed).toBe(false);
+  });
+});
@@ -4,7 +4,20 @@ import { dirname } from 'node:path';

 import { emitEvent } from './event-emitter.js';
 import { nowISO } from './event-emitter.js';
-import type { GateResult } from './types.js';
+import type { GateResult, GateStatus, RunGatesResult } from './types.js';
+
+/** Typed reason stamped on every simulated gate result. */
+export const SIMULATED_GATE_REASON =
+  'simulated execution (explicit simulate opt-in): gate was not evaluated by a real implementation';
+
+/** Options for gate execution (RI-N2 fail-closed / explicit simulation). */
+export interface RunGateOptions {
+  /**
+   * Explicit caller opt-in to simulation. Simulated gates are NOT executed;
+   * every result is typed `simulated` and never satisfies anything.
+   */
+  simulate?: boolean;
+}

 export interface NormalizedGate {
  command: string;
@@ -103,36 +116,91 @@ export function countAIFindings(parsedOutput: unknown): { blockers: number; tota
  return { blockers, total };
 }

+function simulatedResult(gateEntry: NormalizedGate): GateResult {
+  return {
+    command: gateEntry.command,
+    exit_code: 0,
+    type: gateEntry.type,
+    output: SIMULATED_GATE_REASON,
+    timed_out: false,
+    passed: false,
+    status: 'simulated',
+    reason: SIMULATED_GATE_REASON,
+  };
+}
+
+function capabilityFailureResult(
+  gateEntry: NormalizedGate,
+  code: GateResult['capability_code'],
+  reason: string,
+): GateResult {
+  return {
+    command: gateEntry.command,
+    exit_code: 1,
+    type: gateEntry.type,
+    output: '',
+    timed_out: false,
+    passed: false,
+    status: 'capability_failure',
+    capability_code: code,
+    reason,
+  };
+}
+
+function waitingResult(gateEntry: NormalizedGate, reason: string): GateResult {
+  return {
+    command: gateEntry.command,
+    exit_code: 0,
+    type: gateEntry.type,
+    output: '',
+    timed_out: false,
+    passed: false,
+    status: 'waiting',
+    capability_code: 'MACP_AUTHORITY_REQUIRED',
+    reason,
+  };
+}
+
 export function runGate(
  gate: unknown,
  cwd: string,
  logPath: string,
  timeoutSec: number,
+  options: RunGateOptions = {},
 ): GateResult {
  const gateEntry = normalizeGate(gate);
  const gateType = gateEntry.type;
  const command = gateEntry.command;

+  // Explicit simulation only: never executes, typed simulated, never satisfying.
+  if (options.simulate) {
+    return simulatedResult(gateEntry);
+  }
+
+  // Fail closed: no CI provider implementation exists in @mosaicstack/macp,
+  // so a ci-pipeline gate is an absent capability — never a placeholder pass.
  if (gateType === 'ci-pipeline') {
-    return {
-      command,
-      exit_code: 0,
-      type: gateType,
-      output: 'CI pipeline gate placeholder',
-      timed_out: false,
-      passed: true,
-    };
+    return capabilityFailureResult(
+      gateEntry,
+      'MACP_NO_CI_PIPELINE',
+      `ci-pipeline gate '${gateEntry.command || gateType}' has no CI provider implementation wired — refusing placeholder pass`,
+    );
  }

  if (!command) {
-    return {
-      command: '',
-      exit_code: 0,
-      type: gateType,
-      output: '',
-      timed_out: false,
-      passed: true,
-    };
+    // A manual gate with no automation waits for human sign-off: not pass, not fail.
+    if (gateType === 'manual') {
+      return waitingResult(
+        gateEntry,
+        `manual gate has no automation — waiting for human sign-off (type: ${gateType})`,
+      );
+    }
+    // Any other commandless gate is an absent capability — never a vacuous pass.
+    return capabilityFailureResult(
+      gateEntry,
+      gateType === 'ai-review' ? 'MACP_NO_REVIEWER' : 'MACP_NO_COMMAND',
+      `gate of type '${gateType}' has no command to execute — refusing empty-command pass`,
+    );
  }

  const { exitCode, output, timedOut } = runShell(command, cwd, logPath, timeoutSec);
@@ -143,10 +211,12 @@ export function runGate(
    output,
    timed_out: timedOut,
    passed: false,
+    status: 'failed',
  };

  if (gateType !== 'ai-review') {
    result.passed = exitCode === 0;
+    result.status = result.passed ? 'passed' : 'failed';
    return result;
  }

@@ -170,6 +240,7 @@ export function runGate(
  } else {
    result.passed = exitCode === 0 && blockers === 0 && !timedOut && parseError === undefined;
  }
+  result.status = result.passed ? 'passed' : 'failed';

  result.fail_on = failOn;
  result.blockers = blockers;
@@ -191,16 +262,19 @@ export function runGates(
  timeoutSec: number,
  eventsPath: string,
  taskId: string,
-): { allPassed: boolean; gateResults: GateResult[] } {
-  let allPassed = true;
+  options: RunGateOptions = {},
+): RunGatesResult {
  const gateResults: GateResult[] = [];
+  let hasCapabilityFailure = false;
+  let hasSimulated = false;
+  let hasFailed = false;
+  let hasWaiting = false;

  for (const gate of gates) {
    const gateEntry = normalizeGate(gate);
    const gateCmd = gateEntry.command;
-    if (!gateCmd && gateEntry.type !== 'ci-pipeline') continue;
-
    const label = gateCmd || gateEntry.type;
+    // NOTE: no silent skip — every gate produces a typed result (RI-N2).
    emitEvent(
      eventsPath,
      'rail.check.started',
@@ -209,10 +283,10 @@ export function runGates(
      'quality-gate',
      `Running gate: ${label}`,
    );
-    const result = runGate(gate, cwd, logPath, timeoutSec);
+    const result = runGate(gate, cwd, logPath, timeoutSec, options);
    gateResults.push(result);

-    if (result.passed) {
+    if (result.status === 'passed') {
      emitEvent(
        eventsPath,
        'rail.check.passed',
@@ -224,7 +298,46 @@ export function runGates(
      continue;
    }

-    allPassed = false;
+    if (result.status === 'waiting') {
+      hasWaiting = true;
+      emitEvent(
+        eventsPath,
+        'rail.check.waiting',
+        taskId,
+        'gated',
+        'quality-gate',
+        `Gate waiting: ${label} — ${result.reason ?? 'manual gate awaits sign-off'}`,
+      );
+      continue;
+    }
+
+    if (result.status === 'simulated') {
+      hasSimulated = true;
+      emitEvent(
+        eventsPath,
+        'rail.check.simulated',
+        taskId,
+        'gated',
+        'quality-gate',
+        `Gate simulated (non-satisfying): ${label}`,
+      );
+      continue;
+    }
+
+    if (result.status === 'capability_failure') {
+      hasCapabilityFailure = true;
+      emitEvent(
+        eventsPath,
+        'rail.check.failed',
+        taskId,
+        'gated',
+        'quality-gate',
+        `Gate capability failure (${result.capability_code ?? 'MACP_NO_PROVIDER'}): ${label} — ${result.reason ?? 'required capability is absent'}`,
+      );
+      continue;
+    }
+
+    hasFailed = true;
    let message: string;
    if (result.timed_out) {
      message = `Gate timed out after ${timeoutSec}s: ${label}`;
@@ -236,5 +349,15 @@ export function runGates(
    emitEvent(eventsPath, 'rail.check.failed', taskId, 'gated', 'quality-gate', message);
  }

-  return { allPassed, gateResults };
+  const state: GateStatus = hasCapabilityFailure
+    ? 'capability_failure'
+    : hasSimulated
+      ? 'simulated'
+      : hasFailed
+        ? 'failed'
+        : hasWaiting
+          ? 'waiting'
+          : 'passed';
+
+  return { allPassed: state === 'passed', gateResults, state };
 }
@@ -6,11 +6,13 @@ export type {
  DependsOnPolicy,
  GateType,
  GateFailOn,
+  GateStatus,
  GateEntry,
  Task,
  EventType,
  MACPEvent,
  GateResult,
+  RunGatesResult,
  TaskResult,
  ProviderMeta,
  ProviderRegistry,
@@ -18,6 +20,11 @@ export type {

 export { CredentialError } from './types.js';

+// Typed fail-closed capability errors (RI-N2, SDLC-D-035)
+export { MACP_ERROR_CODES, MACPCapabilityError } from './errors.js';
+
+export type { MacpErrorCode } from './errors.js';
+
 // Credential resolver
 export {
  DEFAULT_CREDENTIALS_DIR,
@@ -35,9 +42,16 @@ export {
 export type { ResolveCredentialsOptions } from './credential-resolver.js';

 // Gate runner
-export { normalizeGate, runShell, countAIFindings, runGate, runGates } from './gate-runner.js';
+export {
+  normalizeGate,
+  runShell,
+  countAIFindings,
+  runGate,
+  runGates,
+  SIMULATED_GATE_REASON,
+} from './gate-runner.js';

-export type { NormalizedGate } from './gate-runner.js';
+export type { NormalizedGate, RunGateOptions } from './gate-runner.js';

 // Risk-floor (agent reflection loop — diff review classifier)
 export { evaluateRiskFloor, DEFAULT_RISK_THRESHOLD } from './risk-floor.js';
@@ -1,3 +1,5 @@
+import type { MacpErrorCode } from './errors.js';
+
 /** Task status values. */
 export type TaskStatus = 'pending' | 'running' | 'gated' | 'completed' | 'failed' | 'escalated';

@@ -17,7 +19,17 @@ export type DispatchMode = 'yolo' | 'acp' | 'exec';
 export type DependsOnPolicy = 'all' | 'any' | 'all_terminal';

 /** Quality gate type. */
-export type GateType = 'mechanical' | 'ai-review' | 'ci-pipeline';
+export type GateType = 'mechanical' | 'ai-review' | 'ci-pipeline' | 'manual';
+
+/**
+ * Typed execution state of a gate — closed set (RI-N2, SDLC-D-035).
+ *
+ * Only `passed` means "really executed and green". `simulated` is produced
+ * exclusively under an explicit simulate opt-in and never satisfies anything.
+ * `capability_failure` means a required executor/provider/command was absent.
+ * `waiting` means a manual gate awaits human sign-off (neither pass nor fail).
+ */
+export type GateStatus = 'passed' | 'failed' | 'simulated' | 'waiting' | 'capability_failure';

 /** Gate fail_on mode. */
 export type GateFailOn = 'blocker' | 'any';
@@ -67,7 +79,9 @@ export type EventType =
  | 'task.retry.scheduled'
  | 'rail.check.started'
  | 'rail.check.passed'
-  | 'rail.check.failed';
+  | 'rail.check.failed'
+  | 'rail.check.waiting'
+  | 'rail.check.simulated';

 /** Structured event record. */
 export interface MACPEvent {
@@ -88,7 +102,14 @@ export interface GateResult {
  type: string;
  output: string;
  timed_out: boolean;
+  /** Back-compat boolean view — true ONLY when `status === 'passed'`. */
  passed: boolean;
+  /** Typed discriminator — the authoritative gate outcome (RI-N2). */
+  status: GateStatus;
+  /** Typed capability error code, set when `status === 'capability_failure'`. */
+  capability_code?: MacpErrorCode;
+  /** Why a non-executed state (simulated/waiting/capability_failure) was reached. */
+  reason?: string;
  fail_on?: string;
  blockers?: number;
  findings?: number;
@@ -96,6 +117,22 @@ export interface GateResult {
  parse_error?: string;
 }

+/**
+ * Aggregate outcome of `runGates` (RI-N2).
+ *
+ * `state` is the typed aggregate: it is `passed` only when every gate really
+ * executed green. A `simulated` result makes the aggregate `simulated` (never
+ * `passed`); a `waiting` manual gate keeps the aggregate `waiting`; a missing
+ * capability makes it `capability_failure`. `allPassed` is exactly
+ * `state === 'passed'`, so a simulated or waiting result can never satisfy a
+ * dependency, acceptance criterion, gate, merge, or release check.
+ */
+export interface RunGatesResult {
+  allPassed: boolean;
+  gateResults: GateResult[];
+  state: GateStatus;
+}
+
 /** Result from a completed task. */
 export interface TaskResult {
  task_id: string;
@@ -39,3 +39,20 @@ packages/mosaic/framework/tools/tmux/test-send-message-verdict.sh | requires rea
 # recorded judgement. These lines ARE that judgement, signed.)
 packages/mosaic/framework/tools/orchestrator/smoke-test.sh | behavior smoke checks for coord continue/run workflows, run manually by orchestrator seats; unmeasured in CI; #1017 burndown
 packages/mosaic/framework/tools/wake/validate-973/microtest-wake-assert.sh | #973 instrument self-test, run as a precondition of the validate-973 evidence procedure rather than as a standing CI suite; #1017 burndown candidate
+
+# --- tools/fleet: precondition is unsatisfiable in the CI image (#1271) ---
+# Signed by fred (sb-it-1-dt, 2026-08-16) at origin/next 476db12.
+# This suite asserts the launcher's behaviour when `mosaic` and `pi` are MISSING.
+# It shims fakes into $FAKE_BIN, but the constructed PANE_PATH always ends in the
+# real system path, so on a host that installs those binaries the missing-binary
+# cases cannot be measured at all. The suite's own guard (line 103) says so and
+# fails rather than reporting a pass it cannot back. That guard is correct.
+# The error was wiring the suite into CI: #1017 (c56483eb) enumerated it and
+# dropped this exclusion, and the CI image provides `pi` in the system path, so
+# it has failed on every pipeline since. Measured 2026-08-16 across pipelines
+# 2444 (#1256), 2438 (#1240) and 2441 (#1017-quality): exactly one FAIL line in
+# each full log, identical, this assertion; control `zzz-not-present-zzz` -> 0.
+# Burn-down and the full measurement are tracked in #1271; unwired by PR #1270.
+# Because test:framework-shell is one && chain and this sat at position 44 of 48,
+# the four suites after it had not run at all since the merge.
+packages/mosaic/framework/tools/fleet/test-start-agent-session.sh | precondition unsatisfiable in the CI image: asserts missing-binary behaviour, but PANE_PATH always ends in the system path and the image provides `pi` there; guard at line 103 fails by design rather than passing unmeasured. Burn down by controlling the tail of PANE_PATH inside the test. NOT by removing `pi` from the image: the CI image installs @earendil-works/[email protected] deliberately (measured in pipeline 2444's test-step log), and other suites depend on that pin. Burn-down tracked in #1271
@@ -25,7 +25,7 @@
    "lint": "eslint src",
    "typecheck": "tsc --noEmit",
    "test": "vitest run --passWithNoTests && pnpm run test:framework-shell",
-    "test:framework-shell": "bash framework/tools/quality/scripts/check-test-enumeration.sh && bash framework/tools/quality/scripts/test-check-test-enumeration.sh && python3 src/lease-broker/daemon_deadline_unittest.py && python3 src/lease-broker/normative_fragments_unittest.py && python3 src/lease-broker/promotion_binding_unittest.py && python3 src/lease-broker/promotion_trigger_unittest.py && python3 src/lease-broker/receipt_challenge_unittest.py && python3 src/lease-broker/context_recovery_unittest.py && python3 src/lease-broker/recovery_runtime_unittest.py && python3 src/lease-broker/recovery_b1_adversarial_unittest.py && python3 src/lease-broker/receipt_observer_client_unittest.py && python3 src/lease-broker/invariant_r_unittest.py && python3 src/lease-broker/framework_skill_portability_unittest.py && python3 src/mutator-gate/runtime_tools_unittest.py && python3 src/mutator-gate/runtime_launch_guard_unittest.py && python3 src/mutator-gate/version_coupling_unittest.py && python3 framework/tools/lease-broker/check-runtime-launches.py --root ../.. && bash framework/tools/codex/test-pr-diff-context.sh && bash framework/tools/qa/test-deps-preflight.sh && bash framework/tools/git/test-pr-review-gitea-comment.sh && bash framework/tools/git/test-pr-review-repo-host-override.sh && bash framework/tools/git/test-ci-queue-wait-branch-absent.sh && bash framework/tools/git/test-ci-queue-wait-tristate.sh && bash framework/tools/git/test-ci-queue-wait-github-checks.sh && bash framework/tools/git/test-pr-merge-queue-branch.sh && bash framework/tools/git/test-pr-merge-head-pin.sh && bash framework/tools/git/test-pr-merge-message-field.sh && bash framework/tools/git/test-git-credential-mosaic.sh && bash framework/tools/git/test-gitea-token-identity.sh && bash framework/tools/woodpecker/test-terminal-green-contract.sh && bash framework/tools/_scripts/test-install-ordering-guard.sh && bash framework/tools/_scripts/test-mosaic-init-rce.sh && bash framework/tools/tmux/agent-send.test.sh && bash framework/tools/wake/test-wake-store-ack.sh && bash framework/tools/wake/test-wake-store-enqueue-race.sh && bash framework/tools/wake/test-wake-digest-hmac.sh && bash framework/tools/wake/test-wake-digest-quarantine.sh && bash framework/tools/wake/test-wake-detector.sh && bash framework/tools/wake/test-wake-fn-oracle.sh && bash framework/tools/wake/test-wake-reconcile.sh && bash framework/tools/wake/test-wake-beacon.sh && bash framework/tools/wake/test-wake-preimage.sh && bash framework/tools/wake/test-wake-install.sh && bash framework/tools/fleet/test-start-agent-session.sh && bash framework/tools/glpi/test-list-http-status.sh && bash framework/tools/orchestrator/test-board-roll.sh && bash framework/tools/woodpecker/test-ci-wait-exit-matrix.sh && bash framework/tools/_scripts/test-fleet-transport-check.sh"
+    "test:framework-shell": "bash framework/tools/quality/scripts/check-test-enumeration.sh && bash framework/tools/quality/scripts/test-check-test-enumeration.sh && python3 src/lease-broker/daemon_deadline_unittest.py && python3 src/lease-broker/normative_fragments_unittest.py && python3 src/lease-broker/promotion_binding_unittest.py && python3 src/lease-broker/promotion_trigger_unittest.py && python3 src/lease-broker/receipt_challenge_unittest.py && python3 src/lease-broker/context_recovery_unittest.py && python3 src/lease-broker/recovery_runtime_unittest.py && python3 src/lease-broker/recovery_b1_adversarial_unittest.py && python3 src/lease-broker/receipt_observer_client_unittest.py && python3 src/lease-broker/invariant_r_unittest.py && python3 src/lease-broker/framework_skill_portability_unittest.py && python3 src/mutator-gate/runtime_tools_unittest.py && python3 src/mutator-gate/runtime_launch_guard_unittest.py && python3 src/mutator-gate/version_coupling_unittest.py && python3 framework/tools/lease-broker/check-runtime-launches.py --root ../.. && bash framework/tools/codex/test-pr-diff-context.sh && bash framework/tools/qa/test-deps-preflight.sh && bash framework/tools/git/test-pr-review-gitea-comment.sh && bash framework/tools/git/test-pr-review-repo-host-override.sh && bash framework/tools/git/test-ci-queue-wait-branch-absent.sh && bash framework/tools/git/test-ci-queue-wait-tristate.sh && bash framework/tools/git/test-ci-queue-wait-github-checks.sh && bash framework/tools/git/test-pr-merge-queue-branch.sh && bash framework/tools/git/test-pr-merge-head-pin.sh && bash framework/tools/git/test-pr-merge-message-field.sh && bash framework/tools/git/test-git-credential-mosaic.sh && bash framework/tools/git/test-gitea-token-identity.sh && bash framework/tools/woodpecker/test-terminal-green-contract.sh && bash framework/tools/_scripts/test-install-ordering-guard.sh && bash framework/tools/_scripts/test-mosaic-init-rce.sh && bash framework/tools/tmux/agent-send.test.sh && bash framework/tools/wake/test-wake-store-ack.sh && bash framework/tools/wake/test-wake-store-enqueue-race.sh && bash framework/tools/wake/test-wake-digest-hmac.sh && bash framework/tools/wake/test-wake-digest-quarantine.sh && bash framework/tools/wake/test-wake-detector.sh && bash framework/tools/wake/test-wake-fn-oracle.sh && bash framework/tools/wake/test-wake-reconcile.sh && bash framework/tools/wake/test-wake-beacon.sh && bash framework/tools/wake/test-wake-preimage.sh && bash framework/tools/wake/test-wake-install.sh && bash framework/tools/glpi/test-list-http-status.sh && bash framework/tools/orchestrator/test-board-roll.sh && bash framework/tools/woodpecker/test-ci-wait-exit-matrix.sh && bash framework/tools/_scripts/test-fleet-transport-check.sh"
  },
  "dependencies": {
    "@mosaicstack/brain": "workspace:*",
Author	SHA1	Message	Date
fargo	2097379e25	fix(macp): fail-closed typed gate states — no placeholder/empty-command passes (#1275 ) ci/woodpecker/pr/ci Pipeline was successful Details RI-N2 / SDLC-D-035. Every GateResult now carries a typed status discriminator (passed\|failed\|simulated\|waiting\|capability_failure); passed:true remains true only for really-executed, really-green gates. - ci-pipeline without a CI provider → capability_failure (MACP_NO_CI_PIPELINE), never the placeholder pass - empty-command gate → capability_failure (MACP_NO_COMMAND / MACP_NO_REVIEWER), and runGates no longer silently skips it - manual gate type waits (MACP_AUTHORITY_REQUIRED) — neither pass nor fail - explicit simulation only (simulate option / --simulate): typed simulated results can never make the aggregate passed (RunGatesResult.state) - CLI stubs (tasks list, submit, events tail) exit nonzero with typed MACP_NOT_IMPLEMENTED; macp gate is now wired to runGates with --simulate - legacy __tests__/gate-runner.test.ts consolidated into src/gate-runner.spec.ts (root eslint project service does not cover packages/macp/__tests__)	2026-08-17 16:40:16 -05:00
jason.woltje	8199261caa	Merge pull request 'fix(ci): unwire test-start-agent-session.sh, restore its signed exclusion — unblocks every PR on next' (#1270 ) from fix/1269-ci-chain-unblock into next ci/woodpecker/push/publish Pipeline failed Details Reviewed-on: #1270	2026-08-17 20:44:59 +00:00
fred	57a2f2b40e	docs(ci): point the exclusion at tracking issue #1271 , not the closed first filing ci/woodpecker/pr/ci Pipeline was successful Details The first PR for this change was filed under the retired mos-dt-0 principal (pr-create.sh has no --login flag and find_tea_login_for_host returns the first host match) and was closed and refiled as #1270. That left in-tree references pointing at a closed duplicate PR rather than at the burn-down issue, which is the wrong target for them anyway: the open design question belongs on #1271.	2026-08-16 18:03:02 -05:00
fred	93c1de51e1	fix(ci): unwire test-start-agent-session.sh, restore its signed exclusion (#1269 ) ci/woodpecker/pr/ci Pipeline was canceled Details The `test` step has failed on every `next` pipeline since #1017 on exactly one assertion, and it is the same one on unrelated PRs: FAIL: host provides 'pi' in the system path; missing-binary cases are not measurable here (framework/tools/fleet/test-start-agent-session.sh:103) Measured 2026-08-16 across pipelines 2444 (#1256), 2438 (#1240) and 2441 (#1017-quality): exactly one FAIL line in each full log, identical, this line. Control `zzz-not-present-zzz` -> 0 on all three. Cause. #1241 (`5c35a250`) added the guard: the suite shims fake mosaic/pi/npm into $FAKE_BIN, but the constructed PANE_PATH always ends in the real system path, so on a host that installs those binaries the missing-binary cases cannot be measured and a green run would mean nothing. The guard says so instead of passing. Its own pipeline 2430 was green only because the suite was CI-excluded at the time, so the guard had never run in CI. #1017 (`c56483eb`) then enumerated it and dropped the exclusion. The CI image installs @earendil-works/[email protected].1 on purpose, so the precondition is unsatisfiable there. Both commits are mine. The guard is correct and is not being softened. A check that cannot measure its property and reports success is the failure mode this repo has been cataloguing all week; the error was wiring the suite into an image that violates its precondition, so the wiring is what gets reverted. Second effect, which is the reason this cost a day rather than an hour: test:framework-shell is one && chain and this sat at position 44 of 48, so glpi/test-list-http-status.sh, orchestrator/test-board-roll.sh, woodpecker/test-ci-wait-exit-matrix.sh and _scripts/test-fleet-transport-check.sh have not run at all since the merge. The pipeline reported one failure, never "one failure plus four unrun". All four are green when run directly on sb-it-1-dt, so the mask hid nothing broken -- but that is a local result on one host, not a CI-image result. Verification, with controls: - enumeration guard OK (population 52, enumerated 36, signed-excluded 16). - control A, exclusion line removed while unwired -> FAIL UNENUMERATED. - control B, exclusion line kept while rewired -> FAIL CONTRADICTORY EXCLUSION. The gate discriminates in both directions, so its OK is load-bearing. - the four formerly-masked suites: rc=0 each, run directly. - the full chain cannot be run to completion on sb-it-1-dt: it stops earlier, at the lease-broker Invariant R test, because this host carries the quarantined operator-global pi 0.84.2 against a measured 0.84.1. That is host-specific and out of scope here -- CI pins 0.84.1, and the single FAIL line in those three pipelines proves positions 1-43 passed there. Burn-down is to control the tail of PANE_PATH inside the test, not to remove pi from the image. Recorded in the exclusion reason and in #1269.	2026-08-16 17:58:49 -05:00