fix(ri-050): forge fails closed without providers; explicit typed simulation (#1275 )

2026-08-17 00:58:12 -05:00
28 changed files with 1523 additions and 1428 deletions
@@ -34,6 +34,7 @@ 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',
          ],
        },
@@ -539,3 +539,43 @@ 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`.
@@ -0,0 +1,319 @@
+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, RunManifest, TaskExecutor } from '../src/types.js';
-import type { TaskResult } from '@mosaicstack/macp';
+import type { ForgeTask, ForgeTaskResult, RunManifest, TaskExecutor } from '../src/types.js';
+import { gateLabel, isCommandGate } from '../src/outcomes.js';

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

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

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

    const manifest = loadManifest(result.runDir);
    expect(manifest.stages['00-intake']?.status).toBe('passed');
-    expect(manifest.stages['00b-discovery']?.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',
+    ]);
  });

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

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

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

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

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

-  it('resumes from first incomplete stage', async () => {
-    // First run fails on discovery
-    const executor1 = createMockExecutor({ failStage: '00b-discovery' });
-    let runDir: string;
+  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' },
+      },
+    });

-    try {
-      await runPipeline(briefPath, tmpDir, {
-        executor: executor1,
-        stages: ['00-intake', '00b-discovery', '02-planning-1'],
-      });
-    } catch {
-      // expected
+    // 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');
    }
+  });

-    const runsDir = path.join(tmpDir, '.forge', 'runs');
-    runDir = path.join(runsDir, fs.readdirSync(runsDir)[0]!);
+  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' },
+      },
+    });

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

-    expect(result.manifest.status).toBe('completed');
-    // Should have re-run from 00b-discovery onward
-    expect(result.stages[0]).toBe('00b-discovery');
+    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',
+    });
  });
 });

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

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

 describe('registerForgeCommand', () => {
  it('registers a "forge" command on the parent program', () => {
@@ -55,3 +59,94 @@ 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');
+  });
+});
@@ -5,37 +5,47 @@ 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 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;
-  },
-};
+import { createSimulatedExecutor } from './simulated-executor.js';
+import type { PipelineOptions, RunManifest, RunMode } from './types.js';

 // ---------------------------------------------------------------------------
 // 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();
@@ -44,19 +54,24 @@ 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: ${manifest.updatedAt}`);
  console.log('');
-  console.log('Stage'.padEnd(22) + 'Status'.padEnd(14) + 'Duration');
-  console.log('-'.repeat(50));
+  console.log('Stage'.padEnd(22) + 'Status'.padEnd(24) + 'Duration');
+  console.log('-'.repeat(60));
  for (const stage of STAGE_SEQUENCE) {
    const s = manifest.stages[stage];
    if (!s) continue;
    const label = (STAGE_LABELS[stage] ?? stage).padEnd(22);
-    const status = s.status.padEnd(14);
+    const status = s.status.padEnd(24);
    const dur = formatDuration(s.startedAt, s.completedAt);
    console.log(`${label}${status}${dur}`);
  }
@@ -90,23 +105,58 @@ function listRecentRuns(projectRoot?: string): void {
  }

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

  for (const runId of entries) {
    const runDir = path.join(runsDir, runId);
    try {
      const manifest = loadManifest(runDir);
-      const status = manifest.status.padEnd(14);
+      const status = manifest.status.padEnd(24);
+      const mode = runModeOf(manifest).padEnd(12);
      const brief = path.basename(manifest.brief);
-      console.log(`${runId.padEnd(22)}${status}${brief}`);
+      console.log(`${runId.padEnd(22)}${status}${mode}${brief}`);
    } catch {
-      console.log(`${runId.padEnd(22)}${'(unreadable)'.padEnd(14)}`);
+      console.log(`${runId.padEnd(22)}${'(unreadable)'.padEnd(24)}`);
    }
  }
  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
 // ---------------------------------------------------------------------------
@@ -129,6 +179,11 @@ 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;
@@ -137,6 +192,7 @@ export function registerForgeCommand(parent: Command): void {
        config?: string;
        codebase: string;
        dryRun: boolean;
+        simulate: boolean;
      }) => {
        const briefPath = path.resolve(opts.brief);

@@ -149,14 +205,22 @@ 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}`);
-          const { resumePipeline } = await import('./pipeline-runner.js');
-          const result = await resumePipeline(runDir, stubExecutor);
-          console.log(`[forge] pipeline complete: ${result.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);
+          }
          return;
        }

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

        if (opts.dryRun) {
@@ -180,16 +245,15 @@ 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);
-          console.log(`[forge] pipeline complete: ${result.runId}`);
-          console.log(`[forge] run directory: ${result.runDir}`);
+          applyRunExitPolicy(result, opts.simulate);
        } catch (err) {
-          console.error(
-            `[forge] pipeline failed: ${err instanceof Error ? err.message : String(err)}`,
-          );
-          process.exitCode = 1;
+          handlePipelineError(err);
        }
      },
    );
@@ -224,7 +288,12 @@ 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())
-    .action(async (runId: string, opts: { project: string }) => {
+    .option(
+      '--simulate',
+      'Simulate execution without real providers (every result is typed simulated, never verified)',
+      false,
+    )
+    .action(async (runId: string, opts: { project: string; simulate: boolean }) => {
      const runDir = resolveRunDir(runId, opts.project);

      if (!fs.existsSync(runDir)) {
@@ -234,15 +303,20 @@ 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, stubExecutor);
-        console.log(`[forge] pipeline complete: ${result.runId}`);
-        console.log(`[forge] run directory: ${result.runDir}`);
+        const result = await resumePipeline(runDir, executor, { simulate: opts.simulate });
+        applyRunExitPolicy(result, opts.simulate);
      } catch (err) {
-        console.error(`[forge] resume failed: ${err instanceof Error ? err.message : String(err)}`);
-        process.exitCode = 1;
+        handlePipelineError(err);
      }
    });

@@ -9,7 +9,16 @@ 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. */
+/** 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.
+ */
 export const STAGE_SPECS: Record<string, StageSpec> = {
  '00-intake': {
    number: '00',
@@ -27,7 +36,13 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'discovery-complete',
    promptFile: '00b-discovery.md',
-    qualityGates: ['true'],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 'discovery-complete',
+        reason: 'discovery completion is attested by an authority; no mechanical check exists',
+      },
+    ],
  },
  '01-board': {
    number: '01',
@@ -36,7 +51,13 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'review',
    gate: 'board-approval',
    promptFile: '01-board.md',
-    qualityGates: [{ type: 'ci-pipeline', command: 'board-approval (via board-tasks)' }],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 'board-approval',
+        reason: 'board approval is a board/human decision; no mechanical gate exists',
+      },
+    ],
  },
  '01b-brief-analyzer': {
    number: '01b',
@@ -45,7 +66,13 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'brief-analysis-complete',
    promptFile: '01-board.md',
-    qualityGates: ['true'],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 'brief-analysis-complete',
+        reason: 'brief analysis completion is attested by an authority; no mechanical check exists',
+      },
+    ],
  },
  '02-planning-1': {
    number: '02',
@@ -54,7 +81,13 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'architecture-approval',
    promptFile: '02-planning-1-architecture.md',
-    qualityGates: ['true'],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 'architecture-approval',
+        reason: 'ADR approval requires authority sign-off; no mechanical check exists',
+      },
+    ],
  },
  '03-planning-2': {
    number: '03',
@@ -63,7 +96,14 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'implementation-approval',
    promptFile: '03-planning-2-implementation.md',
-    qualityGates: ['true'],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 'implementation-approval',
+        reason:
+          'implementation spec approval requires authority sign-off; no mechanical check exists',
+      },
+    ],
  },
  '04-planning-3': {
    number: '04',
@@ -72,7 +112,14 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'research',
    gate: 'decomposition-approval',
    promptFile: '04-planning-3-decomposition.md',
-    qualityGates: ['true'],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 'decomposition-approval',
+        reason:
+          'task decomposition approval requires authority sign-off; no mechanical check exists',
+      },
+    ],
  },
  '05-coding': {
    number: '05',
@@ -92,9 +139,10 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    promptFile: '06-review.md',
    qualityGates: [
      {
-        type: 'ai-review',
-        command:
-          'echo \'{"summary":"review-pass","verdict":"approve","findings":[],"stats":{"blockers":0,"should_fix":0,"suggestions":0}}\'',
+        kind: 'provider',
+        capability: 'reviewer',
+        reason:
+          'review verdicts require a wired reviewer provider; synthetic approvals are not permitted',
      },
    ],
  },
@@ -105,7 +153,13 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'coding',
    gate: 're-review',
    promptFile: '07-remediate.md',
-    qualityGates: ['true'],
+    qualityGates: [
+      {
+        kind: 'authority',
+        capability: 're-review',
+        reason: 'remediation re-review is an approval-based gate; no mechanical check exists',
+      },
+    ],
  },
  '08-test': {
    number: '08',
@@ -123,7 +177,13 @@ export const STAGE_SPECS: Record<string, StageSpec> = {
    type: 'deploy',
    gate: 'deploy-verification',
    promptFile: '09-deploy.md',
-    qualityGates: [{ type: 'ci-pipeline', command: 'deploy-verification' }],
+    qualityGates: [
+      {
+        kind: 'provider',
+        capability: 'ci-pipeline',
+        reason: 'deploy verification requires a wired CI pipeline provider',
+      },
+    ],
  },
 };

@@ -0,0 +1,46 @@
+/**
+ * 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';
+  }
+}
@@ -5,6 +5,13 @@ export type {
  StageSpec,
  BriefClass,
  ClassSource,
+  ForgeOutcome,
+  AuthorityGate,
+  ProviderGate,
+  ForgeGate,
+  ForgeGateResult,
+  ForgeTaskResult,
+  RunMode,
  StageStatus,
  RunManifest,
  ForgeTaskStatus,
@@ -81,5 +88,24 @@ 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';
@@ -0,0 +1,147 @@
+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,
+  };
+}
@@ -1,18 +1,33 @@
 import fs from 'node:fs';
 import path from 'node:path';

-import { STAGE_SEQUENCE } from './constants.js';
+import { STAGE_SEQUENCE, STAGE_SPECS } 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.
 */
@@ -47,6 +62,7 @@ function createManifest(opts: {
  briefClass: RunManifest['briefClass'];
  classSource: RunManifest['classSource'];
  forceBoard: boolean;
+  mode: RunMode;
  runDir: string;
 }): RunManifest {
  const ts = nowISO();
@@ -57,6 +73,7 @@ function createManifest(opts: {
    briefClass: opts.briefClass,
    classSource: opts.classSource,
    forceBoard: opts.forceBoard,
+    mode: opts.mode,
    createdAt: ts,
    updatedAt: ts,
    currentStage: '',
@@ -108,20 +125,199 @@ 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. 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
+ * 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
 */
 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');
@@ -146,6 +342,7 @@ export async function runPipeline(
    briefClass,
    classSource,
    forceBoard: options.forceBoard ?? false,
+    mode,
    runDir,
  });

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

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

-    // 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';
+  // All stages reached a terminal state for this mode
+  manifest.status = simulate ? 'simulated' : 'completed';
  saveManifest(runDir, manifest);

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

 /**
- * Resume a pipeline from the last incomplete stage.
+ * Resume a pipeline from the last non-passed stage.
 */
 export async function resumePipeline(
  runDir: string,
-  executor: TaskExecutor,
+  executor?: TaskExecutor,
+  options?: { simulate?: boolean },
 ): 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);

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

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

-    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';
+  manifest.status = simulate ? 'simulated' : 'completed';
  saveManifest(runDir, manifest);

  return {
@@ -0,0 +1,32 @@
+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, TaskResult } from '@mosaicstack/macp';
+import type { GateEntry } from '@mosaicstack/macp';

 /** Stage dispatch mode. */
 export type StageDispatch = 'exec' | 'yolo' | 'pi';
@@ -6,6 +6,58 @@ 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;
@@ -14,7 +66,7 @@ export interface StageSpec {
  type: StageType;
  gate: string;
  promptFile: string;
-  qualityGates: (string | GateEntry)[];
+  qualityGates: ForgeGate[];
 }

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

 /** Per-stage status within a run manifest. */
 export interface StageStatus {
-  status: 'pending' | 'in_progress' | 'passed' | 'failed';
+  status: 'pending' | 'in_progress' | ForgeOutcome;
+  /** Why the stage reached its current (terminal) outcome, when applicable. */
+  reason?: string;
  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;
@@ -38,10 +97,23 @@ 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';
+  status:
+    | 'in_progress'
+    | 'completed'
+    | 'failed'
+    | 'interrupted'
+    | 'rejected'
+    | 'simulated'
+    | 'waiting-for-authority';
  stages: Record<string, StageStatus>;
 }

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

@@ -122,7 +194,16 @@ export interface PipelineOptions {
  stages?: string[];
  skipTo?: string;
  dryRun?: boolean;
-  executor: TaskExecutor;
+  /**
+   * 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;
 }

 /** Pipeline run result. */
@@ -30,7 +30,7 @@ The Gitea API token is **never passed on a curl command line.** An `Authorizatio

 ### `--login` override

-Both `pr-review.sh` and `issue-comment.sh` accept an optional `--login <name>` flag that overrides the automatically detected Gitea login for that single invocation (as of #1280, `pr-create.sh`, `pr-merge.sh` and `issue-create.sh` accept it too, and it wins over `MOSAIC_GIT_IDENTITY` everywhere). The override selects **which credential the REST write, the `/user` identity lookup, and the read-back all use** — its token is resolved from the tea config for that login name (`get_gitea_token_for_login`), falling back to the repo host's credential when no login is named. The resolved login is **host- and port-bound**: the login's configured URL host **and effective port** (the scheme's default port — 80 for `http`, 443 for `https` — applies when a port is omitted, symmetrically on both sides) must match the repo remote's, so a login name shared across hosts (or an override configured for a different Gitea, including one on a different port of the same host) can never send one host's credential to another — a host or port mismatch fails closed rather than leaking a cross-host token. Resolving the acting identity and the read-back from the _same_ login that performs the write is essential: a write performed under an overridden login must be verified against that login's identity, not the host default's. Callers who need a different login than the host default should pass `--login <reviewer-login>`.
+Both `pr-review.sh` and `issue-comment.sh` accept an optional `--login <name>` flag that overrides the automatically detected Gitea login for that single invocation. The override selects **which credential the REST write, the `/user` identity lookup, and the read-back all use** — its token is resolved from the tea config for that login name (`get_gitea_token_for_login`), falling back to the repo host's credential when no login is named. The resolved login is **host- and port-bound**: the login's configured URL host **and effective port** (the scheme's default port — 80 for `http`, 443 for `https` — applies when a port is omitted, symmetrically on both sides) must match the repo remote's, so a login name shared across hosts (or an override configured for a different Gitea, including one on a different port of the same host) can never send one host's credential to another — a host or port mismatch fails closed rather than leaking a cross-host token. Resolving the acting identity and the read-back from the _same_ login that performs the write is essential: a write performed under an overridden login must be verified against that login's identity, not the host default's. Callers who need a different login than the host default should pass `--login <reviewer-login>`.

 As a durable successor to this mechanism, consider giving each reviewer/approver slot its own dedicated Gitea login credential, so that author≠reviewer holds at the credential level rather than relying on wrapper-level `--login` bookkeeping. This is a recommendation for future hardening, not something implemented by this flag.

@@ -58,36 +58,6 @@ token file present, both tools fall through to the existing shared-account path
 unchanged, so this feature is a no-op on any host that hasn't provisioned per-slot
 tokens.

-### Identity-first principal resolution in the wrappers (#1280)
-
-`resolve_gitea_principal()` (detect-platform.sh) gives the write wrappers —
-`pr-create.sh`, `pr-merge.sh`, `pr-review.sh`, `issue-create.sh`, `issue-comment.sh` —
-ONE precedence for choosing the acting principal:
-
-1. an explicit `--login <name>` (now accepted by all five; operator intent beats
-   environment), then
-2. the per-agent identity above (`MOSAIC_GIT_IDENTITY` env / worktree
-   `mosaic.gitIdentity`) when a per-slot token exists — the wrapper then writes via the
-   REST API with that identity's token and never consults `tea`, so the tea login list
-   cannot shadow the requested principal, then
-3. the tea login list — the LAST resort, never the first, because it enumerates
-   whatever logins the host happens to hold and knows nothing about which seat is
-   calling.
-
-A requested identity whose per-slot token is absent, or a `--login` whose token cannot
-resolve host-bound, **fails loud** (nonzero, naming the identity/login and the expected
-slot) instead of silently writing under whatever account `tea` has configured — that
-silent fallthrough is defect #1280 (reviews, comments, merges, PRs and issues filed
-under the wrong account). `pr-merge.sh --dry-run` reports the principal the merge would
-act as, resolved exactly as the real merge resolves it. ⚠ A **workstation-global**
-`mosaic.gitIdentity` shadows every seat on that host (a fresh clone with no local value
-resolves the global one) — set it per-worktree, not with `--global`.
-
-The resolver is covered by `test-gitea-principal-resolution.sh`; the happy-path
-ordering (identity arm REACHED, not sitting behind a tea failure) by
-`test-pr-create-identity-first.sh`; merge credential binding by
-`test-pr-merge-principal-resolution.sh`.
-
 ### Enabling it for a clone

 The framework installer syncs `git-credential-mosaic` to
@@ -497,32 +497,6 @@ get_gitea_url_for_host() {
    return 1
 }

-# Map a Gitea host to the per-agent identity-token slot PREFIX ("gitea-usc" /
-# "gitea-mosaicstack") used by identity-first principal resolution
-# (MOSAIC_GIT_IDENTITY / git config mosaic.gitIdentity; #1280). Returns 1 for
-# hosts with no per-slot scheme — callers treat that as "identity does not
-# bind here" and fall through to existing behavior, never as an error. This is
-# the single source of truth for the slot layout: get_gitea_token and
-# resolve_gitea_principal both derive their slot paths from here, so the two
-# resolutions can never disagree about where an identity's credential lives.
-gitea_identity_slot_prefix() {
-    case "$1" in
-        git.uscllc.com)      echo "gitea-usc" ;;
-        git.mosaicstack.dev) echo "gitea-mosaicstack" ;;
-        *) return 1 ;;
-    esac
-}
-
-# Resolve the per-slot token FILE PATH for an identity on a host. Prints the
-# absolute path on success; returns 1 (no output) when the host has no per-slot
-# scheme. Prints a PATH only — never a token value.
-gitea_identity_token_slot() {
-    local identity="$1" host="$2" prefix
-    [[ -n "$identity" ]] || return 1
-    prefix=$(gitea_identity_slot_prefix "$host") || return 1
-    printf '%s\n' "$HOME/.config/mosaic/secrets/gitea-tokens/${prefix}-${identity}.token"
-}
-
 # Resolve a Gitea API token for the given host.
 # Priority: Mosaic credential loader → GITEA_TOKEN env → ~/.git-credentials
 get_gitea_token() {
@@ -543,8 +517,13 @@ get_gitea_token() {
        _ident_src="git config mosaic.gitIdentity"
    fi
    if [[ -n "$_ident" ]]; then
-        local _idtok=""
-        if _idtok="$(gitea_identity_token_slot "$_ident" "$host" 2>/dev/null)"; then
+        local _idpfx=""
+        case "$host" in
+            git.uscllc.com)      _idpfx=gitea-usc ;;
+            git.mosaicstack.dev) _idpfx=gitea-mosaicstack ;;
+        esac
+        if [[ -n "$_idpfx" ]]; then
+            local _idtok="$HOME/.config/mosaic/secrets/gitea-tokens/${_idpfx}-${_ident}.token"
            if [[ -r "$_idtok" ]]; then
                cat "$_idtok"
                return 0
@@ -1486,81 +1465,6 @@ raise SystemExit(1)
 PY
 }

-# resolve_gitea_principal — identity-first acting-principal resolution shared by
-# the git wrappers (#1280). The defect this fixes: wrappers resolved their
-# acting principal from tea's login list FIRST, and that list enumerates
-# whatever logins happen to be configured on the host — it knows nothing about
-# which seat is calling — so a wrapper invoked with MOSAIC_GIT_IDENTITY=fargo
-# still wrote under whichever account tea held (mos-dt-0), and the correct
-# identity-aware code sat behind arms that only ran when the tea path failed.
-# Precedence here is the contract:
-#   1. an explicit login override ($1, the wrapper's --login) — operator intent
-#      beats environment;
-#   2. MOSAIC_GIT_IDENTITY env, else per-worktree `git config mosaic.gitIdentity`
-#      (mirroring get_gitea_token exactly, so resolver and token resolution can
-#      never disagree) — binds only on hosts with a per-slot token scheme;
-#   3. the tea login list — LAST resort, never the first.
-#
-# Prints exactly one line, three tab-separated fields (machine-readable for
-# wrapper dispatch and tests):
-#   mode        "login" | "identity" | "default"
-#   principal   login name (login) | identity name (identity) | tea login or "" (default)
-#   source      "tea-login:<name>" | "identity-slot:<path>" | "tea-default" | "host-credential"
-#
-# Fails LOUD (nonzero, empty stdout, stderr diagnostic) when an explicit
-# override cannot be honored — a refusal is a good day; silently falling
-# through to whoever tea has configured is the exact defect this resolves:
-#   - login mode: no host-bound token for that tea login. The existence check
-#     runs the same tea-config lookup tea itself uses; the token VALUE is
-#     discarded (never printed, never used).
-#   - identity mode: no per-slot token file for that identity on a recognized
-#     host — the diagnostic names the identity, its source, and the expected
-#     slot path. An identity requested on a host with NO per-slot scheme does
-#     not bind (matching get_gitea_token's containment) and falls to default.
-#
-# NEVER prints a token value — principal names and slot paths only.
-# $1 = explicit login override ("" when absent), $2 = host (default: the
-# origin remote's host).
-resolve_gitea_principal() {
-    local login_override="${1:-}" host="${2:-}" ident ident_src slot login
-    [[ -n "$host" ]] || { host=$(get_remote_host) || return 1; }
-
-    if [[ -n "$login_override" ]]; then
-        get_gitea_token_for_login "$login_override" "$host" >/dev/null || {
-            echo "Error: --login '$login_override' has no host-matched token on host '$host' (tea config lookup); refusing to fall back to any other principal (#1280 identity-first resolution)." >&2
-            return 1
-        }
-        printf 'login\t%s\ttea-login:%s\n' "$login_override" "$login_override"
-        return 0
-    fi
-
-    ident="${MOSAIC_GIT_IDENTITY:-}"
-    ident_src="MOSAIC_GIT_IDENTITY"
-    if [[ -z "$ident" ]]; then
-        ident="$(git config --get mosaic.gitIdentity 2>/dev/null || true)"
-        ident_src="git config mosaic.gitIdentity"
-    fi
-    if [[ -n "$ident" ]] && slot="$(gitea_identity_token_slot "$ident" "$host" 2>/dev/null)"; then
-        if [[ -r "$slot" ]]; then
-            printf 'identity\t%s\tidentity-slot:%s\n' "$ident" "$slot"
-            return 0
-        fi
-        echo "Error: git identity '$ident' requested (via $ident_src) for host '$host', but no per-slot token at $slot (#1280 identity-first resolution)." >&2
-        echo "       Refusing to fall back to the tea login list or shared credentials. Provision the per-slot token, or unset the identity." >&2
-        return 1
-    fi
-
-    # No override requested: tea's login list is the LAST resort. Absence is
-    # not an error here — callers fall back to the host credential, exactly as
-    # they did before this resolver existed (preserved behavior).
-    if login=$(get_gitea_login_for_host "$host" 2>/dev/null); then
-        printf 'default\t%s\ttea-default\n' "$login"
-    else
-        printf 'default\t\thost-credential\n'
-    fi
-    return 0
-}
-
 # Resolve HTTPS basic auth credentials for a Gitea host from ~/.git-credentials.
 # Prints "username:password" for direct curl -u consumption. Callers must not log it.
 get_gitea_basic_auth() {
@@ -76,36 +76,27 @@ fi
 detect_platform >/dev/null

 # Resolve and cache the Gitea REST endpoint + token for the current remote,
-# bound to a SPECIFIC acting principal ($1) selected identity-first (#1280):
-# an explicit --login wins, else MOSAIC_GIT_IDENTITY / git config
-# mosaic.gitIdentity binds the per-slot credential, else the tea login list
-# (last resort). Populates GITEA_API_ROOT (…/api/v1), GITEA_API_BASE
-# (…/api/v1/repos/<slug>), and GITEA_API_TOKEN.
+# bound to a SPECIFIC login identity ($1). Populates GITEA_API_ROOT (…/api/v1),
+# GITEA_API_BASE (…/api/v1/repos/<slug>), and GITEA_API_TOKEN.
 #
-# The token is resolved for the EFFECTIVE principal so that the single
-# credential used for the write ALSO drives the /user identity read and the
-# read-back — write token and read-back token are the same identity by
-# construction (this is the credential-ordering fix: a --login override is no
-# longer written under one credential and verified under a different default
-# one). When $2 is "identity" the principal ($1) is a requested git identity:
-# the token MUST resolve from that identity's per-slot token (get_gitea_token's
-# identity arm), failing closed rather than borrowing the tea default login —
-# the tea login list must never shadow a requested identity (#1280). When $2
-# is "explicit" the principal came from a caller-supplied --login: that exact
-# login's token MUST resolve, and we FAIL CLOSED rather than silently
-# downgrading the write to the host default identity. Otherwise the best-effort
-# default path applies (per-login token, else the host-scoped credential).
-# Returns non-zero (clear stderr) on any resolution failure.
+# The token is resolved for the EFFECTIVE login (the --login override when
+# given, otherwise the detected default) so that the single credential used for
+# the write ALSO drives the /user identity read and the read-back — write token
+# and read-back token are the same identity by construction (this is the
+# credential-ordering fix: a --login override is no longer written under one
+# credential and verified under a different default one). Falls back to the
+# host-scoped credential ONLY when NO --login override was supplied (the
+# best-effort default path). When $2 is "explicit" the login came from a
+# caller-supplied --login: that exact login's token MUST resolve, and we FAIL
+# CLOSED rather than silently downgrading the write to the host default
+# identity — otherwise a caller relying on a dedicated per-role credential would
+# be told the write succeeded as requested while it was attributed to the shared
+# default. Returns non-zero (clear stderr) on any resolution failure.
 gitea_resolve_api_for_login() {
    local effective_login="$1" override_explicit="${2:-}" host configured_url repo

    host=$(get_remote_host)
-    if [[ "$override_explicit" == "identity" ]]; then
-        GITEA_API_TOKEN=$(get_gitea_token "$host") || {
-            echo "Error: could not resolve the per-slot token for requested git identity '$effective_login' on host '$host'; refusing to fall back to the tea login list or shared credentials (comment write/read-back, #1280)." >&2
-            return 1
-        }
-    elif [[ -n "$override_explicit" ]]; then
+    if [[ -n "$override_explicit" ]]; then
        GITEA_API_TOKEN=$(get_gitea_token_for_login "$effective_login" "$host") || {
            echo "Error: could not resolve a host-matched Gitea token for --login '$effective_login' on host '$host'; refusing to fall back to the host default identity or a cross-host credential (comment write/read-back)" >&2
            return 1
@@ -327,31 +318,23 @@ if [[ "$PLATFORM" == "github" ]]; then
    gh issue comment "$ISSUE_NUMBER" --body "$COMMENT"
    echo "Added comment to GitHub issue #$ISSUE_NUMBER"
 elif [[ "$PLATFORM" == "gitea" ]]; then
-    # Resolve the acting principal identity-first (#1280): an explicit --login
-    # wins; otherwise MOSAIC_GIT_IDENTITY / git config mosaic.gitIdentity
-    # selects the principal when a per-slot token exists (fail-loud when it
-    # does not); the tea login list is the LAST resort — it knows nothing about
-    # which seat is calling, so resolving from it first wrote under whichever
-    # account tea had configured (the #1280 family).
-    principal_host=$(get_remote_host)
-    if ! principal_resolved="$(resolve_gitea_principal "$LOGIN_OVERRIDE" "$principal_host")"; then
-        # resolve_gitea_principal already printed the fail-loud diagnostic.
-        exit 1
-    fi
-    PRINCIPAL_MODE="$(printf '%s' "$principal_resolved" | cut -f1)"
-    PRINCIPAL_NAME="$(printf '%s' "$principal_resolved" | cut -f2)"
+    # Resolve the login this comment should be attributed to: the --login
+    # override when given, otherwise the detected default for this repo's host.
+    # A --login override always wins. Otherwise name this repo host's login only
+    # as a best effort: the login name merely selects a per-login token, and
+    # gitea_resolve_api_for_login falls back to the host credential
+    # (get_gitea_token) when no tea login is named, so the default credential
+    # still resolves even when the host tea has no matching login entry.
+    EFFECTIVE_LOGIN="$LOGIN_OVERRIDE"
+    [[ -n "$EFFECTIVE_LOGIN" ]] || EFFECTIVE_LOGIN=$(get_gitea_login 2>/dev/null || true)

-    # Bind the REST endpoint + token to the resolved principal, then derive the
+    # Bind the REST endpoint + token to the effective login, then derive the
    # acting identity from that SAME credential (GET /user). The write below and
    # its read-back both use this credential, so the write is verified against
-    # the identity that actually performed it.
-    if [[ "$PRINCIPAL_MODE" == "identity" ]]; then
-        gitea_resolve_api_for_login "$PRINCIPAL_NAME" identity || exit 1
-    elif [[ "$PRINCIPAL_MODE" == "login" ]]; then
-        gitea_resolve_api_for_login "$PRINCIPAL_NAME" explicit || exit 1
-    else
-        gitea_resolve_api_for_login "$PRINCIPAL_NAME" "" || exit 1
-    fi
+    # the identity that actually performed it. Passing "explicit" when --login
+    # was supplied forbids the host-default fallback: an unresolvable explicit
+    # override fails closed instead of writing under the default identity.
+    gitea_resolve_api_for_login "$EFFECTIVE_LOGIN" "${LOGIN_OVERRIDE:+explicit}" || exit 1
    ACTING_LOGIN=$(gitea_authenticated_login) || exit 1

    comment_id=$(gitea_create_comment_verified "$ISSUE_NUMBER" "$COMMENT" "$ACTING_LOGIN") || {
@@ -1,15 +1,6 @@
 #!/bin/bash
 # issue-create.sh - Create issues on Gitea or GitHub
-# Usage: issue-create.sh -t "Title" [-b "Body"] [-l "label1,label2"] [-m "milestone"] [--login <name>]
-#
-# Acting principal is resolved identity-first (#1280): an explicit --login
-# wins; otherwise MOSAIC_GIT_IDENTITY / per-worktree git config
-# mosaic.gitIdentity selects the principal when a per-slot token exists (and
-# the wrapper then creates the issue through the REST API with that identity's
-# token — tea is never invoked, so the tea login list cannot shadow the
-# requested principal); the tea login list is the LAST resort. A requested
-# identity with no per-slot token fails LOUD rather than writing under
-# whichever account tea happens to hold.
+# Usage: issue-create.sh -t "Title" [-b "Body"] [-l "label1,label2"] [-m "milestone"]

 set -e

@@ -25,14 +16,6 @@ INTERACTIVE=false

 # get_remote_host and get_gitea_token are provided by detect-platform.sh

-# Acting-principal mode set in the Gitea branch below (from
-# resolve_gitea_principal): "login" when --login was given, "identity" when a
-# git identity bound, "default" otherwise. PRINCIPAL_MODE=login makes the API
-# arm resolve the --login principal's token too, so an explicit --login keeps
-# winning even on the tea-FAILURE fallback arm.
-PRINCIPAL_MODE=""
-PRINCIPAL_NAME=""
-
 gitea_issue_create_api() {
    local host repo token url payload
    host=$(get_remote_host) || {
@@ -43,19 +26,10 @@ gitea_issue_create_api() {
        echo "Error: could not determine repo owner/name for API fallback" >&2
        return 1
    }
-    if [[ "$PRINCIPAL_MODE" == "login" ]]; then
-        token=$(get_gitea_token_for_login "$PRINCIPAL_NAME" "$host") || {
-            echo "Error: could not resolve a host-matched Gitea token for --login '$PRINCIPAL_NAME' on host '$host' (API path)" >&2
-            return 1
-        }
-    else
-        # Identity-first when MOSAIC_GIT_IDENTITY / git config mosaic.gitIdentity
-        # is set (per-slot token, fail-loud on absence); shared default otherwise.
-        token=$(get_gitea_token "$host") || {
-            echo "Error: Gitea token not found for API fallback (set GITEA_TOKEN or configure ~/.git-credentials)" >&2
-            return 1
-        }
-    fi
+    token=$(get_gitea_token "$host") || {
+        echo "Error: Gitea token not found for API fallback (set GITEA_TOKEN or configure ~/.git-credentials)" >&2
+        return 1
+    }

    if [[ -n "$LABELS" || -n "$MILESTONE" ]]; then
        echo "Warning: API fallback currently applies title/body only; labels/milestone require authenticated tea setup." >&2
@@ -93,7 +67,6 @@ Options:
    -b, --body BODY         Issue body/description
    -l, --labels LABELS     Comma-separated labels (e.g., "bug,feature")
    -m, --milestone NAME    Milestone name to assign
-        --login NAME        Act as this Gitea tea login (wins over MOSAIC_GIT_IDENTITY)
    -i, --interactive       Prompt for missing issue fields
    -h, --help              Show this help message

@@ -124,10 +97,6 @@ while [[ $# -gt 0 ]]; do
            MILESTONE="$2"
            shift 2
            ;;
-        --login)
-            LOGIN_OVERRIDE="$2"
-            shift 2
-            ;;
        -i|--interactive)
            INTERACTIVE=true
            shift
@@ -165,37 +134,13 @@ case "$PLATFORM" in
        "${CMD[@]}"
        ;;
    gitea)
-        # Resolve the acting principal identity-first (#1280). The tea login
-        # list is the LAST resort: it knows nothing about which seat is calling,
-        # and a login resolved from it first is what attributed issues to the
-        # wrong account even when MOSAIC_GIT_IDENTITY was set.
-        principal_host=$(get_remote_host 2>/dev/null || true)
-        if ! principal_resolved="$(resolve_gitea_principal "${LOGIN_OVERRIDE:-}" "$principal_host")"; then
-            # resolve_gitea_principal already printed the fail-loud diagnostic.
-            exit 1
-        fi
-        PRINCIPAL_MODE="$(printf '%s' "$principal_resolved" | cut -f1)"
-        PRINCIPAL_NAME="$(printf '%s' "$principal_resolved" | cut -f2)"
-
-        if [[ "$PRINCIPAL_MODE" == "identity" ]]; then
-            # HAPPY PATH for a requested identity: create through the REST API
-            # with the per-slot token and never invoke tea — the identity arm
-            # must be REACHED, not sit behind a tea failure (#1280).
-            gitea_issue_create_api
-            exit $?
-        fi
-
        if command -v tea >/dev/null 2>&1; then
            REPO_SLUG=$(get_repo_slug)
-            if [[ "$PRINCIPAL_MODE" == "login" ]]; then
-                GITEA_LOGIN_NAME="$PRINCIPAL_NAME"
-            else
-                GITEA_LOGIN_NAME=$(get_gitea_login) || {
-                    echo "Warning: could not resolve Gitea login for tea; trying Gitea API fallback..." >&2
-                    gitea_issue_create_api
-                    exit $?
-                }
-            fi
+            GITEA_LOGIN_NAME=$(get_gitea_login) || {
+                echo "Warning: could not resolve Gitea login for tea; trying Gitea API fallback..." >&2
+                gitea_issue_create_api
+                exit $?
+            }
            if ! get_gitea_authenticated_user "$GITEA_LOGIN_NAME" >/dev/null; then
                echo "Warning: Tea authenticated-user validation failed (possible stale user/login); trying Gitea API fallback..." >&2
                gitea_issue_create_api
@@ -1,15 +1,6 @@
 #!/bin/bash
 # pr-create.sh - Create pull requests on Gitea or GitHub
-# Usage: pr-create.sh -t "Title" [-b "Body"] [-B base] [-H head] [-l "labels"] [-m "milestone"] [--login <name>]
-#
-# Acting principal is resolved identity-first (#1280): an explicit --login
-# wins; otherwise MOSAIC_GIT_IDENTITY / per-worktree git config
-# mosaic.gitIdentity selects the principal when a per-slot token exists (and
-# the wrapper then creates the PR through the REST API with that identity's
-# token — tea is never invoked, so the tea login list cannot shadow the
-# requested principal); the tea login list is the LAST resort. A requested
-# identity with no per-slot token fails LOUD rather than writing under
-# whichever account tea happens to hold.
+# Usage: pr-create.sh -t "Title" [-b "Body"] [-B base] [-H head] [-l "labels"] [-m "milestone"]

 set -e

@@ -28,15 +19,6 @@ ISSUE=""

 # get_remote_host, get_gitea_token, get_repo_info, and get_gitea_repo_args are provided by detect-platform.sh

-# Acting-principal mode set in the Gitea branch below (from
-# resolve_gitea_principal): "login" when --login was given, "identity" when a
-# git identity bound, "default" otherwise. PRINCIPAL_MODE=login makes the API
-# arm resolve the --login principal's token too, so an explicit --login keeps
-# winning even on the tea-FAILURE fallback arm (otherwise the fallback would
-# silently re-resolve to the environment identity or shared credential).
-PRINCIPAL_MODE=""
-PRINCIPAL_NAME=""
-
 gitea_pr_create_api() {
    local host repo token url payload
    host=$(get_remote_host) || {
@@ -47,19 +29,10 @@ gitea_pr_create_api() {
        echo "Error: could not determine repo owner/name for API fallback" >&2
        return 1
    }
-    if [[ "$PRINCIPAL_MODE" == "login" ]]; then
-        token=$(get_gitea_token_for_login "$PRINCIPAL_NAME" "$host") || {
-            echo "Error: could not resolve a host-matched Gitea token for --login '$PRINCIPAL_NAME' on host '$host' (API path)" >&2
-            return 1
-        }
-    else
-        # Identity-first when MOSAIC_GIT_IDENTITY / git config mosaic.gitIdentity
-        # is set (per-slot token, fail-loud on absence); shared default otherwise.
-        token=$(get_gitea_token "$host") || {
-            echo "Error: Gitea token not found for API fallback (set GITEA_TOKEN or configure ~/.git-credentials)" >&2
-            return 1
-        }
-    fi
+    token=$(get_gitea_token "$host") || {
+        echo "Error: Gitea token not found for API fallback (set GITEA_TOKEN or configure ~/.git-credentials)" >&2
+        return 1
+    }

    if [[ -n "$LABELS" || -n "$MILESTONE" || "$DRAFT" == true ]]; then
        echo "Warning: API fallback applies title/body/head/base only; labels/milestone/draft require authenticated tea setup." >&2
@@ -103,7 +76,6 @@ Options:
    -H, --head BRANCH       Head branch with changes (default: current branch)
    -l, --labels LABELS     Comma-separated labels
    -m, --milestone NAME    Milestone name
-        --login NAME        Act as this Gitea tea login (wins over MOSAIC_GIT_IDENTITY)
    -i, --issue NUMBER      Link to issue (auto-generates title if not provided)
    -d, --draft             Create as draft PR
    -h, --help              Show this help message
@@ -144,10 +116,6 @@ while [[ $# -gt 0 ]]; do
            MILESTONE="$2"
            shift 2
            ;;
-        --login)
-            LOGIN_OVERRIDE="$2"
-            shift 2
-            ;;
        -i|--issue)
            ISSUE="$2"
            shift 2
@@ -206,41 +174,15 @@ case "$PLATFORM" in
        "${CMD[@]}"
        ;;
    gitea)
-        # Resolve the acting principal identity-first (#1280). The tea login
-        # list is the LAST resort: it knows nothing about which seat is calling,
-        # and a login resolved from it first is what attributed PRs to the wrong
-        # account even when MOSAIC_GIT_IDENTITY was set.
-        principal_host=$(get_remote_host 2>/dev/null || true)
-        if ! principal_resolved="$(resolve_gitea_principal "${LOGIN_OVERRIDE:-}" "$principal_host")"; then
-            # resolve_gitea_principal already printed the fail-loud diagnostic.
-            exit 1
-        fi
-        PRINCIPAL_MODE="$(printf '%s' "$principal_resolved" | cut -f1)"
-        PRINCIPAL_NAME="$(printf '%s' "$principal_resolved" | cut -f2)"
-
-        if [[ "$PRINCIPAL_MODE" == "identity" ]]; then
-            # HAPPY PATH for a requested identity: the per-slot token IS the
-            # credential, so create through the REST API directly and never
-            # invoke tea — the identity arm must be REACHED, not sit behind a
-            # tea failure (#1280). Fail-loud on a missing slot already happened
-            # in resolve_gitea_principal.
-            gitea_pr_create_api
-            exit $?
-        fi
-
        # tea pull create syntax. Always pass --repo because tea repo inference
        # is unreliable in Mosaic worktrees/profile shells. Use arrays instead
        # of eval so markdown backticks/body content are not shell-executed.
        REPO_SLUG=$(get_repo_slug)
-        if [[ "$PRINCIPAL_MODE" == "login" ]]; then
-            GITEA_LOGIN_NAME="$PRINCIPAL_NAME"
-        else
-            GITEA_LOGIN_NAME=$(get_gitea_login) || {
-                echo "Warning: could not resolve Gitea login for tea; trying Gitea API fallback..." >&2
-                gitea_pr_create_api
-                exit $?
-            }
-        fi
+        GITEA_LOGIN_NAME=$(get_gitea_login) || {
+            echo "Warning: could not resolve Gitea login for tea; trying Gitea API fallback..." >&2
+            gitea_pr_create_api
+            exit $?
+        }
        if ! get_gitea_authenticated_user "$GITEA_LOGIN_NAME" >/dev/null; then
            echo "Warning: Tea authenticated-user validation failed (possible stale user/login); trying Gitea API fallback..." >&2
            gitea_pr_create_api
@@ -1,13 +1,6 @@
 #!/bin/bash
 # pr-merge.sh - Merge pull requests on Gitea or GitHub
-# Usage: pr-merge.sh -n PR_NUMBER [-m squash] [-d] [--expect-head SHA] [--co-author-trailers --escalate-to PRINCIPAL] [--login <name>]
-#
-# Acting principal is resolved identity-first (#1280): an explicit --login
-# wins; otherwise MOSAIC_GIT_IDENTITY / per-worktree git config
-# mosaic.gitIdentity selects the credential (per-slot token, fail-loud when
-# absent); the shared host credential is the last resort. The merge is
-# performed with the resolved credential only — never a cross-principal
-# fallback (an HTTP 401 from the identity-bound token is a hard stop).
+# Usage: pr-merge.sh -n PR_NUMBER [-m squash] [-d] [--expect-head SHA] [--co-author-trailers --escalate-to PRINCIPAL]

 set -euo pipefail

@@ -23,7 +16,6 @@ DRY_RUN=false
 EXPECT_HEAD=""
 CO_AUTHOR_TRAILERS=false
 ESCALATE_TO=""
-LOGIN_OVERRIDE=""

 usage() {
    cat <<EOF
@@ -39,7 +31,6 @@ Options:
        --expect-head SHA   Refuse unless the PR head matches this full commit SHA
        --co-author-trailers Build verified trailers from linked PR commit authors
        --escalate-to NAME  Named principal for an unresolved-author BLOCK
-        --login NAME        Act as this Gitea tea login (wins over MOSAIC_GIT_IDENTITY)
    -h, --help              Show this help message

 Examples:
@@ -48,7 +39,6 @@ Examples:
    $(basename "$0") -n 42 -d                 # Squash merge and delete branch
    $(basename "$0") -n 42 --expect-head 0123456789abcdef0123456789abcdef01234567
    $(basename "$0") -n 42 --co-author-trailers --escalate-to tl-mosaic
-    $(basename "$0") -n 42 --login fred-ms    # Merge under the fred-ms tea login
 EOF
    exit "${1:-1}"
 }
@@ -92,14 +82,6 @@ while [[ $# -gt 0 ]]; do
            ESCALATE_TO="$2"
            shift 2
            ;;
-        --login|-l)
-            if [[ $# -lt 2 ]]; then
-                echo "Error: --login requires one tea login name." >&2
-                exit 1
-            fi
-            LOGIN_OVERRIDE="$2"
-            shift 2
-            ;;
        -h|--help)
            usage 0
            ;;
@@ -590,22 +572,9 @@ PY
 merge_gitea_with_api() {
    local host="$1" token attempt_rc

-    # Identity-first principal resolution (#1280): an explicit --login wins
-    # over MOSAIC_GIT_IDENTITY (operator intent beats environment); otherwise
-    # get_gitea_token resolves the identity's per-slot token when an identity
-    # is requested (fail-loud when absent) and the shared host credential only
-    # when no identity is set. No cross-principal fallback: whatever resolves
-    # here is the ONLY credential the merge is attempted with.
-    if [[ -n "$LOGIN_OVERRIDE" ]]; then
-        if ! token=$(get_gitea_token_for_login "$LOGIN_OVERRIDE" "$host"); then
-            echo "Error: --login '$LOGIN_OVERRIDE' has no host-matched token on host '$host'; refusing to merge under any other principal (#1280 identity-first resolution)." >&2
-            return 1
-        fi
-    else
-        if ! token=$(get_gitea_token "$host"); then
-            echo "Error: Could not resolve the required Gitea token; refusing merge without changing principals." >&2
-            return 1
-        fi
+    if ! token=$(get_gitea_token "$host"); then
+        echo "Error: Could not resolve the required Gitea token; refusing merge without changing principals." >&2
+        return 1
    fi
    if [[ -z "$token" ]]; then
        echo "Error: Required Gitea token resolved empty; refusing merge without changing principals." >&2
@@ -633,25 +602,10 @@ if [[ "$DRY_RUN" == true ]]; then
            echo "Error: Cannot determine host from origin remote URL" >&2
            exit 1
        }
-        # Report the acting principal the merge WOULD use, resolved the same
-        # way the real merge resolves it (#1280) — a dry run that names a
-        # different principal than the merge would act as is a lie.
-        if ! principal_resolved="$(resolve_gitea_principal "$LOGIN_OVERRIDE" "$HOST")"; then
-            # Fail-loud diagnostic already printed (unresolvable --login or a
-            # requested identity with no per-slot token).
-            exit 1
-        fi
-        DRY_PRINCIPAL_MODE="$(printf '%s' "$principal_resolved" | cut -f1)"
-        DRY_PRINCIPAL_NAME="$(printf '%s' "$principal_resolved" | cut -f2)"
-        case "$DRY_PRINCIPAL_MODE" in
-            login)    DRY_PRINCIPAL_DESC="tea login '$DRY_PRINCIPAL_NAME'" ;;
-            identity) DRY_PRINCIPAL_DESC="git identity '$DRY_PRINCIPAL_NAME' (per-slot credential)" ;;
-            *)        DRY_PRINCIPAL_DESC="default host credential" ;;
-        esac
        if [[ "$CO_AUTHOR_TRAILERS" == true ]]; then
-            echo "Dry run: would verify PR commit authors and merge PR #$PR_NUMBER on $HOST as $DRY_PRINCIPAL_DESC with authenticated Gitea API message fields (base=$BASE_BRANCH, method=squash)."
+            echo "Dry run: would verify PR commit authors and merge PR #$PR_NUMBER on $HOST with authenticated Gitea API message fields (base=$BASE_BRANCH, method=squash)."
        else
-            echo "Dry run: would merge PR #$PR_NUMBER on $HOST as $DRY_PRINCIPAL_DESC with the authenticated exact-head Gitea API path (base=$BASE_BRANCH, method=squash)."
+            echo "Dry run: would merge PR #$PR_NUMBER on $HOST with the authenticated exact-head Gitea API path (base=$BASE_BRANCH, method=squash)."
        fi
    else
        echo "Dry run: would merge PR #$PR_NUMBER on $PLATFORM (base=$BASE_BRANCH, method=squash)."
@@ -76,7 +76,7 @@ while [[ $# -gt 0 ]]; do
            echo "  -n, --number   PR number (required)"
            echo "  -a, --action   Review action: approve, request-changes, comment (required)"
            echo "  -c, --comment  Review comment (required for request-changes)"
-            echo "  -l, --login    Override the detected Gitea tea login (all actions; wins over MOSAIC_GIT_IDENTITY)"
+            echo "  -l, --login    Override the detected Gitea tea login (approve/request-changes only)"
            echo "  -r, --repo     Explicit owner/repo slug (skips git-remote slug inference)"
            echo "  -H, --host     Explicit Gitea host (skips remote-host inference)"
            echo "  -h, --help     Show this help"
@@ -346,14 +346,7 @@ gitea_resolve_api_for_login() {
    else
        host=$(get_remote_host)
    fi
-    if [[ "$override_explicit" == "identity" ]]; then
-        # Requested git identity (#1280): the per-slot token MUST resolve via
-        # get_gitea_token's identity arm; never borrow the tea default login.
-        GITEA_API_TOKEN=$(get_gitea_token "$host") || {
-            echo "Error: could not resolve the per-slot token for requested git identity '$effective_login' on host '$host'; refusing to fall back to the tea login list or shared credentials (review write/read-back, #1280)." >&2
-            return 1
-        }
-    elif [[ -n "$override_explicit" ]]; then
+    if [[ -n "$override_explicit" ]]; then
        GITEA_API_TOKEN=$(get_gitea_token_for_login "$effective_login" "$host") || {
            echo "Error: could not resolve a host-matched Gitea token for --login '$effective_login' on host '$host'; refusing to fall back to the host default identity or a cross-host credential (review write/read-back)" >&2
            return 1
@@ -683,32 +676,29 @@ if [[ "$PLATFORM" == "github" ]]; then
            ;;
    esac
 elif [[ "$PLATFORM" == "gitea" ]]; then
-    # Resolve the acting principal ONCE for every action, identity-first
-    # (#1280): an explicit --login wins; otherwise MOSAIC_GIT_IDENTITY /
-    # per-worktree git config mosaic.gitIdentity selects the principal when a
-    # per-slot token exists (fail-loud when it does not); the tea login list is
-    # the LAST resort — it enumerates whatever logins this host happens to hold
-    # and knows nothing about which seat is calling, so resolving from it first
-    # wrote under whichever account tea had configured (the #1280 family).
-    principal_host="${HOST_OVERRIDE:-$(get_remote_host 2>/dev/null || true)}"
-    if ! principal_resolved="$(resolve_gitea_principal "$LOGIN_OVERRIDE" "$principal_host")"; then
-        # resolve_gitea_principal already printed the fail-loud diagnostic.
-        exit 1
-    fi
-    PRINCIPAL_MODE="$(printf '%s' "$principal_resolved" | cut -f1)"
-    PRINCIPAL_NAME="$(printf '%s' "$principal_resolved" | cut -f2)"
    case $ACTION in
        approve)
-            # Identity-first principal resolution (#1280): PRINCIPAL_MODE /
-            # PRINCIPAL_NAME were resolved once above from --login >
-            # MOSAIC_GIT_IDENTITY / git config > tea login list (last resort).
-            if [[ "$PRINCIPAL_MODE" == "identity" ]]; then
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" identity || exit 1
-            elif [[ "$PRINCIPAL_MODE" == "login" ]]; then
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" explicit || exit 1
-            else
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" "" || exit 1
-            fi
+            # Best-effort host for the tea-login GUESS only (gitea_resolve_api_for_login
+            # below re-derives the real host from HOST_OVERRIDE/remote independently and
+            # is authoritative). Prefer an explicit -H/--host; otherwise best-effort
+            # git-remote inference, tolerating its ABSENCE (a bare `get_remote_host` here
+            # under `set -e`, with no origin and no -H, previously killed the script
+            # SILENTLY — exit 1, zero output — even though -r/-H are exactly the flags
+            # that support running with no usable origin at all).
+            host="${HOST_OVERRIDE:-$(get_remote_host 2>/dev/null || true)}"
+            # A --login override always wins. Otherwise name this host's login
+            # only as a best effort: the login name merely selects a per-login
+            # token, and gitea_resolve_api_for_login falls back to the host
+            # credential (get_gitea_token) when no tea login is named — so a host
+            # tea's login list need not enumerate exotic (e.g. ported) hosts for
+            # the default credential to resolve. The single resolved token is
+            # then used for the write, the /user identity, and the read-back.
+            EFFECTIVE_LOGIN="$LOGIN_OVERRIDE"
+            [[ -n "$EFFECTIVE_LOGIN" ]] || EFFECTIVE_LOGIN=$(get_gitea_login_for_host "$host" 2>/dev/null || true)
+            # Bind the REST endpoint + token to the effective login, then derive
+            # the acting identity from that SAME credential so the review submit
+            # and its read-back verify against the identity that performed them.
+            gitea_resolve_api_for_login "$EFFECTIVE_LOGIN" "${LOGIN_OVERRIDE:+explicit}" || exit 1
            ACTING_LOGIN=$(gitea_authenticated_login) || exit 1
            head_sha=$(gitea_pr_head_sha "$PR_NUMBER") || exit 1
            # The review body (if any) travels with the review itself in the REST
@@ -725,16 +715,24 @@ elif [[ "$PLATFORM" == "gitea" ]]; then
                echo "Error: Comment required for request-changes"
                exit 1
            fi
-            # Identity-first principal resolution (#1280): PRINCIPAL_MODE /
-            # PRINCIPAL_NAME were resolved once above from --login >
-            # MOSAIC_GIT_IDENTITY / git config > tea login list (last resort).
-            if [[ "$PRINCIPAL_MODE" == "identity" ]]; then
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" identity || exit 1
-            elif [[ "$PRINCIPAL_MODE" == "login" ]]; then
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" explicit || exit 1
-            else
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" "" || exit 1
-            fi
+            # Best-effort host for the tea-login GUESS only (gitea_resolve_api_for_login
+            # below re-derives the real host from HOST_OVERRIDE/remote independently and
+            # is authoritative). Prefer an explicit -H/--host; otherwise best-effort
+            # git-remote inference, tolerating its ABSENCE (a bare `get_remote_host` here
+            # under `set -e`, with no origin and no -H, previously killed the script
+            # SILENTLY — exit 1, zero output — even though -r/-H are exactly the flags
+            # that support running with no usable origin at all).
+            host="${HOST_OVERRIDE:-$(get_remote_host 2>/dev/null || true)}"
+            # A --login override always wins. Otherwise name this host's login
+            # only as a best effort: the login name merely selects a per-login
+            # token, and gitea_resolve_api_for_login falls back to the host
+            # credential (get_gitea_token) when no tea login is named — so a host
+            # tea's login list need not enumerate exotic (e.g. ported) hosts for
+            # the default credential to resolve. The single resolved token is
+            # then used for the write, the /user identity, and the read-back.
+            EFFECTIVE_LOGIN="$LOGIN_OVERRIDE"
+            [[ -n "$EFFECTIVE_LOGIN" ]] || EFFECTIVE_LOGIN=$(get_gitea_login_for_host "$host" 2>/dev/null || true)
+            gitea_resolve_api_for_login "$EFFECTIVE_LOGIN" "${LOGIN_OVERRIDE:+explicit}" || exit 1
            ACTING_LOGIN=$(gitea_authenticated_login) || exit 1
            head_sha=$(gitea_pr_head_sha "$PR_NUMBER") || exit 1
            review_id=$(gitea_submit_review_verified "$PR_NUMBER" "REQUEST_CHANGES" "$COMMENT" "$ACTING_LOGIN" "$head_sha") || {
@@ -748,16 +746,24 @@ elif [[ "$PLATFORM" == "gitea" ]]; then
                echo "Error: Comment required"
                exit 1
            fi
-            # Identity-first principal resolution (#1280): PRINCIPAL_MODE /
-            # PRINCIPAL_NAME were resolved once above from --login >
-            # MOSAIC_GIT_IDENTITY / git config > tea login list (last resort).
-            if [[ "$PRINCIPAL_MODE" == "identity" ]]; then
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" identity || exit 1
-            elif [[ "$PRINCIPAL_MODE" == "login" ]]; then
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" explicit || exit 1
-            else
-                gitea_resolve_api_for_login "$PRINCIPAL_NAME" "" || exit 1
-            fi
+            # Best-effort host for the tea-login GUESS only (gitea_resolve_api_for_login
+            # below re-derives the real host from HOST_OVERRIDE/remote independently and
+            # is authoritative). Prefer an explicit -H/--host; otherwise best-effort
+            # git-remote inference, tolerating its ABSENCE (a bare `get_remote_host` here
+            # under `set -e`, with no origin and no -H, previously killed the script
+            # SILENTLY — exit 1, zero output — even though -r/-H are exactly the flags
+            # that support running with no usable origin at all).
+            host="${HOST_OVERRIDE:-$(get_remote_host 2>/dev/null || true)}"
+            # A --login override always wins. Otherwise name this host's login
+            # only as a best effort: the login name merely selects a per-login
+            # token, and gitea_resolve_api_for_login falls back to the host
+            # credential (get_gitea_token) when no tea login is named — so a host
+            # tea's login list need not enumerate exotic (e.g. ported) hosts for
+            # the default credential to resolve. The single resolved token is
+            # then used for the write, the /user identity, and the read-back.
+            EFFECTIVE_LOGIN="$LOGIN_OVERRIDE"
+            [[ -n "$EFFECTIVE_LOGIN" ]] || EFFECTIVE_LOGIN=$(get_gitea_login_for_host "$host" 2>/dev/null || true)
+            gitea_resolve_api_for_login "$EFFECTIVE_LOGIN" "${LOGIN_OVERRIDE:+explicit}" || exit 1
            ACTING_LOGIN=$(gitea_authenticated_login) || exit 1
            comment_id=$(gitea_create_comment_verified "$PR_NUMBER" "$COMMENT" "$ACTING_LOGIN") || {
                echo "Error: could not create and verify a comment on Gitea PR #$PR_NUMBER via a provider-returned created id (#865)." >&2
@@ -1,255 +0,0 @@
-#!/usr/bin/env bash
-# Regression harness for detect-platform.sh's resolve_gitea_principal() — the
-# identity-first acting-principal resolution shared by the git wrappers
-# (mosaicstack/stack #1280).
-#
-# The contract under test (precedence: --login > MOSAIC_GIT_IDENTITY /
-# git config mosaic.gitIdentity > tea login list, which is the LAST resort):
-#   1. identity env + per-slot token present -> mode=identity, principal=
-#      identity name, source names the identity's slot PATH (never a token
-#      value).
-#   2. identity env + per-slot token ABSENT -> FAIL LOUD: nonzero, empty
-#      stdout, stderr naming the identity and the expected slot path.
-#   3. identity env + --login -> --login wins (login mode resolves even when
-#      the identity has no slot — operator intent beats environment).
-#   4. identity unset + no --login -> default mode: the tea login list
-#      resolves the principal exactly as before (preserved behavior).
-#   5. no identity + no host-matching tea login -> default/host-credential
-#      (preserved behavior; absence is not an error on the default path).
-#   6. identity on an UNRECOGNIZED host (no per-slot scheme) -> does not bind;
-#      default mode (containment, mirroring get_gitea_token).
-#   7. --login with no host-bound token for that login -> FAIL LOUD, stderr
-#      naming the login and the host.
-#   8. git config mosaic.gitIdentity is honored when the env var is unset.
-#   9. The resolver NEVER emits a token value — stdout/stderr of every
-#      successful resolution must not contain the slot file's contents.
-#
-# Uses a stubbed tea binary, stubbed tea config.yml, stubbed credentials.json
-# and stubbed per-slot token files under a fake HOME. NEVER reads real secrets.
-
-set -euo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-WORK_DIR="${MOSAIC_TEST_WORK_DIR:-$PWD/.mosaic-test-work/gitea-principal-resolution}"
-FAKE_HOME="$WORK_DIR/home"
-REPO_DIR="$WORK_DIR/repo"
-BIN_DIR="$WORK_DIR/bin"
-CREDENTIALS_FILE="$FAKE_HOME/.config/mosaic/credentials.json"
-
-rm -rf "$WORK_DIR"
-mkdir -p "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens" "$FAKE_HOME/.config/tea" "$REPO_DIR" "$BIN_DIR"
-
-git -C "$REPO_DIR" init -q
-git -C "$REPO_DIR" remote add origin https://git.mosaicstack.dev/mosaicstack/stack.git
-
-cat > "$CREDENTIALS_FILE" <<'JSON'
-{
-  "gitea": {
-    "mosaicstack": {
-      "url": "https://git.mosaicstack.dev",
-      "token": "shared-mosaicstack-token"
-    },
-    "usc": {
-      "url": "https://git.uscllc.com",
-      "token": "shared-usc-token"
-    }
-  }
-}
-JSON
-
-# tea's own config store: the source get_gitea_token_for_login reads. Logins
-# "alice" (mosaicstack) and "bob-usc" (usc) carry sentinel token values that
-# the assertions prove are NEVER emitted by the resolver.
-cat > "$FAKE_HOME/.config/tea/config.yml" <<'YAML'
-logins:
- name: alice
-  url: https://git.mosaicstack.dev
-  token: SECRET-alice-tea-token
- name: bob-usc
-  url: https://git.uscllc.com
-  token: SECRET-bob-usc-tea-token
-YAML
-
-# Stubbed tea: only what login resolution needs (`login list --output json`).
-cat > "$BIN_DIR/tea" <<'SH'
-#!/usr/bin/env bash
-set -euo pipefail
-if [[ "$*" == "login list --output json" ]]; then
-    cat <<'JSON'
-[
-  {"name":"alice","url":"https://git.mosaicstack.dev","default":true},
-  {"name":"bob-usc","url":"https://git.uscllc.com"}
-]
-JSON
-    exit 0
-fi
-exit 0
-SH
-chmod +x "$BIN_DIR/tea"
-
-# Per-slot identity token with a sentinel value the assertions prove is never
-# emitted (proving "token came from the identity's slot BY PATH, not by value").
-echo -n "SECRET-agentX-slot-token" > "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentX.token"
-
-fail=0
-assert_eq() {
-  local desc="$1" expected="$2" actual="$3"
-  if [[ "$expected" != "$actual" ]]; then
-    echo "FAIL: $desc — expected '$expected', got '$actual'" >&2
-    fail=1
-  fi
-}
-assert_contains() {
-  local desc="$1" haystack="$2" needle="$3"
-  if [[ "$haystack" != *"$needle"* ]]; then
-    echo "FAIL: $desc — missing '$needle' in: $haystack" >&2
-    fail=1
-  fi
-}
-assert_not_contains() {
-  local desc="$1" haystack="$2" needle="$3"
-  if [[ "$haystack" == *"$needle"* ]]; then
-    echo "FAIL: $desc — must not contain '$needle', got: $haystack" >&2
-    fail=1
-  fi
-}
-
-# Runs resolve_gitea_principal for $1=login_override $2=host inside REPO_DIR
-# (per-worktree git config resolves there) under a fake HOME, stubbed tea, and
-# stubbed credentials. Extra env (e.g. MOSAIC_GIT_IDENTITY) via $@.
-call_resolver() {
-  local login="$1" host="$2"; shift 2
-  (
-    cd "$REPO_DIR"
-    env -i HOME="$FAKE_HOME" PATH="$BIN_DIR:$PATH" \
-      GIT_CONFIG_GLOBAL=/dev/null GIT_CONFIG_SYSTEM=/dev/null \
-      MOSAIC_CREDENTIALS_FILE="$CREDENTIALS_FILE" \
-      DETECT_PLATFORM_SH="$SCRIPT_DIR/detect-platform.sh" "$@" \
-      bash -c 'source "$DETECT_PLATFORM_SH"; resolve_gitea_principal "$1" "$2"' _ "$login" "$host"
-  )
-}
-
-field() { printf '%s' "$1" | cut -f"$2"; }
-
-# ---------------------------------------------------------------------------
-# 1. Identity env + slot present -> identity mode, slot named BY PATH, and no
-#    token value ever emitted.
-# ---------------------------------------------------------------------------
-git -C "$REPO_DIR" config --unset mosaic.gitIdentity 2>/dev/null || true
-out=$(call_resolver "" "git.mosaicstack.dev" MOSAIC_GIT_IDENTITY=agentX)
-assert_eq "identity mode"            "identity" "$(field "$out" 1)"
-assert_eq "identity principal"       "agentX"   "$(field "$out" 2)"
-assert_eq "identity slot source" \
-  "identity-slot:$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentX.token" \
-  "$(field "$out" 3)"
-assert_not_contains "identity stdout leaks token" "$out" "SECRET"
-
-# ---------------------------------------------------------------------------
-# 2. Identity env + slot ABSENT -> fail loud: nonzero, empty stdout, stderr
-#    naming the identity and the expected slot path.
-# ---------------------------------------------------------------------------
-stderr_file="$WORK_DIR/stderr.tmp"
-set +e
-out=$(call_resolver "" "git.mosaicstack.dev" MOSAIC_GIT_IDENTITY=agentNoSlot 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -eq 0 ]]; then
-  echo "FAIL: missing slot — expected nonzero return, got 0 (stdout='$out')" >&2
-  fail=1
-fi
-if [[ -n "$out" ]]; then
-  echo "FAIL: missing slot — expected empty stdout, got '$out'" >&2
-  fail=1
-fi
-err=$(cat "$stderr_file")
-assert_contains "missing slot names identity"  "$err" "agentNoSlot"
-assert_contains "missing slot names slot path" "$err" \
-  "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentNoSlot.token"
-assert_not_contains "missing-slot stderr leaks token" "$err" "SECRET"
-
-# ---------------------------------------------------------------------------
-# 3. Identity + --login -> --login wins. Also wins when the identity has NO
-#    slot (no identity check may veto an explicit login).
-# ---------------------------------------------------------------------------
-out=$(call_resolver "alice" "git.mosaicstack.dev" MOSAIC_GIT_IDENTITY=agentX)
-assert_eq "login beats identity (mode)"      "login" "$(field "$out" 1)"
-assert_eq "login beats identity (principal)" "alice" "$(field "$out" 2)"
-assert_eq "login source" "tea-login:alice" "$(field "$out" 3)"
-out=$(call_resolver "alice" "git.mosaicstack.dev" MOSAIC_GIT_IDENTITY=agentNoSlot)
-assert_eq "login beats slot-less identity" "login" "$(field "$out" 1)"
-
-# ---------------------------------------------------------------------------
-# 4. No identity, no --login -> default mode via the tea login list
-#    (preserved behavior).
-# ---------------------------------------------------------------------------
-out=$(call_resolver "" "git.mosaicstack.dev")
-assert_eq "default mode"          "default" "$(field "$out" 1)"
-assert_eq "default principal"     "alice"   "$(field "$out" 2)"
-assert_eq "default source"        "tea-default" "$(field "$out" 3)"
-
-# ---------------------------------------------------------------------------
-# 5. No identity, no --login, no host-matching tea login -> default with the
-#    host credential (absence is not an error on the default path).
-# ---------------------------------------------------------------------------
-out=$(call_resolver "" "git.unknown.test")
-assert_eq "no-match default mode"  "default" "$(field "$out" 1)"
-assert_eq "no-match default principal" "" "$(field "$out" 2)"
-assert_eq "no-match default source" "host-credential" "$(field "$out" 3)"
-
-# ---------------------------------------------------------------------------
-# 6. Identity on an UNRECOGNIZED host -> does not bind; default mode
-#    (containment, mirroring get_gitea_token's scope).
-# ---------------------------------------------------------------------------
-out=$(call_resolver "" "github.com" MOSAIC_GIT_IDENTITY=agentX)
-assert_eq "unrecognized host falls to default" "default" "$(field "$out" 1)"
-
-# ---------------------------------------------------------------------------
-# 7. --login with no host-bound token for that login -> fail loud, stderr
-#    naming the login and the host.
-# ---------------------------------------------------------------------------
-: > "$stderr_file"
-set +e
-out=$(call_resolver "ghost-login" "git.mosaicstack.dev" 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -eq 0 ]]; then
-  echo "FAIL: unknown --login — expected nonzero return, got 0 (stdout='$out')" >&2
-  fail=1
-fi
-err=$(cat "$stderr_file")
-assert_contains "unknown login names login" "$err" "ghost-login"
-assert_contains "unknown login names host"  "$err" "git.mosaicstack.dev"
-# A cross-host login (exists, but for usc) must ALSO fail loud for mosaicstack.
-set +e
-out=$(call_resolver "bob-usc" "git.mosaicstack.dev" 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -eq 0 ]]; then
-  echo "FAIL: cross-host --login — expected nonzero return, got 0" >&2
-  fail=1
-fi
-
-# ---------------------------------------------------------------------------
-# 8. git config mosaic.gitIdentity honored when env is unset.
-# ---------------------------------------------------------------------------
-git -C "$REPO_DIR" config mosaic.gitIdentity agentX
-out=$(call_resolver "" "git.mosaicstack.dev")
-assert_eq "git-config identity mode" "identity" "$(field "$out" 1)"
-assert_eq "git-config identity principal" "agentX" "$(field "$out" 2)"
-git -C "$REPO_DIR" config --unset mosaic.gitIdentity
-
-# ---------------------------------------------------------------------------
-# 9. Cross-host slot layout: the usc slot path is chosen for the usc host.
-# ---------------------------------------------------------------------------
-echo -n "SECRET-agentX-usc-token" > "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-usc-agentX.token"
-out=$(call_resolver "" "git.uscllc.com" MOSAIC_GIT_IDENTITY=agentX)
-assert_eq "usc identity mode" "identity" "$(field "$out" 1)"
-assert_eq "usc slot source" \
-  "identity-slot:$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-usc-agentX.token" \
-  "$(field "$out" 3)"
-
-if [[ "$fail" -eq 0 ]]; then
-  echo "resolve_gitea_principal identity-first resolution regression passed"
-fi
-
-exit "$fail"
@@ -77,30 +77,12 @@ exit 0
 SH
 chmod +x "$BIN_DIR/tea"

-# TRIPWIRE provider stub: this harness tests argv construction, so ANY curl
-# call is a failure of that contract (and, before this stub existed, a LIVE
-# write — the #1282–#1287 incident: the seat's real HOME leaked a global
-# mosaic.gitIdentity, flipping the wrapper into identity mode whose real
-# per-slot token created real issues on the forge). Fail loudly instead.
-cat > "$BIN_DIR/curl" <<'SH'
-#!/usr/bin/env bash
-echo "FAIL: body-safety harness reached a provider request — this test must never curl" >&2
-exit 99
-SH
-chmod +x "$BIN_DIR/curl"
-
-# Hermetic invocation: fake HOME (no credentials, no tea config, no token
-# slots) and GIT_CONFIG_GLOBAL severed — `git config --get mosaic.gitIdentity`
-# otherwise resolves the WORKSTATION's global identity (mos-dt-0 on the seat
-# that wrote this) and reroutes the wrapper into identity mode (#1280 family).
 (
    cd "$REPO_DIR"
-    env -i HOME="$WORK_DIR/home" PATH="$BIN_DIR:$PATH" \
-      GIT_CONFIG_GLOBAL=/dev/null GIT_CONFIG_SYSTEM=/dev/null \
-      MOSAIC_TEST_RECEIVED="$RECEIVED_FILE" \
-      "$SCRIPT_DIR/issue-create.sh" -t "Body safety test" -b "$BODY"
+    PATH="$BIN_DIR:$PATH" \
+    MOSAIC_TEST_RECEIVED="$RECEIVED_FILE" \
+    "$SCRIPT_DIR/issue-create.sh" -t "Body safety test" -b "$BODY"
 ) >/dev/null
-mkdir -p "$WORK_DIR/home"

 # 1. No command substitution executed anywhere in the pipeline.
 if [[ -e "$SENTINEL" ]]; then
@@ -47,31 +47,14 @@ SH
 chmod +x "$BIN_DIR/tea" "$BIN_DIR/curl"

 run_wrapper() {
-  # Hermetic: fake HOME (fixture credentials only, no token slots, no tea
-  # config) and GIT_CONFIG_GLOBAL severed — `git config --get
-  # mosaic.gitIdentity` otherwise resolves the WORKSTATION's global identity
-  # and reroutes the wrapper into identity mode before the tea paths this
-  # harness exercises (#1280 family; see test-issue-create-body-safety.sh).
-  # An `env …` prefix (used for MOSAIC_TEA_STALE_USER) is re-wrapped, not
-  # doubled: arguments beginning with "env" are shifted past.
-  local env_pairs=()
-  if [[ "${1:-}" == "env" ]]; then
-    shift
-    while [[ "$#" -gt 0 && "$1" == *=* ]]; do
-      env_pairs+=("$1")
-      shift
-    done
-  fi
  (
    cd "$REPO_DIR"
-    env -i HOME="$WORK_DIR/home" PATH="$BIN_DIR:$PATH" \
-      GIT_CONFIG_GLOBAL=/dev/null GIT_CONFIG_SYSTEM=/dev/null \
-      MOSAIC_CREDENTIALS_FILE="$CREDENTIALS_FILE" \
-      MOSAIC_TEST_LOG="$LOG_FILE" "${env_pairs[@]}" \
-      "$@"
+    PATH="$BIN_DIR:$PATH" \
+    MOSAIC_CREDENTIALS_FILE="$CREDENTIALS_FILE" \
+    MOSAIC_TEST_LOG="$LOG_FILE" \
+    "$@"
  )
 }
-mkdir -p "$WORK_DIR/home"

 : > "$LOG_FILE"
 printf 'Interactive title\nInteractive body\nlabel-a,label-b\nM1\n' | run_wrapper "$SCRIPT_DIR/issue-create.sh" -i >/dev/null
@@ -1,244 +0,0 @@
-#!/usr/bin/env bash
-# Load-bearing regression harness for pr-create.sh identity-first principal
-# resolution (mosaicstack/stack #1280).
-#
-# The failure this harness is written down to catch: `MOSAIC_GIT_IDENTITY=fargo
-# pr-create.sh …` produces a PR attributed to `mos-dt-0` (whichever account the
-# tea login list happens to hold). Before #1280 the identity-aware code existed
-# but sat on the API arm that only ran when the tea path FAILED — tea succeeded,
-# so the identity arm never executed, and every test that did not check ORDERING
-# passed. This harness checks ordering directly:
-#
-#   1. identity set + slot present -> the PR is created via the REST API with
-#      the identity's per-slot token (asserted by sentinel value AT the fake
-#      provider), and tea's `pr create` is NEVER invoked.
-#   2. identity set + slot ABSENT -> nonzero, stderr naming the identity and
-#      the expected slot path; neither tea `pr create` nor any API request
-#      fires. No silent fallback to the tea login list.
-#   3. identity set + --login -> --login wins: tea runs WITH the explicit
-#      --login, no API request.
-#   4. nothing set -> preserved behavior: tea path with the tea-list login.
-#
-# Uses a stubbed tea, a stubbed curl provider, stubbed credentials.json and
-# per-slot token under a fake HOME. NEVER reads real secrets or hits a live
-# forge — all assertions are against the stubs' logs.
-
-set -euo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-WORK_DIR="${MOSAIC_TEST_WORK_DIR:-$PWD/.mosaic-test-work/pr-create-identity-first}"
-FAKE_HOME="$WORK_DIR/home"
-REPO_DIR="$WORK_DIR/repo"
-TOOLS_DIR="$WORK_DIR/tools"
-BIN_DIR="$WORK_DIR/bin"
-LOG_FILE="$WORK_DIR/calls.log"
-CREDENTIALS_FILE="$FAKE_HOME/.config/mosaic/credentials.json"
-
-rm -rf "$WORK_DIR"
-mkdir -p "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens" "$FAKE_HOME/.config/tea" \
-  "$REPO_DIR" "$TOOLS_DIR/git" "$TOOLS_DIR/_lib" "$BIN_DIR"
-
-# Fixture: the real scripts under test, copied so sibling stubs (and the
-# ../_lib credential loader) resolve inside the fixture tree.
-cp "$SCRIPT_DIR/pr-create.sh" "$TOOLS_DIR/git/pr-create.sh"
-cp "$SCRIPT_DIR/detect-platform.sh" "$TOOLS_DIR/git/detect-platform.sh"
-cp "$SCRIPT_DIR/../_lib/credentials.sh" "$TOOLS_DIR/_lib/credentials.sh"
-chmod +x "$TOOLS_DIR/git/pr-create.sh"
-
-git -C "$REPO_DIR" init -q
-git -C "$REPO_DIR" remote add origin https://git.mosaicstack.dev/mosaicstack/stack.git
-
-cat > "$CREDENTIALS_FILE" <<'JSON'
-{
-  "gitea": {
-    "mosaicstack": {
-      "url": "https://git.mosaicstack.dev",
-      "token": "shared-mosaicstack-token"
-    }
-  }
-}
-JSON
-
-cat > "$FAKE_HOME/.config/tea/config.yml" <<'YAML'
-logins:
- name: alice
-  url: https://git.mosaicstack.dev
-  token: SECRET-alice-tea-token
-YAML
-
-echo -n "SECRET-agentX-slot-token" > "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentX.token"
-
-: > "$LOG_FILE"
-
-# Stubbed tea: records every invocation; `login list` feeds login resolution;
-# `api --login <n> /user` feeds get_gitea_authenticated_user; `pr create` marks
-# the marker file (its presence fails the identity-mode assertions).
-cat > "$BIN_DIR/tea" <<SH
-#!/usr/bin/env bash
-set -euo pipefail
-printf 'TEA: %s\n' "\$*" >> "$LOG_FILE"
-if [[ "\$*" == "login list --output json" ]]; then
-    cat <<'JSON'
-[
-  {"name":"alice","url":"https://git.mosaicstack.dev","default":true}
-]
-JSON
-    exit 0
-fi
-if [[ "\${1:-}" == "api" ]]; then
-    printf '%s\n' '{"login":"alice"}'
-    exit 0
-fi
-if [[ "\$*" == pr\ create* ]]; then
-    echo "TEA-PR-CREATE-INVOKED" >> "$LOG_FILE"
-    exit 0
-fi
-exit 0
-SH
-chmod +x "$BIN_DIR/tea"
-
-# Stubbed provider: records the URL and the Authorization header VALUE it
-# received, answers 201 with a created-PR object. The sentinel token values are
-# synthetic fixtures — asserting them at the provider proves WHICH slot's
-# credential carried the write.
-cat > "$BIN_DIR/curl" <<SH
-#!/usr/bin/env bash
-set -euo pipefail
-url=""
-auth=""
-while [[ \$# -gt 0 ]]; do
-    case "\$1" in
-        -H)
-            case "\$2" in
-                Authorization*) auth="\$2" ;;
-            esac
-            shift 2
-            ;;
-        *)  [[ -n "\$1" && "\$1" != -* ]] && url="\$1"
-            shift
-            ;;
-    esac
-done
-printf 'CURL-URL: %s\nCURL-AUTH: %s\n' "\$url" "\$auth" >> "$LOG_FILE"
-cat <<'JSON'
-{"number": 1299, "html_url": "https://git.mosaicstack.dev/mosaicstack/stack/pulls/1299"}
-JSON
-exit 0
-SH
-chmod +x "$BIN_DIR/curl"
-
-fail=0
-assert_contains() {
-  local desc="$1" needle="$2"
-  if ! grep -qF -- "$needle" "$LOG_FILE"; then
-    echo "FAIL: $desc — log does not contain '$needle':" >&2
-    cat "$LOG_FILE" >&2
-    fail=1
-  fi
-}
-assert_not_contains() {
-  local desc="$1" needle="$2"
-  if grep -qF -- "$needle" "$LOG_FILE"; then
-    echo "FAIL: $desc — log must not contain '$needle':" >&2
-    cat "$LOG_FILE" >&2
-    fail=1
-  fi
-}
-
-EXTRA_ARGS=""
-run_pr_create() {
-  # "$@" carries ONLY environment assignments (VAR=value); EXTRA_ARGS (if
-  # set) carries wrapper arguments, so `env` never mistakes a wrapper flag
-  # like --login for one of its own.
-  (
-    cd "$REPO_DIR"
-    # shellcheck disable=SC2086 # EXTRA_ARGS is deliberately word-split wrapper args
-    env -i HOME="$FAKE_HOME" PATH="$BIN_DIR:$PATH" \
-      GIT_CONFIG_GLOBAL=/dev/null GIT_CONFIG_SYSTEM=/dev/null \
-      MOSAIC_CREDENTIALS_FILE="$CREDENTIALS_FILE" "$@" \
-      bash "$TOOLS_DIR/git/pr-create.sh" -t "Test PR" -B next -H fix/test $EXTRA_ARGS
-  )
-}
-
-# ---------------------------------------------------------------------------
-# 1. HAPPY PATH (the load-bearing ordering test): identity set + slot present
-#    -> REST API with the per-slot token; tea `pr create` NEVER invoked.
-# ---------------------------------------------------------------------------
-set +e
-out=$(run_pr_create MOSAIC_GIT_IDENTITY=agentX 2>"$WORK_DIR/stderr-1.tmp")
-rc=$?
-set -e
-if [[ "$rc" -ne 0 ]]; then
-  echo "FAIL: identity happy path — expected rc=0, got $rc" >&2
-  cat "$WORK_DIR/stderr-1.tmp" >&2
-  fail=1
-fi
-assert_contains "identity happy path reaches the API" "CURL-URL: https://git.mosaicstack.dev/api/v1/repos/mosaicstack/stack/pulls"
-assert_contains "identity happy path carries the slot token" "CURL-AUTH: Authorization: token SECRET-agentX-slot-token"
-assert_not_contains "identity happy path must NOT invoke tea pr create" "TEA-PR-CREATE-INVOKED"
-
-# ---------------------------------------------------------------------------
-# 2. Identity set + slot ABSENT -> fail loud BEFORE any write: nonzero, stderr
-#    naming identity + slot path, no tea pr create, no API request.
-# ---------------------------------------------------------------------------
-: > "$LOG_FILE"
-set +e
-out=$(run_pr_create MOSAIC_GIT_IDENTITY=agentNoSlot 2>"$WORK_DIR/stderr-2.tmp")
-rc=$?
-set -e
-if [[ "$rc" -eq 0 ]]; then
-  echo "FAIL: missing slot — expected nonzero return, got 0 (stdout='$out')" >&2
-  fail=1
-fi
-err=$(cat "$WORK_DIR/stderr-2.tmp")
-if [[ "$err" != *"agentNoSlot"* ]]; then
-  echo "FAIL: missing slot — stderr does not name the identity:" >&2
-  echo "$err" >&2
-  fail=1
-fi
-if [[ "$err" != *"$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentNoSlot.token"* ]]; then
-  echo "FAIL: missing slot — stderr does not name the expected slot path:" >&2
-  echo "$err" >&2
-  fail=1
-fi
-assert_not_contains "missing slot must not reach tea pr create" "TEA-PR-CREATE-INVOKED"
-assert_not_contains "missing slot must not reach the API" "CURL-URL"
-
-# ---------------------------------------------------------------------------
-# 3. Identity set + --login -> --login wins: tea runs WITH the explicit login.
-# ---------------------------------------------------------------------------
-: > "$LOG_FILE"
-EXTRA_ARGS="--login alice"
-set +e
-out=$(run_pr_create MOSAIC_GIT_IDENTITY=agentX 2>"$WORK_DIR/stderr-3.tmp")
-rc=$?
-set -e
-EXTRA_ARGS=""
-if [[ "$rc" -ne 0 ]]; then
-  echo "FAIL: login override — expected rc=0, got $rc" >&2
-  cat "$WORK_DIR/stderr-3.tmp" >&2
-  fail=1
-fi
-assert_contains "login override drives tea with the explicit login" "TEA: pr create --repo mosaicstack/stack --login alice"
-assert_not_contains "login override must not hit the API" "CURL-URL"
-
-# ---------------------------------------------------------------------------
-# 4. Nothing set -> preserved behavior: tea path with the tea-list login.
-# ---------------------------------------------------------------------------
-: > "$LOG_FILE"
-set +e
-out=$(run_pr_create 2>"$WORK_DIR/stderr-4.tmp")
-rc=$?
-set -e
-if [[ "$rc" -ne 0 ]]; then
-  echo "FAIL: default path — expected rc=0, got $rc" >&2
-  cat "$WORK_DIR/stderr-4.tmp" >&2
-  fail=1
-fi
-assert_contains "default path still uses the tea-list login" "TEA: pr create --repo mosaicstack/stack --login alice"
-
-if [[ "$fail" -eq 0 ]]; then
-  echo "pr-create identity-first happy-path regression passed"
-fi
-
-exit "$fail"
@@ -1,247 +0,0 @@
-#!/usr/bin/env bash
-# Regression harness for pr-merge.sh identity-first principal resolution
-# (mosaicstack/stack #1280).
-#
-# Covers:
-#   1. --dry-run reports the acting principal the merge WOULD use, resolved the
-#      same way the real merge resolves it: --login > MOSAIC_GIT_IDENTITY /
-#      git config mosaic.gitIdentity > shared host credential. (The pre-#1280
-#      deployed copy reported a tea login that the merge would not act as.)
-#   2. --dry-run fails closed when the requested principal has no credential:
-#      unknown --login, or an identity with no per-slot token (stderr names
-#      the login / the identity and its slot path).
-#   3. The real merge POST carries the resolved principal's credential and no
-#      other: --login merges with that login's tea-config token; an identity
-#      merges with the per-slot token; an unresolvable --login never reaches
-#      the provider.
-#
-# Fixture pattern from test-pr-merge-head-pin.sh: the scripts under test are
-# copied into a fixture tree with stubbed pr-metadata.sh / ci-queue-wait.sh
-# siblings; the provider is a stubbed curl that records the credential it
-# received. NEVER reads real secrets or hits a live forge.
-
-set -euo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-WORK_DIR="${MOSAIC_TEST_WORK_DIR:-$PWD/.mosaic-test-work/pr-merge-principal-resolution}"
-FAKE_HOME="$WORK_DIR/home"
-REPO_DIR="$WORK_DIR/repo"
-TOOLS_DIR="$WORK_DIR/tools"
-BIN_DIR="$WORK_DIR/bin"
-LOG_FILE="$WORK_DIR/calls.log"
-CREDENTIALS_FILE="$FAKE_HOME/.config/mosaic/credentials.json"
-SHA=0123456789abcdef0123456789abcdef01234567
-
-rm -rf "$WORK_DIR"
-mkdir -p "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens" "$FAKE_HOME/.config/tea" \
-  "$REPO_DIR" "$TOOLS_DIR/git" "$TOOLS_DIR/_lib" "$BIN_DIR"
-
-cp "$SCRIPT_DIR/pr-merge.sh" "$TOOLS_DIR/git/pr-merge.sh"
-cp "$SCRIPT_DIR/detect-platform.sh" "$TOOLS_DIR/git/detect-platform.sh"
-cp "$SCRIPT_DIR/../_lib/credentials.sh" "$TOOLS_DIR/_lib/credentials.sh"
-chmod +x "$TOOLS_DIR/git/pr-merge.sh"
-
-git -C "$REPO_DIR" init -q
-git -C "$REPO_DIR" remote add origin https://git.mosaicstack.dev/mosaicstack/stack.git
-
-# Stubbed siblings pr-merge.sh resolves relative to its own SCRIPT_DIR.
-cat > "$TOOLS_DIR/git/pr-metadata.sh" <<SH
-#!/usr/bin/env bash
-printf '%s\n' '{"baseRefName":"next","headRefName":"fix/pinned","headRefOid":"$SHA","headRepository":"mosaicstack/stack","title":"Test PR","author":{"login":"contributor"}}'
-SH
-cat > "$TOOLS_DIR/git/ci-queue-wait.sh" <<'SH'
-#!/usr/bin/env bash
-exit 0
-SH
-chmod +x "$TOOLS_DIR/git/pr-metadata.sh" "$TOOLS_DIR/git/ci-queue-wait.sh"
-
-cat > "$CREDENTIALS_FILE" <<'JSON'
-{
-  "gitea": {
-    "mosaicstack": {
-      "url": "https://git.mosaicstack.dev",
-      "token": "shared-mosaicstack-token"
-    }
-  }
-}
-JSON
-
-cat > "$FAKE_HOME/.config/tea/config.yml" <<'YAML'
-logins:
- name: fred-ms
-  url: https://git.mosaicstack.dev
-  token: SECRET-fred-ms-tea-token
-YAML
-
-echo -n "SECRET-agentX-slot-token" > "$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentX.token"
-
-: > "$LOG_FILE"
-
-# Stubbed tea for login-list resolution only.
-cat > "$BIN_DIR/tea" <<'SH'
-#!/usr/bin/env bash
-set -euo pipefail
-if [[ "$*" == "login list --output json" ]]; then
-    cat <<'JSON'
-[
-  {"name":"fred-ms","url":"https://git.mosaicstack.dev","default":true}
-]
-JSON
-    exit 0
-fi
-exit 0
-SH
-chmod +x "$BIN_DIR/tea"
-
-# Stubbed provider. pr-merge passes curl config on STDIN with -K -; the stub
-# reads stdin, records the Authorization header it received, answers 200.
-cat > "$BIN_DIR/curl" <<SH
-#!/usr/bin/env bash
-set -euo pipefail
-url=""
-out_file=""
-stdin_config=""
-if [[ ! -t 0 ]]; then
-    stdin_config="\$(cat || true)"
-fi
-while [[ \$# -gt 0 ]]; do
-    case "\$1" in
-        -o) out_file="\$2"; shift 2 ;;
-        -K|-w|--max-filesize|--max-time|--connect-timeout|-sS) shift 2 ;;
-        *)  [[ -n "\$1" && "\$1" != -* && -z "\$url" ]] && url="\$1"
-            shift
-            ;;
-    esac
-done
-auth="\$(printf '%s' "\$stdin_config" | grep -o 'Authorization: token [^"]*' || true)"
-printf 'CURL-URL: %s\nCURL-AUTH: %s\n' "\$url" "\$auth" >> "$LOG_FILE"
-[[ -n "\$out_file" ]] && printf '{}' > "\$out_file"
-printf '200\n'
-exit 0
-SH
-chmod +x "$BIN_DIR/curl"
-
-fail=0
-assert_contains_log() {
-  local desc="$1" needle="$2"
-  if ! grep -qF -- "$needle" "$LOG_FILE"; then
-    echo "FAIL: $desc — log does not contain '$needle':" >&2
-    cat "$LOG_FILE" >&2
-    fail=1
-  fi
-}
-assert_not_contains_log() {
-  local desc="$1" needle="$2"
-  if grep -qF -- "$needle" "$LOG_FILE"; then
-    echo "FAIL: $desc — log must not contain '$needle':" >&2
-    cat "$LOG_FILE" >&2
-    fail=1
-  fi
-}
-
-run_pr_merge() {
-  local extra_args="$1"; shift
-  (
-    cd "$REPO_DIR"
-    # shellcheck disable=SC2086 # extra_args is deliberately word-split wrapper args
-    env -i HOME="$FAKE_HOME" PATH="$BIN_DIR:$PATH" \
-      GIT_CONFIG_GLOBAL=/dev/null GIT_CONFIG_SYSTEM=/dev/null \
-      MOSAIC_CREDENTIALS_FILE="$CREDENTIALS_FILE" "$@" \
-      bash "$TOOLS_DIR/git/pr-merge.sh" -n 42 $extra_args
-  )
-}
-
-# ---------------------------------------------------------------------------
-# 1. --dry-run reports the resolved acting principal truthfully.
-# ---------------------------------------------------------------------------
-out=$(run_pr_merge "--dry-run" MOSAIC_GIT_IDENTITY=agentX)
-if [[ "$out" != *"as git identity 'agentX' (per-slot credential)"* ]]; then
-  echo "FAIL: dry-run identity — principal not reported: $out" >&2
-  fail=1
-fi
-
-out=$(run_pr_merge "--dry-run --login fred-ms" MOSAIC_GIT_IDENTITY=agentX)
-if [[ "$out" != *"as tea login 'fred-ms'"* ]]; then
-  echo "FAIL: dry-run login override — login not reported (must beat env identity): $out" >&2
-  fail=1
-fi
-
-out=$(run_pr_merge "--dry-run")
-if [[ "$out" != *"as default host credential"* ]]; then
-  echo "FAIL: dry-run default — not reported: $out" >&2
-  fail=1
-fi
-
-# ---------------------------------------------------------------------------
-# 2. --dry-run fails closed when the requested principal has no credential.
-# ---------------------------------------------------------------------------
-stderr_file="$WORK_DIR/stderr.tmp"
-set +e
-out=$(run_pr_merge "--dry-run --login ghost" 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -eq 0 ]] || [[ "$(cat "$stderr_file")" != *"ghost"* ]]; then
-  echo "FAIL: dry-run unknown --login — expected fail-loud naming 'ghost', rc=$rc" >&2
-  cat "$stderr_file" >&2
-  fail=1
-fi
-
-: > "$stderr_file"
-set +e
-out=$(run_pr_merge "--dry-run" MOSAIC_GIT_IDENTITY=agentNoSlot 2>"$stderr_file")
-rc=$?
-set -e
-err=$(cat "$stderr_file")
-if [[ "$rc" -eq 0 ]] || [[ "$err" != *"agentNoSlot"* ]] \
-  || [[ "$err" != *"$FAKE_HOME/.config/mosaic/secrets/gitea-tokens/gitea-mosaicstack-agentNoSlot.token"* ]]; then
-  echo "FAIL: dry-run identity without slot — expected fail-loud naming identity + slot path, rc=$rc" >&2
-  echo "$err" >&2
-  fail=1
-fi
-
-# ---------------------------------------------------------------------------
-# 3. The real merge POST carries the resolved principal's credential ONLY.
-# ---------------------------------------------------------------------------
-: > "$LOG_FILE"
-set +e
-out=$(run_pr_merge "--login fred-ms" MOSAIC_GIT_IDENTITY=agentX 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -ne 0 ]]; then
-  echo "FAIL: merge with --login — expected rc=0, got $rc" >&2
-  cat "$stderr_file" >&2
-  fail=1
-fi
-assert_contains_log "merge --login uses the login token" "CURL-AUTH: Authorization: token SECRET-fred-ms-tea-token"
-assert_not_contains_log "merge --login must not use the identity slot token" "SECRET-agentX-slot-token"
-assert_not_contains_log "merge --login must not use the shared token" "shared-mosaicstack-token"
-
-: > "$LOG_FILE"
-set +e
-out=$(run_pr_merge "" MOSAIC_GIT_IDENTITY=agentX 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -ne 0 ]]; then
-  echo "FAIL: merge with identity — expected rc=0, got $rc" >&2
-  cat "$stderr_file" >&2
-  fail=1
-fi
-assert_contains_log "merge identity uses the per-slot token" "CURL-AUTH: Authorization: token SECRET-agentX-slot-token"
-assert_not_contains_log "merge identity must not use the shared token" "shared-mosaicstack-token"
-
-: > "$LOG_FILE"
-set +e
-out=$(run_pr_merge "--login ghost" 2>"$stderr_file")
-rc=$?
-set -e
-if [[ "$rc" -eq 0 ]]; then
-  echo "FAIL: merge with unknown --login — expected nonzero, got 0" >&2
-  fail=1
-fi
-assert_not_contains_log "merge with unknown --login must not reach the provider" "CURL-URL"
-
-if [[ "$fail" -eq 0 ]]; then
-  echo "pr-merge identity-first principal resolution regression passed"
-fi
-
-exit "$fail"
@@ -39,20 +39,3 @@ 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/git/test-gitea-principal-resolution.sh && bash framework/tools/git/test-pr-create-identity-first.sh && bash framework/tools/git/test-pr-merge-principal-resolution.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"
+    "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"
  },
  "dependencies": {
    "@mosaicstack/brain": "workspace:*",