sdd-agent-platform 0.3.0 → 0.4.1

package/node_modules/@sdd-agent-platform/core/src/verification/test-runtime.ts

@@ -0,0 +1,1032 @@
+import { createHash } from 'node:crypto';
+import { spawn } from 'node:child_process';
+import { appendEvent } from '../run-state/events.js';
+import { appendInvocationLedgerEntry } from '../run-state/invocation-ledger.js';
+import type { RunState, RunStateTaskRuntime } from '../run-state/model.js';
+import { createRun, readRunState, writeRunState } from '../run-state/run-state.js';
+import { writeArtifact } from '../run-state/artifacts.js';
+import { resolveSddContext } from '../sdd-docs/context.js';
+import { bindRunStateToTask } from '../sdd-docs/run-binding.js';
+import { parseSddBranch, type SddTask } from '../sdd-docs/task-parser.js';
+import { inspectSddTask } from '../sdd-docs/task-inspection.js';
+import { recordRuntimeAcceptanceEvidenceMap, recordRuntimeDurableGap, recordRuntimeProjection, recordRuntimeTestRun, recordRuntimeTestStep, recordRuntimeValidationEnvironmentSession, recordRuntimeValidationWaveRun, runtimeScopedId, updateRuntimeDurableGapStatus } from '../storage/runtime-store.js';
+import { ACCEPTANCE_POLICY_RULESET_VERSION, SDD_EVIDENCE_CONTRACT, SDD_EVIDENCE_VERSION, SDD_RESULT_CONTRACT, SDD_RESULT_VERSION, TEST_EVIDENCE_RUN_CONTRACT_VERSION, WORKFLOW_HANDOFF_CONTRACT_VERSION } from '../contracts.js';
+import type { LifecycleRiskDecision } from '../risk/contracts.js';
+import { inspectVerifyContract, type VerifyContractInspection } from './verify-contract.js';
+import type { AcceptanceEvidenceCoverage, CapabilityEvidenceClassification, EvidenceCoverageStatus, TestEvidenceStatus, UnifiedTestEvidenceRun } from '../evidence-runtime.js';
+import { ensureTaskOrchestration, inspectOrchestrationGate } from '../orchestration/runtime.js';
+import { recordStageRunProjection, recordWorkflowHandoffProjection, validateWorkflowHandoff } from '../stage-runtime/runtime.js';
+import type { StageRun, WorkflowHandoff } from '../stage-runtime/contracts.js';
+import { evaluateTaskWorkflowGate, verifyContractBlockedGate, type ApprovalPolicy, type LifecycleRiskProfile, type LifecycleWorkflowGate } from '../risk.js';
+import { validateSddResultArtifact } from '../artifacts/sdd-result.js';
+import { dependencyBlockingReasonsForTask } from '../workflow-state/dependencies.js';
+import { routeSddTask } from '../router/route-sdd-task.js';
+import type { AgentCapabilityRouteDecision } from '../router/agent-runtime.js';
+import { evaluateAndRecordWorkflowGateDecision } from '../workflow-gate/evidence-packet.js';
+import type { WorkflowGateDecision, WorkflowGateStatus } from '../workflow-gate/types.js';
+
+const DEFAULT_TEST_TIMEOUT_MS = 120_000;
+const MAX_CAPTURE_BYTES = 256 * 1024;
+
+export type SddTestStatus = 'PASS' | 'FAIL' | 'BLOCKED';
+export type SddTestStepStatus = 'pass' | 'fail' | 'blocked';
+
+export interface SddTestCommandInput {
+  command?: string;
+  argv?: string[];
+}
+
+export interface SddTestCommandStep {
+  stepId: string;
+  command: string;
+  argv: string[] | null;
+  shell: boolean;
+  acceptanceRefs: string[];
+  status: SddTestStepStatus;
+  exitCode: number | null;
+  signal: string | null;
+  durationMs: number;
+  outputArtifact: string;
+  stdoutBytes: number;
+  stderrBytes: number;
+  truncated: boolean;
+}
+
+interface NormalizedSddTestCommand {
+  command: string;
+  argv: string[] | null;
+  shell: boolean;
+}
+
+type VerifyContractAction = 'none' | 'created' | 'refreshed' | 'blocked';
+
+type RuntimeTestJudgment = WorkflowGateStatus;
+
+export interface SddTestResult {
+  contract: 'sdd-test-runtime-v1';
+  runId: string;
+  testRunId: string;
+  validationWaveRunId: string;
+  validationEnvironmentSessionId: string;
+  branch: string;
+  taskId: string;
+  status: SddTestStatus;
+  validationStatus: SddTestStatus;
+  workflowGateStatus: WorkflowGateStatus;
+  runtimeJudgment: RuntimeTestJudgment;
+  workflowGateDecision: WorkflowGateDecision;
+  verifyContractStatus: string;
+  verifyContractAction: VerifyContractAction;
+  lifecycleGate: LifecycleWorkflowGate;
+  lifecycleProfile: LifecycleRiskProfile | null;
+  approvalPolicy: ApprovalPolicy | null;
+  requiredStages: string[];
+  primaryReason: string;
+  commandStatus: TestEvidenceStatus;
+  evidenceCoverage: EvidenceCoverageStatus;
+  policyJudgment: TestEvidenceStatus;
+  acceptanceCoverage: AcceptanceEvidenceCoverage[];
+  capabilityEvidence: CapabilityEvidenceClassification[];
+  syncBackReady: boolean;
+  commands: string[];
+  steps: SddTestCommandStep[];
+  validationArtifact: string | null;
+  indexArtifact: string | null;
+  gaps: string[];
+  next: string;
+}
+
+export interface RunSddTestOptions {
+  taskId: string;
+  branch?: string | null;
+  runId?: string | null;
+  commands?: string[];
+  commandInputs?: SddTestCommandInput[];
+  timeoutMs?: number;
+  approved?: boolean;
+  validationWave?: { waveRunId: string; environmentSessionId: string; taskIds: string[]; acceptanceRefsByTask?: Record<string, string[]> };
+}
+
+export async function runSddTest(projectRoot: string, options: RunSddTestOptions): Promise<SddTestResult> {
+  const context = await resolveSddContext(projectRoot, { branch: options.branch ?? undefined, branchSource: options.branch ? 'cli_option' : undefined });
+  const model = await parseSddBranch(projectRoot, context.partition);
+  const inspected = inspectSddTask(model, options.taskId);
+  const task = inspected.task;
+  const verifyContract = await ensureVerifyContractForTest(projectRoot, context.partition);
+  const verifyInspection = verifyContract.inspection;
+  const initialState = options.runId ? await readRunState(projectRoot, options.runId) : await createRun(projectRoot);
+  const state = await bindRunStateToTask(projectRoot, initialState, context, model, task, options.taskId);
+  const testRunId = runtimeScopedId(state.runId, options.taskId, new Date().toISOString(), 'sdd-test');
+  const commandInputs = normalizeTestCommandInputs(options.commandInputs, options.commands, task?.validation ?? []);
+  const commands = commandInputs.map((input) => input.command);
+  const gaps: string[] = [];
+  const startedAt = new Date().toISOString();
+  const ownsValidationWave = !options.validationWave;
+  const validationWaveRunId = options.validationWave?.waveRunId ?? runtimeScopedId(context.partition, options.taskId, state.runId, testRunId, 'validation-wave');
+  const validationEnvironmentSessionId = options.validationWave?.environmentSessionId ?? runtimeScopedId(context.partition, validationWaveRunId, 'validation-env');
+  const validationWaveTaskIds = options.validationWave?.taskIds ?? [options.taskId];
+  const validationWaveAcceptanceRefs = options.validationWave?.acceptanceRefsByTask?.[options.taskId];
+  if (ownsValidationWave) {
+    await recordRuntimeValidationEnvironmentSession(projectRoot, {
+      sessionId: validationEnvironmentSessionId,
+      partition: context.partition,
+      runId: state.runId,
+      waveRunId: validationWaveRunId,
+      status: 'active',
+      reuseKey: `${context.partition}:${options.taskId}`,
+      createdAt: startedAt,
+      updatedAt: startedAt,
+      payload: { contract: 'phase-8.17-validation-wave-runtime-v1', mode: 'single-task' }
+    });
+    await recordRuntimeValidationWaveRun(projectRoot, {
+      waveRunId: validationWaveRunId,
+      partition: context.partition,
+      runId: state.runId,
+      taskIds: validationWaveTaskIds,
+      status: 'RUNNING',
+      environmentSessionId: validationEnvironmentSessionId,
+      startedAt,
+      completedAt: startedAt,
+      payload: { contract: 'phase-8.17-validation-wave-runtime-v1', mode: 'single-task', taskId: options.taskId }
+    });
+  }
+
+  await appendEvent(projectRoot, state.runId, {
+    event: 'test_runtime_started',
+    runId: state.runId,
+    summary: `SDD test runtime started for ${options.taskId}`,
+    data: { taskId: options.taskId, branch: context.partition, testRunId, commands }
+  });
+
+  if (!task) {
+    gaps.push(`Task ${options.taskId} was not found in specs/${context.partition}/tasks.md.`);
+  }
+  if (task) {
+    gaps.push(...inspected.gaps.filter((gap) => gap.severity === 'blocking').map((gap) => `${gap.field}: ${gap.message}`));
+    gaps.push(...dependencyBlockingReasonsForTask(model, options.taskId));
+  }
+  if (verifyContract.action === 'blocked') {
+    gaps.push(verifyContractBlocker(verifyInspection));
+  }
+  if (commands.length === 0) {
+    gaps.push(`Task ${options.taskId} has no validation commands.`);
+  }
+  const orchestration = await ensureTaskOrchestration(projectRoot, model, task, {
+    branch: context.partition,
+    runId: state.runId,
+    taskId: options.taskId,
+    agent: 'validator',
+    stage: 'test',
+    status: 'active'
+  });
+  const orchestrationGate = await inspectOrchestrationGate(projectRoot, {
+    branch: context.partition,
+    runId: state.runId,
+    taskId: options.taskId,
+    target: 'test',
+    riskDecision: orchestration.riskDecision,
+    stageRun: orchestration.stageRun,
+    contextLoadSignal: orchestration.contextLoadSignal,
+    contextOffloadDecision: orchestration.contextOffloadDecision
+  });
+  const reviewerCheckpointSatisfied = await hasReviewerCheckpoint(projectRoot, state, options.taskId);
+  const workflowGate = verifyContract.action === 'blocked'
+    ? verifyContractBlockedGate(options.taskId)
+    : evaluateTaskWorkflowGate({ task, taskId: options.taskId, riskDecision: orchestration.riskDecision, approved: options.approved, reviewerCheckpointSatisfied });
+  if (workflowGate.blocksTest) {
+    gaps.push(workflowGate.primaryReason);
+  }
+  gaps.push(...orchestrationGate.blockingReasons);
+
+  await recordRuntimeTestRun(projectRoot, {
+    testRunId,
+    runId: state.runId,
+    partition: context.partition,
+    taskId: options.taskId,
+    status: 'RUNNING',
+    startedAt,
+    completedAt: startedAt,
+    payload: { verifyContractStatus: verifyInspection.status, verifyContractAction: verifyContract.action, lifecycleGate: workflowGate.lifecycleGate, lifecycleProfile: workflowGate.lifecycleProfile, approvalPolicy: workflowGate.approvalPolicy, requiredStages: workflowGate.requiredStages, primaryReason: workflowGate.primaryReason, commands, commandInputs, evidence: [], gaps }
+  });
+
+  const steps: SddTestCommandStep[] = [];
+  if (gaps.length === 0) {
+    for (const [index, commandInput] of commandInputs.entries()) {
+      const step = await runCommandStep(projectRoot, state.runId, context.partition, options.taskId, testRunId, index + 1, commandInput, acceptanceRefsForCommand(task, commandInput.command, validationWaveAcceptanceRefs), options.timeoutMs ?? DEFAULT_TEST_TIMEOUT_MS);
+      steps.push(step);
+      await appendEvent(projectRoot, state.runId, {
+        event: 'test_step_completed',
+        runId: state.runId,
+        summary: `SDD test step ${step.status}: ${step.command}`,
+        data: { taskId: options.taskId, testRunId, step }
+      });
+    }
+  }
+
+  const commandStatus = deriveCommandStatus(gaps, steps);
+  const acceptanceCoverage = buildAcceptanceCoverage(task, steps, commandStatus, validationWaveAcceptanceRefs);
+  const evidenceCoverage = summarizeEvidenceCoverage(acceptanceCoverage);
+  const policyJudgment = derivePolicyJudgment(commandStatus, evidenceCoverage);
+  await recordAcceptanceEvidenceMaps(projectRoot, validationWaveRunId, testRunId, context.partition, state.runId, options.taskId, acceptanceCoverage);
+  const validationStatus = policyJudgment;
+  const syncBackReady = false;
+  const capabilityRoute = task ? await routeSddTask(projectRoot, { taskId: options.taskId, branch: context.partition, approved: options.approved }) : null;
+  const capabilityEvidence = buildCapabilityEvidenceClassification(capabilityRoute?.capabilityDecision ?? null, steps);
+  const validationArtifact = task ? await writeValidationArtifact(projectRoot, state.runId, task, validationStatus, steps, gaps, capabilityEvidence) : null;
+  const evidenceBeforeIndex = [validationArtifact?.runRelativePath, ...steps.map((step) => step.outputArtifact)].filter((item): item is string => Boolean(item));
+  await persistTestRunState(projectRoot, state, options.taskId, validationStatus, commands, evidenceBeforeIndex, validationArtifact?.runRelativePath ?? null);
+  await resolveTestRuntimeDurableGap(projectRoot, context.partition, state.runId, options.taskId, validationStatus, gaps);
+
+  const gateDecision = (await evaluateAndRecordWorkflowGateDecision(projectRoot, {
+    branch: context.partition,
+    taskId: options.taskId,
+    runId: state.runId,
+    decisionKind: 'test'
+  })).decision;
+  const runtimeJudgment = gateDecision.status;
+  const status = finalStatusForTest(validationStatus, runtimeJudgment);
+  await recordTestRuntimeDurableGap(projectRoot, context.partition, state.runId, options.taskId, status, validationStatus, runtimeJudgment, gaps, evidenceBeforeIndex);
+  const unifiedEvidence = buildUnifiedTestEvidenceRun(testRunId, context.partition, state.runId, options.taskId, commandStatus, evidenceCoverage, policyJudgment, status, runtimeJudgment, steps, acceptanceCoverage, capabilityEvidence, syncBackReady, gaps, workflowGate.nextAction, gateDecision);
+  const indexArtifact = await writeIndexArtifact(projectRoot, state.runId, {
+    testRunId,
+    validationWaveRunId,
+    validationEnvironmentSessionId,
+    branch: context.partition,
+    taskId: options.taskId,
+    status,
+    validationStatus,
+    workflowGateStatus: gateDecision.status,
+    runtimeJudgment,
+    workflowGateDecision: gateDecision,
+    verifyContractStatus: verifyInspection.status,
+    verifyContractAction: verifyContract.action,
+    lifecycleGate: workflowGate.lifecycleGate,
+    lifecycleProfile: workflowGate.lifecycleProfile,
+    approvalPolicy: workflowGate.approvalPolicy,
+    requiredStages: workflowGate.requiredStages,
+    primaryReason: workflowGate.primaryReason,
+    commandStatus,
+    evidenceCoverage,
+    policyJudgment,
+    acceptanceCoverage,
+    capabilityEvidence,
+    syncBackReady,
+    commands,
+    steps,
+    validationArtifact: validationArtifact?.runRelativePath ?? null,
+    gaps
+  });
+  const completedAt = new Date().toISOString();
+  const evidence = [validationArtifact?.runRelativePath, indexArtifact.runRelativePath, ...steps.map((step) => step.outputArtifact)].filter((item): item is string => Boolean(item));
+
+  await recordRuntimeTestRun(projectRoot, {
+    testRunId,
+    runId: state.runId,
+    partition: context.partition,
+    taskId: options.taskId,
+    status,
+    startedAt,
+    completedAt,
+    payload: { verifyContractStatus: verifyInspection.status, verifyContractAction: verifyContract.action, lifecycleGate: workflowGate.lifecycleGate, lifecycleProfile: workflowGate.lifecycleProfile, approvalPolicy: workflowGate.approvalPolicy, requiredStages: workflowGate.requiredStages, primaryReason: workflowGate.primaryReason, commandStatus, evidenceCoverage, policyJudgment, validationStatus, workflowGateStatus: gateDecision.status, runtimeJudgment, workflowGateDecision: gateDecision, acceptanceCoverage, capabilityEvidence, syncBackReady, commands, commandInputs, evidence, gaps }
+  });
+  await recordRuntimeProjection(projectRoot, 'test_runtime', `${context.partition}:${options.taskId}:${state.runId}`, {
+    contract: 'sdd-test-runtime-v1',
+    testRunId,
+    runId: state.runId,
+    taskId: options.taskId,
+    status,
+    validationStatus,
+    workflowGateStatus: gateDecision.status,
+    runtimeJudgment,
+    lifecycleGate: workflowGate.lifecycleGate,
+    primaryReason: workflowGate.primaryReason,
+    evidence,
+    gaps
+  });
+  await recordRuntimeProjection(projectRoot, 'test_evidence_run', `${context.partition}:${options.taskId}:${state.runId}`, unifiedEvidence);
+  await recordTestWorkflowProjection(projectRoot, {
+    taskId: options.taskId,
+    stageRun: orchestration.stageRun,
+    status,
+    completedAt,
+    evidence,
+    gaps,
+    riskDecision: orchestration.riskDecision
+  });
+  await persistTestGateOutcome(projectRoot, state.runId, options.taskId, status, validationStatus, commands, evidence, validationArtifact?.runRelativePath ?? null, gateDecision);
+  if (ownsValidationWave) {
+    await recordRuntimeValidationWaveRun(projectRoot, {
+      waveRunId: validationWaveRunId,
+      partition: context.partition,
+      runId: state.runId,
+      taskIds: validationWaveTaskIds,
+      status,
+      environmentSessionId: validationEnvironmentSessionId,
+      startedAt,
+      completedAt,
+      payload: { contract: 'phase-8.17-validation-wave-runtime-v1', mode: 'single-task', taskId: options.taskId, testRunId, evidence, gaps, workflowGateDecision: gateDecision }
+    });
+    await recordRuntimeValidationEnvironmentSession(projectRoot, {
+      sessionId: validationEnvironmentSessionId,
+      partition: context.partition,
+      runId: state.runId,
+      waveRunId: validationWaveRunId,
+      status: status === 'PASS' ? 'completed' : status === 'FAIL' ? 'failed' : 'blocked',
+      reuseKey: `${context.partition}:${options.taskId}`,
+      createdAt: startedAt,
+      updatedAt: completedAt,
+      payload: { contract: 'phase-8.17-validation-wave-runtime-v1', mode: 'single-task', taskId: options.taskId, validationStatus, workflowGateStatus: gateDecision.status, status }
+    });
+  }
+
+  await appendEvent(projectRoot, state.runId, {
+    event: status === 'PASS' ? 'test_runtime_passed' : 'test_runtime_blocked',
+    runId: state.runId,
+    summary: `SDD test runtime ${status} for ${options.taskId}`,
+    data: { taskId: options.taskId, testRunId, status, validationStatus, evidence, gaps, gateDecisionId: gateDecision.decisionId, gateStatus: gateDecision.status }
+  });
+
+  return {
+    contract: 'sdd-test-runtime-v1',
+    runId: state.runId,
+    testRunId,
+    validationWaveRunId,
+    validationEnvironmentSessionId,
+    branch: context.partition,
+    taskId: options.taskId,
+    status,
+    validationStatus,
+    workflowGateStatus: gateDecision.status,
+    runtimeJudgment,
+    workflowGateDecision: gateDecision,
+    verifyContractStatus: verifyInspection.status,
+    verifyContractAction: verifyContract.action,
+    lifecycleGate: workflowGate.lifecycleGate,
+    lifecycleProfile: workflowGate.lifecycleProfile,
+    approvalPolicy: workflowGate.approvalPolicy,
+    requiredStages: workflowGate.requiredStages,
+    primaryReason: workflowGate.primaryReason,
+    commandStatus,
+    evidenceCoverage,
+    policyJudgment,
+    acceptanceCoverage,
+    capabilityEvidence,
+    syncBackReady,
+    commands,
+    steps,
+    validationArtifact: validationArtifact?.runRelativePath ?? null,
+    indexArtifact: indexArtifact.runRelativePath,
+    gaps,
+    next: nextForTestResult(status, runtimeJudgment, context.partition, options.taskId, state.runId, indexArtifact.runRelativePath, workflowGate.nextAction, gateDecision)
+  };
+}
+
+async function recordAcceptanceEvidenceMaps(projectRoot: string, waveRunId: string, testRunId: string, partition: string, runId: string, taskId: string, acceptanceCoverage: AcceptanceEvidenceCoverage[]): Promise<void> {
+  const createdAt = new Date().toISOString();
+  for (const coverage of acceptanceCoverage) {
+    await recordRuntimeAcceptanceEvidenceMap(projectRoot, {
+      mapId: runtimeScopedId(waveRunId, testRunId, taskId, coverage.acceptanceRef),
+      waveRunId,
+      testRunId,
+      partition,
+      runId,
+      taskId,
+      acceptanceRef: coverage.acceptanceRef,
+      status: coverage.status,
+      evidenceRefs: coverage.evidenceRefs.map((ref) => ref.ref),
+      gaps: coverage.gaps,
+      createdAt,
+      payload: coverage
+    });
+  }
+}
+
+async function recordTestWorkflowProjection(projectRoot: string, input: { taskId: string; stageRun: StageRun; status: SddTestStatus; completedAt: string; evidence: string[]; gaps: string[]; riskDecision: LifecycleRiskDecision }): Promise<void> {
+  const outputRefs = input.evidence.map((ref) => ({ kind: 'artifact' as const, ref }));
+  const completedStage: StageRun = {
+    ...input.stageRun,
+    status: input.status === 'PASS' ? 'completed' : input.status === 'FAIL' ? 'failed' : 'blocked',
+    outputRefs,
+    blockingReasons: input.status === 'PASS' ? [] : input.gaps.length > 0 ? input.gaps : [`SDD test ${input.status}.`],
+    updatedAt: input.completedAt
+  };
+  await recordStageRunProjection(projectRoot, completedStage);
+
+  const handoff: WorkflowHandoff = {
+    contract: WORKFLOW_HANDOFF_CONTRACT_VERSION,
+    id: `${completedStage.id}:handoff:goal-verify`,
+    scope: completedStage.scope,
+    fromStage: 'test',
+    toStage: 'goal-verify',
+    fromAgent: 'validator',
+    toAgent: 'verifier',
+    status: input.status === 'PASS' ? 'proposed' : 'blocked',
+    outputRefs,
+    requiredInputRefs: [{ kind: 'task', ref: input.taskId }],
+    riskDecisionRef: input.stageRun.decisionRefs[0] ?? { kind: 'task', ref: input.taskId },
+    evidenceRefs: outputRefs,
+    openQuestions: [],
+    blockingGaps: input.status === 'PASS' ? [] : completedStage.blockingReasons,
+    createdAt: input.completedAt,
+    decidedAt: input.completedAt
+  };
+  const validation = validateWorkflowHandoff({ handoff, sourceStageRun: completedStage, lifecycleRiskDecision: input.riskDecision });
+  await recordWorkflowHandoffProjection(projectRoot, validation.valid ? handoff : { ...handoff, status: 'blocked', blockingGaps: validation.issues, decidedAt: input.completedAt });
+}
+
+
+async function ensureVerifyContractForTest(projectRoot: string, branch: string): Promise<{ inspection: VerifyContractInspection; action: VerifyContractAction }> {
+  const inspection = await inspectVerifyContract(projectRoot, { branch, branchSource: 'cli_option' });
+  return { inspection, action: inspection.status === 'PASS' ? 'none' : 'blocked' };
+}
+
+function verifyContractBlocker(inspection: VerifyContractInspection): string {
+  const issueSummary = inspection.issues.map((issue) => `${issue.field}: ${issue.message}`).join(' ');
+  return `verify.md contract is ${inspection.status}; ${issueSummary || 'inspect verify.md before executing tests.'}`;
+}
+
+async function hasReviewerCheckpoint(projectRoot: string, state: RunState, taskId: string): Promise<boolean> {
+  const artifactPaths = new Set([
+    ...state.artifacts
+      .filter((artifact) => artifact.task === taskId && (artifact.agent === 'reviewer' || artifact.kind === 'review'))
+      .map((artifact) => artifact.path),
+    `artifacts/review-${taskId}.md`
+  ]);
+
+  for (const artifactPath of artifactPaths) {
+    const report = await validateSddResultArtifact(projectRoot, state.runId, artifactPath, { expectedTask: taskId, expectedAgent: 'reviewer' });
+    if (report.valid && report.result?.status === 'PASS') {
+      return true;
+    }
+  }
+
+  return false;
+}
+
+export function renderSddTestResult(result: SddTestResult): string {
+  return [
+    `SDD test ${result.taskId}`,
+    '',
+    resultSentenceForTest(result),
+    '',
+    'Decision:',
+    `- validation_status=${result.validationStatus}`,
+    `- workflow_gate_status=${result.workflowGateStatus}`,
+    `- workflow_gate_decision=${result.workflowGateDecision.decisionId}`,
+    '',
+    'Why:',
+    `- ${result.primaryReason}`,
+    `- capability_evidence=${capabilityEvidenceSummary(result.capabilityEvidence)}`,
+    '',
+    'Next:',
+    `- ${result.next}`
+  ].join('\n');
+}
+
+
+async function runCommandStep(projectRoot: string, runId: string, branch: string, taskId: string, testRunId: string, sequence: number, commandInput: NormalizedSddTestCommand, acceptanceRefs: string[], timeoutMs: number): Promise<SddTestCommandStep> {
+  const started = Date.now();
+  const executed = await executeCommand(projectRoot, commandInput, timeoutMs);
+  const durationMs = Date.now() - started;
+  const status: SddTestStepStatus = executed.timedOut || executed.error ? 'blocked' : executed.exitCode === 0 ? 'pass' : 'fail';
+  const stepId = `${testRunId}-${String(sequence).padStart(3, '0')}`;
+  const output = renderCommandOutput(commandInput, status, executed, durationMs);
+  const outputArtifact = await writeArtifact(projectRoot, runId, `test-${taskId}-${String(sequence).padStart(3, '0')}.log`, output);
+  await appendInvocationLedgerEntry(projectRoot, {
+    runId,
+    taskId,
+    branch,
+    kind: 'command',
+    ref: commandInput.command,
+    status,
+    artifactPath: outputArtifact.runRelativePath,
+    outputHash: hashDocumentContent(output),
+    materialRefs: [outputArtifact.runRelativePath],
+    metadata: {
+      source: 'sdd-test',
+      exitCode: executed.exitCode,
+      durationMs,
+      stdoutBytes: executed.stdoutBytes,
+      stderrBytes: executed.stderrBytes,
+      truncated: executed.truncated,
+      acceptanceRefs: acceptanceRefs.join(','),
+      shell: commandInput.shell,
+      argv: commandInput.argv ? JSON.stringify(commandInput.argv) : null
+    }
+  });
+  const step: SddTestCommandStep = {
+    stepId,
+    command: commandInput.command,
+    argv: commandInput.argv,
+    shell: commandInput.shell,
+    acceptanceRefs,
+    status,
+    exitCode: executed.exitCode,
+    signal: executed.signal,
+    durationMs,
+    outputArtifact: outputArtifact.runRelativePath,
+    stdoutBytes: executed.stdoutBytes,
+    stderrBytes: executed.stderrBytes,
+    truncated: executed.truncated
+  };
+  await recordRuntimeTestStep(projectRoot, {
+    stepId,
+    testRunId,
+    runId,
+    taskId,
+    command: commandInput.command,
+    status,
+    exitCode: executed.exitCode,
+    durationMs,
+    outputArtifact: outputArtifact.runRelativePath,
+    payload: step
+  });
+  return step;
+}
+
+function executeCommand(projectRoot: string, commandInput: NormalizedSddTestCommand, timeoutMs: number): Promise<{ exitCode: number | null; signal: string | null; stdout: string; stderr: string; stdoutBytes: number; stderrBytes: number; truncated: boolean; timedOut: boolean; error: string | null }> {
+  return new Promise((resolve) => {
+    const child = commandInput.argv
+      ? spawn(commandInput.argv[0], commandInput.argv.slice(1), { cwd: projectRoot, shell: false, windowsHide: true, env: process.env })
+      : spawn(commandInput.command, { cwd: projectRoot, shell: true, windowsHide: true, env: process.env });
+    let stdout = '';
+    let stderr = '';
+    let stdoutBytes = 0;
+    let stderrBytes = 0;
+    let truncated = false;
+    let settled = false;
+    let timedOut = false;
+    const timer = setTimeout(() => {
+      timedOut = true;
+      child.kill();
+    }, timeoutMs);
+    const finish = (result: { exitCode: number | null; signal: string | null; error: string | null }) => {
+      if (settled) {
+        return;
+      }
+      settled = true;
+      clearTimeout(timer);
+      resolve({ ...result, stdout, stderr, stdoutBytes, stderrBytes, truncated, timedOut });
+    };
+    child.stdout?.on('data', (chunk: Buffer) => {
+      stdoutBytes += chunk.length;
+      const next = chunk.toString('utf8');
+      if (Buffer.byteLength(stdout, 'utf8') < MAX_CAPTURE_BYTES) {
+        stdout += next;
+      } else {
+        truncated = true;
+      }
+    });
+    child.stderr?.on('data', (chunk: Buffer) => {
+      stderrBytes += chunk.length;
+      const next = chunk.toString('utf8');
+      if (Buffer.byteLength(stderr, 'utf8') < MAX_CAPTURE_BYTES) {
+        stderr += next;
+      } else {
+        truncated = true;
+      }
+    });
+    child.on('error', (error) => finish({ exitCode: null, signal: null, error: error.message }));
+    child.on('close', (code, signal) => finish({ exitCode: code, signal, error: null }));
+  });
+}
+
+async function writeValidationArtifact(projectRoot: string, runId: string, task: SddTask, status: SddTestStatus, steps: SddTestCommandStep[], gaps: string[], capabilityEvidence: CapabilityEvidenceClassification[]): Promise<{ absolutePath: string; runRelativePath: string }> {
+  const artifactPath = `test-validation-${task.id}.md`;
+  const runRelativePath = `artifacts/${artifactPath}`;
+  const resultStatus = status === 'PASS' ? 'PASS' : status === 'FAIL' ? 'FAIL' : 'BLOCKED';
+  const content = `# Test Validation ${task.id}\n\n\`\`\`sdd-result\ncontract: ${SDD_RESULT_CONTRACT}\nversion: ${SDD_RESULT_VERSION}\nagent: validator\ntask: ${task.id}\nstatus: ${resultStatus}\nartifacts:\n  - ${runRelativePath}\n\`\`\`\n\n## Test Runtime\n\n- status: ${status}\n- commands:\n${steps.length > 0 ? steps.map((step) => `  - [${step.status}] ${step.command}`).join('\n') : '  - none'}\n- gaps:\n${gaps.length > 0 ? gaps.map((gap) => `  - ${gap}`).join('\n') : '  - none'}\n\n## Capability Evidence Classification\n\n${renderCapabilityEvidenceClassification(capabilityEvidence)}\n\n## Acceptance Evidence\n\n${renderEvidenceBlocks(task, status, runRelativePath, steps)}\n`;
+  return writeArtifact(projectRoot, runId, artifactPath, content);
+}
+
+
+async function writeIndexArtifact(projectRoot: string, runId: string, payload: Omit<SddTestResult, 'contract' | 'runId' | 'indexArtifact' | 'next'>): Promise<{ absolutePath: string; runRelativePath: string }> {
+  return writeArtifact(projectRoot, runId, `test-index-${payload.taskId}.json`, `${JSON.stringify({ contract: 'sdd-test-runtime-v1', runId, ...payload }, null, 2)}\n`);
+}
+
+async function persistTestRunState(projectRoot: string, state: RunState, taskId: string, validationStatus: SddTestStatus, commands: string[], evidence: string[], validationArtifact: string | null): Promise<void> {
+  const latest = await readRunState(projectRoot, state.runId);
+  const knownArtifacts = new Set(latest.artifacts.map((artifact) => artifact.path));
+  const now = new Date().toISOString();
+  const nextArtifacts = evidence
+    .filter((artifactPath) => !knownArtifacts.has(artifactPath))
+    .map((artifactPath) => ({ path: artifactPath, kind: testArtifactKind(artifactPath), task: taskId, agent: 'test-runtime', createdAt: now }));
+  const existingTaskState = latest.tasks[taskId];
+  await writeRunState(projectRoot, {
+    ...latest,
+    status: validationStatus === 'PASS' ? 'running' : validationStatus === 'FAIL' ? 'failed' : 'blocked',
+    phase: 'test',
+    currentTask: taskId,
+    tasks: {
+      ...latest.tasks,
+      [taskId]: {
+        ...baseRuntimeTaskState(existingTaskState),
+        status: validationStatus === 'PASS' ? 'validation_passed_pending_gate' : validationStatus === 'FAIL' ? 'validation_failed' : 'validation_blocked',
+        implementationStatus: existingTaskState?.implementationStatus ?? 'implemented',
+        verificationStatus: verificationStatusFromTest(validationStatus),
+        testStatus: validationStatus,
+        evidence
+      }
+    },
+    artifacts: [...latest.artifacts, ...nextArtifacts],
+    validation: {
+      status: validationStatus === 'PASS' ? 'pass' : validationStatus === 'FAIL' ? 'fail' : 'blocked',
+      commands,
+      evidence
+    },
+    syncBack: latest.syncBack
+  });
+}
+
+async function persistTestGateOutcome(projectRoot: string, runId: string, taskId: string, status: SddTestStatus, validationStatus: SddTestStatus, commands: string[], evidence: string[], validationArtifact: string | null, gateDecision: WorkflowGateDecision): Promise<void> {
+  const latest = await readRunState(projectRoot, runId);
+  const knownArtifacts = new Set(latest.artifacts.map((artifact) => artifact.path));
+  const now = new Date().toISOString();
+  const nextArtifacts = evidence
+    .filter((artifactPath) => !knownArtifacts.has(artifactPath))
+    .map((artifactPath) => ({ path: artifactPath, kind: testArtifactKind(artifactPath), task: taskId, agent: 'test-runtime', createdAt: now }));
+  const existingTaskState = latest.tasks[taskId];
+  await writeRunState(projectRoot, {
+    ...latest,
+    status: status === 'PASS' ? 'completed' : status === 'FAIL' ? 'failed' : 'blocked',
+    phase: 'test',
+    currentTask: taskId,
+    tasks: {
+      ...latest.tasks,
+      [taskId]: {
+        ...baseRuntimeTaskState(existingTaskState),
+        status: runtimeTaskStatusAfterGate(status, validationStatus),
+        implementationStatus: existingTaskState?.implementationStatus ?? 'implemented',
+        verificationStatus: verificationStatusFromTest(validationStatus),
+        testStatus: status,
+        validationStatus,
+        workflowGateStatus: gateDecision.status,
+        workflowGateDecisionId: gateDecision.decisionId,
+        evidence
+      }
+    },
+    artifacts: [...latest.artifacts, ...nextArtifacts],
+    validation: {
+      status: validationStatus === 'PASS' ? 'pass' : validationStatus === 'FAIL' ? 'fail' : 'blocked',
+      commands,
+      evidence
+    },
+    syncBack: latest.syncBack
+  });
+}
+
+function testArtifactKind(artifactPath: string): string {
+  const fileName = artifactPath.split('/').pop() ?? artifactPath;
+  if (fileName.startsWith('test-validation-')) {
+    return 'test-validation';
+  }
+  if (fileName.startsWith('test-index-')) {
+    return 'test-index';
+  }
+  return 'test';
+}
+
+function baseRuntimeTaskState(existing: RunStateTaskRuntime | undefined): RunStateTaskRuntime {
+  return {
+    status: existing?.status ?? 'not_started',
+    implementationStatus: existing?.implementationStatus ?? 'not_started',
+    verificationStatus: existing?.verificationStatus ?? 'not_run',
+    validationBatch: existing?.validationBatch ?? null,
+    validationTiming: existing?.validationTiming ?? 'task_end',
+    requiresVerifyBeforeNext: existing?.requiresVerifyBeforeNext ?? true,
+    gaps: existing?.gaps,
+    artifacts: existing?.artifacts,
+    testStatus: existing?.testStatus,
+    workflowGateStatus: existing?.workflowGateStatus,
+    workflowGateDecisionId: existing?.workflowGateDecisionId,
+    evidence: existing?.evidence
+  };
+}
+
+function verificationStatusFromTest(validationStatus: SddTestStatus): RunStateTaskRuntime['verificationStatus'] {
+  if (validationStatus === 'PASS') {
+    return 'pass';
+  }
+  return validationStatus === 'FAIL' ? 'failed' : 'blocked';
+}
+
+function runtimeTaskStatusAfterGate(status: SddTestStatus, validationStatus: SddTestStatus): string {
+  if (status === 'PASS') {
+    return 'implemented_verified';
+  }
+  if (validationStatus === 'PASS') {
+    return 'workflow_gate_blocked';
+  }
+  return validationStatus === 'FAIL' ? 'validation_failed' : 'validation_blocked';
+}
+
+function deriveCommandStatus(gaps: string[], steps: SddTestCommandStep[]): TestEvidenceStatus {
+  if (gaps.length > 0 || steps.some((step) => step.status === 'blocked')) {
+    return 'BLOCKED';
+  }
+  if (steps.some((step) => step.status === 'fail')) {
+    return 'FAIL';
+  }
+  return 'PASS';
+}
+
+function buildAcceptanceCoverage(task: SddTask | null, steps: SddTestCommandStep[], commandStatus: TestEvidenceStatus, acceptanceRefsOverride: string[] | undefined): AcceptanceEvidenceCoverage[] {
+  const acceptanceRefs = acceptanceRefsOverride && acceptanceRefsOverride.length > 0 ? [...new Set(acceptanceRefsOverride)] : task ? taskAcceptanceRefs(task) : [];
+  return acceptanceRefs.map((acceptanceRef) => {
+    const mappedSteps = steps.filter((step) => step.acceptanceRefs.includes(acceptanceRef));
+    const hasPassingEvidence = mappedSteps.some((step) => step.status === 'pass');
+    const hasFailingEvidence = mappedSteps.some((step) => step.status === 'fail' || step.status === 'blocked');
+    const status: EvidenceCoverageStatus = hasPassingEvidence && !hasFailingEvidence && commandStatus === 'PASS'
+      ? 'complete'
+      : mappedSteps.length > 0
+        ? 'partial'
+        : 'missing';
+    return {
+      acceptanceRef,
+      status,
+      evidenceRefs: mappedSteps.map((step) => ({ kind: 'artifact', ref: step.outputArtifact })),
+      gaps: status === 'complete' ? [] : [`Acceptance ${acceptanceRef} has no complete non-stale evidence from sdd test task.`]
+    };
+  });
+}
+
+function summarizeEvidenceCoverage(acceptanceCoverage: AcceptanceEvidenceCoverage[]): EvidenceCoverageStatus {
+  if (acceptanceCoverage.length === 0) {
+    return 'missing';
+  }
+  if (acceptanceCoverage.every((coverage) => coverage.status === 'complete')) {
+    return 'complete';
+  }
+  if (acceptanceCoverage.some((coverage) => coverage.status === 'complete' || coverage.status === 'partial')) {
+    return 'partial';
+  }
+  return 'missing';
+}
+
+function derivePolicyJudgment(commandStatus: TestEvidenceStatus, evidenceCoverage: EvidenceCoverageStatus): SddTestStatus {
+  if (commandStatus === 'FAIL') {
+    return 'FAIL';
+  }
+  if (commandStatus === 'BLOCKED' || evidenceCoverage !== 'complete') {
+    return 'BLOCKED';
+  }
+  return 'PASS';
+}
+
+async function resolveTestRuntimeDurableGap(
+  projectRoot: string,
+  branch: string,
+  runId: string,
+  taskId: string,
+  validationStatus: SddTestStatus,
+  gaps: string[]
+): Promise<void> {
+  if (validationStatus !== 'PASS' || gaps.length > 0) {
+    return;
+  }
+  await updateRuntimeDurableGapStatus(projectRoot, {
+    gapId: testRuntimeGapId(branch, runId, taskId),
+    status: 'resolved',
+    source: 'gate_policy',
+    payload: { validationStatus, gaps }
+  });
+}
+
+async function recordTestRuntimeDurableGap(
+  projectRoot: string,
+  branch: string,
+  runId: string,
+  taskId: string,
+  status: SddTestStatus,
+  validationStatus: SddTestStatus,
+  runtimeJudgment: WorkflowGateStatus,
+  gaps: string[],
+  evidenceRefs: string[]
+): Promise<void> {
+  if (status === 'PASS' && gaps.length === 0) {
+    return;
+  }
+  const message = gaps[0] ?? (runtimeJudgment === 'PASS' ? `Validation status is ${validationStatus}.` : `Workflow gate status is ${runtimeJudgment}.`);
+  await recordRuntimeDurableGap(projectRoot, {
+    gapId: testRuntimeGapId(branch, runId, taskId),
+    partition: branch,
+    taskId,
+    runId,
+    stage: 'test',
+    gate: 'test',
+    source: 'runtime',
+    category: runtimeJudgment === 'PASS' ? 'validation' : 'workflow_gate',
+    severity: 'blocking',
+    status: 'open',
+    message,
+    recommendation: `Resolve test runtime gaps for ${taskId}, then rerun sdd test task ${taskId} --branch ${branch}.`,
+    evidenceRefs,
+    proposalRefs: [],
+    sourceRefs: [],
+    payload: { status, validationStatus, runtimeJudgment, gaps }
+  });
+}
+
+function testRuntimeGapId(branch: string, _runId: string, taskId: string): string {
+  return runtimeScopedId(branch, taskId, 'test-runtime-gap');
+}
+
+function finalStatusForTest(validationStatus: SddTestStatus, runtimeJudgment: WorkflowGateStatus): SddTestStatus {
+  if (validationStatus !== 'PASS') {
+    return validationStatus;
+  }
+  return runtimeJudgment === 'PASS' ? 'PASS' : 'BLOCKED';
+}
+
+function buildUnifiedTestEvidenceRun(id: string, branch: string, runId: string, taskId: string, commandStatus: TestEvidenceStatus, evidenceCoverage: EvidenceCoverageStatus, policyJudgment: TestEvidenceStatus, status: SddTestStatus, runtimeJudgment: WorkflowGateStatus, steps: SddTestCommandStep[], acceptanceCoverage: AcceptanceEvidenceCoverage[], capabilityEvidence: CapabilityEvidenceClassification[], syncBackReady: boolean, gaps: string[], gateNextAction: string | null, gateDecision: WorkflowGateDecision | null): UnifiedTestEvidenceRun {
+  return {
+    contract: TEST_EVIDENCE_RUN_CONTRACT_VERSION,
+    id,
+    scope: { branch, taskId, runId },
+    commandStatus,
+    evidenceCoverage,
+    policyJudgment,
+    commands: steps.map((step) => ({
+      command: step.command,
+      status: step.status === 'pass' ? 'PASS' : step.status === 'fail' ? 'FAIL' : 'BLOCKED',
+      outputRef: { kind: 'artifact', ref: step.outputArtifact },
+      evidenceRefs: [{ kind: 'artifact', ref: step.outputArtifact }],
+      acceptanceRefs: step.acceptanceRefs,
+      startedAt: new Date(Date.now() - step.durationMs).toISOString(),
+      completedAt: new Date().toISOString()
+    })),
+    acceptanceCoverage,
+    capabilityEvidence,
+    syncBackReady,
+    gaps: [...gaps, ...acceptanceCoverage.flatMap((coverage) => coverage.gaps)],
+    next: nextForTestResult(status, runtimeJudgment, branch, taskId, runId, `artifacts/test-index-${taskId}.json`, gateNextAction, gateDecision),
+    generatedAt: new Date().toISOString()
+  };
+}
+
+function buildCapabilityEvidenceClassification(decision: AgentCapabilityRouteDecision | null, steps: SddTestCommandStep[]): CapabilityEvidenceClassification[] {
+  if (!decision) {
+    return [{
+      class: 'diagnostic',
+      source: 'runtime_diagnostic',
+      domainOrSourceId: 'capability-routing',
+      evidenceRefs: [],
+      acceptanceRefs: [],
+      provenanceRefs: [],
+      reason: 'Capability routing did not run; no capability output is accepted as test evidence.'
+    }];
+  }
+  const acceptanceRefs = [...new Set(steps.flatMap((step) => step.acceptanceRefs))];
+  const professionalEvidence = decision.selectedDomains.map((domain) => ({
+    class: 'candidate' as const,
+    source: 'professional_capability' as const,
+    domainOrSourceId: domain.domain,
+    evidenceRefs: [],
+    acceptanceRefs,
+    provenanceRefs: [{ kind: 'projection' as const, ref: `capability:${domain.capabilityId}` }],
+    reason: `${domain.reason}; capability output is advisory candidate evidence until accepted by command evidence and policy refs.`
+  }));
+  const externalEvidence = decision.rejectedExternalSources.map((source) => ({
+    class: capabilityClassForRejectedSource(source.quarantineStatus) as CapabilityEvidenceClassification['class'],
+    source: 'external_source' as const,
+    domainOrSourceId: source.sourceId,
+    evidenceRefs: [],
+    acceptanceRefs: [],
+    provenanceRefs: [{ kind: 'external' as const, ref: source.sourceId }],
+    reason: source.reason
+  }));
+  return [...professionalEvidence, ...externalEvidence];
+}
+
+function capabilityClassForRejectedSource(status: AgentCapabilityRouteDecision['rejectedExternalSources'][number]['quarantineStatus']): CapabilityEvidenceClassification['class'] {
+  if (status === 'denied') {
+    return 'blocked';
+  }
+  if (status === 'required' || status === 'quarantined') {
+    return 'quarantined';
+  }
+  return 'diagnostic';
+}
+
+function renderCapabilityEvidenceClassification(items: CapabilityEvidenceClassification[]): string {
+  if (items.length === 0) {
+    return '- none';
+  }
+  return items.map((item) => `- [${item.class}] ${item.source}:${item.domainOrSourceId} — ${item.reason}`).join('\n');
+}
+
+function capabilityEvidenceSummary(items: CapabilityEvidenceClassification[]): string {
+  if (items.length === 0) {
+    return 'none';
+  }
+  const counts = new Map<CapabilityEvidenceClassification['class'], number>();
+  for (const item of items) {
+    counts.set(item.class, (counts.get(item.class) ?? 0) + 1);
+  }
+  return [...counts.entries()].map(([kind, count]) => `${kind}:${count}`).join(',');
+}
+
+function nextForTestResult(status: SddTestStatus, runtimeJudgment: WorkflowGateStatus, branch: string, taskId: string, _runId: string, indexArtifact: string, gateNextAction: string | null, gateDecision: WorkflowGateDecision | null): string {
+  if (status === 'PASS') {
+    return `sdd ship --branch ${branch} --dry-run`;
+  }
+  if (gateNextAction) {
+    return gateNextAction;
+  }
+  if (runtimeJudgment === 'HUMAN_REQUIRED') {
+    return gateDecision ? `Create a decision card for workflow gate ${gateDecision.decisionId}, then rerun sdd test task ${taskId} --branch ${branch}.` : `Create a decision card, then rerun sdd test task ${taskId} --branch ${branch}.`;
+  }
+  if (runtimeJudgment === 'WARN') {
+    return gateDecision ? `Review workflow gate ${gateDecision.decisionId} warnings, then rerun sdd test task ${taskId} --branch ${branch} or proceed only with explicit review.` : `Review workflow gate warnings, then rerun sdd test task ${taskId} --branch ${branch}.`;
+  }
+  if (runtimeJudgment === 'ADVISORY_ONLY') {
+    return `Inspect advisor assessments for ${taskId}; advisory output cannot satisfy the test gate.`;
+  }
+  if (status === 'FAIL') {
+    return `Inspect ${indexArtifact}, fix failing validation commands, then rerun sdd test task ${taskId} --branch ${branch}.`;
+  }
+  return gateDecision ? `Inspect ${indexArtifact} and workflow gate ${gateDecision.decisionId}, resolve blockers, then rerun sdd test task ${taskId} --branch ${branch}.` : `Inspect ${indexArtifact}, fix command/evidence gaps, then rerun sdd test task ${taskId} --branch ${branch}.`;
+}
+
+function resultSentenceForTest(result: SddTestResult): string {
+  if (result.status === 'PASS') {
+    return 'Validation and workflow gate passed; proceed to release readiness.';
+  }
+  if (result.validationStatus === 'PASS' && result.workflowGateStatus !== 'PASS') {
+    return `Validation passed, but workflow gate returned ${result.workflowGateStatus}.`;
+  }
+  if (result.commandStatus === 'BLOCKED') {
+    return 'Blocked before validation commands ran.';
+  }
+  return result.status === 'FAIL' ? 'Validation failed.' : 'Validation did not produce complete evidence.';
+}
+
+function renderEvidenceBlocks(task: SddTask, status: SddTestStatus, sourceArtifact: string, steps: SddTestCommandStep[]): string {
+  const acceptances = task.acceptanceRefs.length > 0 ? task.acceptanceRefs : task.acceptance;
+  if (acceptances.length === 0) {
+    return 'No acceptance targets declared.';
+  }
+  const mappedEvidence = acceptances
+    .map((acceptance) => ({ acceptance, steps: steps.filter((step) => step.acceptanceRefs.includes(acceptance)) }))
+    .filter((item) => item.steps.length > 0);
+  if (mappedEvidence.length === 0) {
+    return 'No acceptance evidence emitted; validation commands are not explicitly mapped to acceptance refs.';
+  }
+  return mappedEvidence.map(({ acceptance, steps: mappedSteps }) => {
+    const evidenceStatus = evidenceStatusForMappedSteps(status, mappedSteps);
+    return `\`\`\`sdd-evidence\ncontract: ${SDD_EVIDENCE_CONTRACT}\nversion: ${SDD_EVIDENCE_VERSION}\ntask: ${task.id}\nacceptance: ${acceptance}\nstatus: ${evidenceStatus}\nclaim: Explicit validation mapping ${mappedSteps.map((step) => step.command).join(' && ')} produced ${evidenceStatus} for ${acceptance}.\nsource_artifact: ${sourceArtifact}\nevidence_refs:\n${mappedSteps.map((step) => `  - command:${step.command}\n  - artifact:${step.outputArtifact}`).join('\n')}\nprovenance_refs:\n  - artifact:${sourceArtifact}\n${mappedSteps.map((step) => `  - command:${step.command}`).join('\n')}\npolicy_refs:\n  - ${ACCEPTANCE_POLICY_RULESET_VERSION}:require-source-evidence\n  - ${ACCEPTANCE_POLICY_RULESET_VERSION}:require-provenance\n  - ${ACCEPTANCE_POLICY_RULESET_VERSION}:require-policy-rule\n\`\`\``;
+  }).join('\n\n');
+}
+
+function normalizeTestCommandInputs(commandInputs: SddTestCommandInput[] | undefined, commands: string[] | undefined, taskValidation: string[]): NormalizedSddTestCommand[] {
+  if (commandInputs && commandInputs.length > 0) {
+    return commandInputs.map(normalizeTestCommandInput);
+  }
+  return (commands && commands.length > 0 ? commands : taskValidation).map((command) => ({ command, argv: null, shell: true }));
+}
+
+function normalizeTestCommandInput(input: SddTestCommandInput): NormalizedSddTestCommand {
+  if (input.argv) {
+    const argv = input.argv.filter((item) => item.length > 0);
+    if (argv.length === 0) {
+      throw new Error('Command argv input must include an executable.');
+    }
+    return { command: argv.join(' '), argv, shell: false };
+  }
+  if (input.command) {
+    return { command: input.command, argv: null, shell: true };
+  }
+  throw new Error('Command input must include command or argv.');
+}
+
+function acceptanceRefsForCommand(task: SddTask | null, command: string, acceptanceRefsOverride: string[] | undefined): string[] {
+  if (acceptanceRefsOverride && acceptanceRefsOverride.length > 0) {
+    return [...new Set(acceptanceRefsOverride)];
+  }
+  return [...new Set((task?.validationCommands ?? [])
+    .filter((entry) => entry.command === command)
+    .flatMap((entry) => entry.acceptanceRefs))];
+}
+
+function taskAcceptanceRefs(task: SddTask): string[] {
+  const refs = task.acceptanceRefs.length > 0 ? task.acceptanceRefs : task.acceptance;
+  return [...new Set(refs)];
+}
+
+function evidenceStatusForMappedSteps(status: SddTestStatus, steps: SddTestCommandStep[]): SddTestStatus {
+  if (steps.some((step) => step.status === 'fail')) {
+    return 'FAIL';
+  }
+  if (status === 'BLOCKED' || steps.some((step) => step.status === 'blocked')) {
+    return 'BLOCKED';
+  }
+  return 'PASS';
+}
+
+function renderCommandOutput(commandInput: NormalizedSddTestCommand, status: SddTestStepStatus, executed: { exitCode: number | null; signal: string | null; stdout: string; stderr: string; truncated: boolean; timedOut: boolean; error: string | null }, durationMs: number): string {
+  return `# Test Command Output\n\n- command: ${commandInput.command}\n- shell: ${commandInput.shell}\n- argv: ${commandInput.argv ? JSON.stringify(commandInput.argv) : 'none'}\n- status: ${status}\n- exit_code: ${executed.exitCode ?? 'none'}\n- signal: ${executed.signal ?? 'none'}\n- duration_ms: ${durationMs}\n- timed_out: ${executed.timedOut}\n- truncated: ${executed.truncated}\n- error: ${executed.error ?? 'none'}\n\n## stdout\n\n\`\`\`text\n${executed.stdout}\n\`\`\`\n\n## stderr\n\n\`\`\`text\n${executed.stderr}\n\`\`\`\n`;
+}
+
+function hashDocumentContent(raw: string): string {
+  return createHash('sha256').update(raw.replace(/\r\n/g, '\n'), 'utf8').digest('hex');
+}