sdd-agent-platform 0.4.0 → 0.4.2

package/node_modules/@sdd-agent-platform/core/dist/verification/test-runtime.js

@@ -1,32 +1,36 @@
 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 { createRun, readRunState, writeRunState } from '../run-state/run-state.js';
-import { writeArtifact } from '../run-state/artifacts.js';
+import { appendArtifactHashLedgerEntry, appendInvocationLedgerEntry } from '../run-state/invocation-ledger.js';
+import { createRun, readAllRunStates, readRunState, writeRunState } from '../run-state/run-state.js';
+import { recordRuntimeOnlyArtifact, recordStageEvidenceArtifact } from '../run-state/artifacts.js';
+import { toBranchStageEvidenceRef } from '../runtime-paths.js';
 import { resolveSddContext } from '../sdd-docs/context.js';
 import { bindRunStateToTask } from '../sdd-docs/run-binding.js';
 import { parseSddBranch } from '../sdd-docs/task-parser.js';
-import { recordRuntimeProjection, recordRuntimeTestRun, recordRuntimeTestStep, runtimeScopedId } from '../storage/runtime-store.js';
+import { inspectSddTask } from '../sdd-docs/task-inspection.js';
+import { listRuntimeArtifactPayloads, readRuntimeValidationCacheEntry, recordRuntimeAcceptanceEvidenceMap, recordRuntimeDurableGap, recordRuntimeProjection, recordRuntimeTestRun, recordRuntimeTestStep, recordRuntimeValidationCacheEntry, recordRuntimeValidationCacheUse, 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 { inspectVerifyContract, writeVerifyContract } from './verify-contract.js';
+import { inspectVerifyContract } from './verify-contract.js';
 import { ensureTaskOrchestration, inspectOrchestrationGate } from '../orchestration/runtime.js';
 import { recordStageRunProjection, recordWorkflowHandoffProjection, validateWorkflowHandoff } from '../stage-runtime/runtime.js';
 import { evaluateTaskWorkflowGate, verifyContractBlockedGate } from '../risk.js';
 import { validateSddResultArtifact } from '../artifacts/sdd-result.js';
 import { dependencyBlockingReasonsForTask } from '../workflow-state/dependencies.js';
+import { selectLatestEligibleRunsByTask } from '../workflow-state/latest-eligible-run.js';
+import { latestRuntimeTaskStates } from '../workflow-state/resolve.js';
+import { routeSddTask } from '../router/route-sdd-task.js';
+import { evaluateAndRecordWorkflowGateDecision } from '../workflow-gate/evidence-packet.js';
+import { buildValidationCachePlan } from './validation-cache.js';
 const DEFAULT_TEST_TIMEOUT_MS = 120_000;
 const MAX_CAPTURE_BYTES = 256 * 1024;
 export async function runSddTest(projectRoot, options) {
     const context = await resolveSddContext(projectRoot, { branch: options.branch ?? undefined, branchSource: options.branch ? 'cli_option' : undefined });
-    let model = await parseSddBranch(projectRoot, context.partition);
-    let task = model.tasks.find((candidate) => candidate.id === options.taskId) ?? null;
+    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;
-    if (verifyContract.action === 'created' || verifyContract.action === 'refreshed') {
-        model = await parseSddBranch(projectRoot, context.partition);
-        task = model.tasks.find((candidate) => candidate.id === options.taskId) ?? null;
-    }
     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');
@@ -34,17 +38,51 @@ export async function runSddTest(projectRoot, options) {
     const commands = commandInputs.map((input) => input.command);
     const gaps = [];
     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];
+    const validationWaveScopeAcceptanceRefs = [...new Set(Object.values(options.validationWave?.acceptanceRefsByTask ?? { [options.taskId]: task?.acceptanceRefs ?? [] }).flat())].sort();
+    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 }
     });
+    const states = await readAllRunStates(projectRoot);
+    const latestEligibleRunsByTask = selectLatestEligibleRunsByTask({ states, model, partition: context.partition, currentGitBranch: context.currentGitBranch });
+    const runtimeByTask = latestRuntimeTaskStates(latestEligibleRunsByTask, states);
     if (!task) {
         gaps.push(`Task ${options.taskId} was not found in specs/${context.partition}/tasks.md.`);
     }
     if (task) {
-        gaps.push(...dependencyBlockingReasonsForTask(model, options.taskId));
+        gaps.push(...inspected.gaps.filter((gap) => gap.severity === 'blocking').map((gap) => `${gap.field}: ${gap.message}`));
+        gaps.push(...dependencyBlockingReasonsForTask(model, options.taskId, { runtimeByTask }));
     }
     if (verifyContract.action === 'blocked') {
         gaps.push(verifyContractBlocker(verifyInspection));
@@ -91,7 +129,7 @@ export async function runSddTest(projectRoot, options) {
     const steps = [];
     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), options.timeoutMs ?? DEFAULT_TEST_TIMEOUT_MS);
+            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, model, task, validationWaveTaskIds, validationWaveScopeAcceptanceRefs);
             steps.push(step);
             await appendEvent(projectRoot, state.runId, {
                 event: 'test_step_completed',
@@ -102,40 +140,29 @@ export async function runSddTest(projectRoot, options) {
         }
     }
     const commandStatus = deriveCommandStatus(gaps, steps);
-    const acceptanceCoverage = buildAcceptanceCoverage(task, steps, commandStatus);
+    const acceptanceCoverage = buildAcceptanceCoverage(task, steps, commandStatus, validationWaveAcceptanceRefs);
     const evidenceCoverage = summarizeEvidenceCoverage(acceptanceCoverage);
     const policyJudgment = derivePolicyJudgment(commandStatus, evidenceCoverage);
-    const status = policyJudgment;
-    const syncBackReady = status === 'PASS';
-    const validationArtifact = task ? await writeValidationArtifact(projectRoot, state.runId, task, status, steps, gaps) : null;
-    const unifiedEvidence = buildUnifiedTestEvidenceRun(testRunId, context.partition, state.runId, options.taskId, commandStatus, evidenceCoverage, policyJudgment, steps, acceptanceCoverage, syncBackReady, gaps);
-    const indexArtifact = await writeIndexArtifact(projectRoot, state.runId, {
-        testRunId,
+    await recordAcceptanceEvidenceMaps(projectRoot, validationWaveRunId, testRunId, context.partition, state.runId, options.taskId, acceptanceCoverage);
+    const validationStatus = policyJudgment;
+    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, context.partition, task, validationStatus, steps, gaps, capabilityEvidence) : null;
+    const evidenceBeforeGate = runtimeEvidenceRefs(validationArtifact?.runRelativePath ?? null, steps);
+    await persistTestRunState(projectRoot, state, options.taskId, validationStatus, commands, evidenceBeforeGate, 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,
-        status,
-        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,
-        syncBackReady,
-        commands,
-        steps,
-        validationArtifact: validationArtifact?.runRelativePath ?? null,
-        gaps
-    });
+        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, evidenceBeforeGate);
+    const unifiedEvidence = buildUnifiedTestEvidenceRun(testRunId, context.partition, state.runId, options.taskId, commandStatus, evidenceCoverage, policyJudgment, status, runtimeJudgment, steps, acceptanceCoverage, capabilityEvidence, gaps, workflowGate.nextAction, gateDecision);
     const completedAt = new Date().toISOString();
-    const evidence = [validationArtifact?.runRelativePath, indexArtifact.runRelativePath, ...steps.map((step) => step.outputArtifact)].filter((item) => Boolean(item));
-    const syncBackProposal = syncBackReady
-        ? await writeTestSyncBackProposal(projectRoot, state.runId, options.taskId, evidence, acceptanceCoverage)
-        : null;
+    const evidence = runtimeEvidenceRefs(validationArtifact?.runRelativePath ?? null, steps);
     await recordRuntimeTestRun(projectRoot, {
         testRunId,
         runId: state.runId,
@@ -144,7 +171,7 @@ export async function runSddTest(projectRoot, options) {
         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, acceptanceCoverage, syncBackReady, commands, commandInputs, evidence, gaps, syncBackProposal: syncBackProposal?.path ?? null }
+        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, commands, commandInputs, evidence, gaps }
     });
     await recordRuntimeProjection(projectRoot, 'test_runtime', `${context.partition}:${options.taskId}:${state.runId}`, {
         contract: 'sdd-test-runtime-v1',
@@ -152,6 +179,9 @@ export async function runSddTest(projectRoot, options) {
         runId: state.runId,
         taskId: options.taskId,
         status,
+        validationStatus,
+        workflowGateStatus: gateDecision.status,
+        runtimeJudgment,
         lifecycleGate: workflowGate.lifecycleGate,
         primaryReason: workflowGate.primaryReason,
         evidence,
@@ -167,20 +197,50 @@ export async function runSddTest(projectRoot, options) {
         gaps,
         riskDecision: orchestration.riskDecision
     });
-    await persistTestRunState(projectRoot, state, options.taskId, status, commands, evidence, validationArtifact?.runRelativePath ?? null, syncBackProposal);
+    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_failed',
+        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, evidence, gaps }
+        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,
@@ -192,15 +252,34 @@ export async function runSddTest(projectRoot, options) {
         evidenceCoverage,
         policyJudgment,
         acceptanceCoverage,
-        syncBackReady,
+        capabilityEvidence,
         commands,
         steps,
         validationArtifact: validationArtifact?.runRelativePath ?? null,
-        indexArtifact: indexArtifact.runRelativePath,
+        indexArtifact: null,
         gaps,
-        next: nextForTestResult(status, context.partition, options.taskId, state.runId, indexArtifact.runRelativePath, workflowGate.nextAction)
+        next: nextForTestResult(status, runtimeJudgment, context.partition, options.taskId, workflowGate.nextAction, gateDecision)
     };
 }
+async function recordAcceptanceEvidenceMaps(projectRoot, waveRunId, testRunId, partition, runId, taskId, acceptanceCoverage) {
+    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, input) {
     const outputRefs = input.evidence.map((ref) => ({ kind: 'artifact', ref }));
     const completedStage = {
@@ -233,31 +312,25 @@ async function recordTestWorkflowProjection(projectRoot, input) {
     await recordWorkflowHandoffProjection(projectRoot, validation.valid ? handoff : { ...handoff, status: 'blocked', blockingGaps: validation.issues, decidedAt: input.completedAt });
 }
 async function ensureVerifyContractForTest(projectRoot, branch) {
-    let inspection = await inspectVerifyContract(projectRoot, { branch, branchSource: 'cli_option' });
-    if (inspection.status === 'BLOCKED') {
-        return { inspection, action: 'blocked' };
-    }
-    if (inspection.status === 'PASS') {
-        return { inspection, action: 'none' };
-    }
-    const written = await writeVerifyContract(projectRoot, { branch, branchSource: 'cli_option', force: inspection.exists });
-    inspection = await inspectVerifyContract(projectRoot, { branch, branchSource: 'cli_option' });
-    if (inspection.status !== 'PASS') {
-        return { inspection, action: 'blocked' };
-    }
-    return { inspection, action: written.status === 'created' ? 'created' : 'refreshed' };
+    const inspection = await inspectVerifyContract(projectRoot, { branch, branchSource: 'cli_option' });
+    return { inspection, action: inspection.status === 'PASS' ? 'none' : 'blocked' };
 }
 function verifyContractBlocker(inspection) {
     const issueSummary = inspection.issues.map((issue) => `${issue.field}: ${issue.message}`).join(' ');
     return `verify.md contract is ${inspection.status}; ${issueSummary || 'inspect verify.md before executing tests.'}`;
 }
 async function hasReviewerCheckpoint(projectRoot, state, taskId) {
-    const artifactPaths = new Set([
-        ...state.artifacts
-            .filter((artifact) => artifact.task === taskId && (artifact.agent === 'reviewer' || artifact.kind === 'review'))
-            .map((artifact) => artifact.path),
-        `artifacts/review-${taskId}.md`
-    ]);
+    const artifactPaths = new Set(state.artifacts
+        .filter((artifact) => artifact.task === taskId && (artifact.agent === 'reviewer' || artifact.kind === 'review'))
+        .map((artifact) => artifact.path));
+    const branch = state.partition ?? state.gitBranch ?? 'unscoped';
+    const payloads = await listRuntimeArtifactPayloads(projectRoot, { runId: state.runId, taskId });
+    for (const payload of payloads) {
+        const fileName = payload.logicalRef.replace(/\\/g, '/').split('/').filter(Boolean).pop();
+        if (fileName && (payload.artifactRole === 'review' || /review/i.test(fileName))) {
+            artifactPaths.add(toBranchStageEvidenceRef(branch, 'do', fileName));
+        }
+    }
     for (const artifactPath of artifactPaths) {
         const report = await validateSddResultArtifact(projectRoot, state.runId, artifactPath, { expectedTask: taskId, expectedAgent: 'reviewer' });
         if (report.valid && report.result?.status === 'PASS') {
@@ -272,21 +345,52 @@ export function renderSddTestResult(result) {
         '',
         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:',
         `- ${result.next}`
     ].join('\n');
 }
-async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, timeoutMs) {
+async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, timeoutMs, model, task, validationScopeTaskIds, validationScopeAcceptanceRefs) {
+    const cachePlan = buildValidationCachePlan({ branch, model, task, command: commandInput.command, argv: commandInput.argv, shell: commandInput.shell, validationScopeTaskIds, acceptanceRefs: validationScopeAcceptanceRefs });
+    const cached = cachePlan.eligible ? await readRuntimeValidationCacheEntry(projectRoot, { branchSlug: branch, cacheKey: cachePlan.cacheKey }) : null;
+    if (cached) {
+        return recordCachedCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, cachePlan.cacheKey, cached);
+    }
     const started = Date.now();
+    const startedAt = new Date(started).toISOString();
     const executed = await executeCommand(projectRoot, commandInput, timeoutMs);
-    const durationMs = Date.now() - started;
+    const ended = Date.now();
+    const endedAt = new Date(ended).toISOString();
+    const durationMs = ended - started;
     const status = executed.timedOut || executed.error ? 'blocked' : executed.exitCode === 0 ? 'pass' : 'fail';
     const stepId = `${testRunId}-${String(sequence).padStart(3, '0')}`;
-    const output = renderCommandOutput(commandInput, status, executed, durationMs);
-    const outputArtifact = await writeArtifact(projectRoot, runId, `test-${taskId}-${String(sequence).padStart(3, '0')}.log`, output);
+    const cacheStatus = cachePlan.eligible ? 'miss' : 'unsafe';
+    const shouldPersistOutputArtifact = shouldPersistCommandOutputArtifact(status, executed, cacheStatus, cachePlan.unsafeReasons);
+    const outputFileName = shouldPersistOutputArtifact ? `test-${taskId}-${String(sequence).padStart(3, '0')}.log` : null;
+    const outputRef = outputFileName ? toBranchStageEvidenceRef(branch, 'test', outputFileName) : null;
+    const output = outputFileName ? renderCommandOutput(commandInput, status, executed, durationMs, cacheStatus, cachePlan.cacheKey, null, cachePlan.unsafeReasons) : null;
+    if (outputFileName && outputRef && output) {
+        await recordRuntimeOnlyArtifact(projectRoot, runId, outputFileName, output, { logicalRef: outputRef, branch, taskId, artifactRole: 'test-command-output' });
+        await appendArtifactHashLedgerEntry(projectRoot, {
+            runId,
+            taskId,
+            branch,
+            artifactPath: outputRef,
+            content: output,
+            status: 'recorded'
+        });
+    }
+    const stdoutDigest = hashDocumentContent(executed.stdout);
+    const stderrDigest = hashDocumentContent(executed.stderr);
+    const outputSummary = summarizeCommandOutput(executed);
     await appendInvocationLedgerEntry(projectRoot, {
         runId,
         taskId,
@@ -294,10 +398,11 @@ async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, seq
         kind: 'command',
         ref: commandInput.command,
         status,
-        artifactPath: outputArtifact.runRelativePath,
-        outputHash: hashDocumentContent(output),
-        materialRefs: [outputArtifact.runRelativePath],
+        artifactPath: outputRef,
+        outputHash: output ? hashDocumentContent(output) : null,
+        materialRefs: outputRef ? [outputRef] : [],
         metadata: {
+            stepId,
             source: 'sdd-test',
             exitCode: executed.exitCode,
             durationMs,
@@ -306,7 +411,12 @@ async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, seq
             truncated: executed.truncated,
             acceptanceRefs: acceptanceRefs.join(','),
             shell: commandInput.shell,
-            argv: commandInput.argv ? JSON.stringify(commandInput.argv) : null
+            argv: commandInput.argv ? JSON.stringify(commandInput.argv) : null,
+            stdoutDigest,
+            stderrDigest,
+            cacheKey: cachePlan.cacheKey,
+            cacheStatus,
+            cacheUnsafeReasons: cachePlan.unsafeReasons.join(',')
         }
     });
     const step = {
@@ -319,10 +429,20 @@ async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, seq
         exitCode: executed.exitCode,
         signal: executed.signal,
         durationMs,
-        outputArtifact: outputArtifact.runRelativePath,
+        outputArtifact: outputRef,
         stdoutBytes: executed.stdoutBytes,
         stderrBytes: executed.stderrBytes,
-        truncated: executed.truncated
+        truncated: executed.truncated,
+        startedAt,
+        endedAt,
+        cwd: projectRoot,
+        stdoutDigest,
+        stderrDigest,
+        outputSummary,
+        cacheStatus,
+        cacheKey: cachePlan.cacheKey,
+        cacheSourceTestRunId: null,
+        cacheUnsafeReasons: cachePlan.unsafeReasons
     };
     await recordRuntimeTestStep(projectRoot, {
         stepId,
@@ -333,9 +453,95 @@ async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, seq
         status,
         exitCode: executed.exitCode,
         durationMs,
-        outputArtifact: outputArtifact.runRelativePath,
+        outputArtifact: outputRef,
+        payload: step
+    });
+    if (status === 'pass' && cachePlan.eligible) {
+        const now = new Date().toISOString();
+        await recordRuntimeValidationCacheEntry(projectRoot, {
+            cacheKey: cachePlan.cacheKey,
+            branchSlug: branch,
+            command: commandInput.command,
+            status: 'valid',
+            sourceTestRunId: testRunId,
+            sourceRunId: runId,
+            sourceEvidenceSetId: null,
+            outputArtifact: outputRef,
+            stdoutDigest,
+            stderrDigest,
+            createdAt: now,
+            lastUsedAt: now,
+            payload: { cachePlan, step }
+        });
+    }
+    return step;
+}
+async function recordCachedCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, cacheKey, cached) {
+    const now = new Date().toISOString();
+    const stepId = `${testRunId}-${String(sequence).padStart(3, '0')}`;
+    const step = {
+        stepId,
+        command: commandInput.command,
+        argv: commandInput.argv,
+        shell: commandInput.shell,
+        acceptanceRefs,
+        status: 'pass',
+        exitCode: 0,
+        signal: null,
+        durationMs: 0,
+        outputArtifact: null,
+        stdoutBytes: 0,
+        stderrBytes: 0,
+        truncated: false,
+        startedAt: now,
+        endedAt: now,
+        cwd: projectRoot,
+        stdoutDigest: cached.stdoutDigest,
+        stderrDigest: cached.stderrDigest,
+        outputSummary: `cache_hit source_test_run=${cached.sourceTestRunId} source_artifact=${cached.outputArtifact ?? 'none'}`,
+        cacheStatus: 'hit',
+        cacheKey,
+        cacheSourceTestRunId: cached.sourceTestRunId,
+        cacheUnsafeReasons: []
+    };
+    await appendInvocationLedgerEntry(projectRoot, {
+        runId,
+        taskId,
+        branch,
+        kind: 'command',
+        ref: commandInput.command,
+        status: 'pass',
+        artifactPath: null,
+        outputHash: null,
+        materialRefs: cached.outputArtifact ? [cached.outputArtifact] : [],
+        metadata: { source: 'sdd-test-cache', stepId, cacheKey, sourceTestRunId: cached.sourceTestRunId, sourceEvidenceSetId: cached.sourceEvidenceSetId, acceptanceRefs: acceptanceRefs.join(',') }
+    });
+    await recordRuntimeTestStep(projectRoot, {
+        stepId,
+        testRunId,
+        runId,
+        taskId,
+        command: commandInput.command,
+        status: 'pass',
+        exitCode: 0,
+        durationMs: 0,
+        outputArtifact: null,
         payload: step
     });
+    await recordRuntimeValidationCacheUse(projectRoot, {
+        useId: runtimeScopedId(cacheKey, testRunId, taskId, stepId),
+        cacheKey,
+        branchSlug: branch,
+        testRunId,
+        runId,
+        taskId,
+        sourceTestRunId: cached.sourceTestRunId,
+        sourceEvidenceSetId: cached.sourceEvidenceSetId,
+        reusedAt: now,
+        mappedTaskIds: [taskId],
+        reason: 'same validation cache key within compatible branch contract scope',
+        payload: { sourceRunId: cached.sourceRunId, sourceArtifact: cached.outputArtifact }
+    });
     return step;
 }
 function executeCommand(projectRoot, commandInput, timeoutMs) {
@@ -386,28 +592,57 @@ function executeCommand(projectRoot, commandInput, timeoutMs) {
         child.on('close', (code, signal) => finish({ exitCode: code, signal, error: null }));
     });
 }
-async function writeValidationArtifact(projectRoot, runId, task, status, steps, gaps) {
-    const artifactPath = `validation-${task.id}.md`;
-    const runRelativePath = `artifacts/${artifactPath}`;
+async function writeValidationArtifact(projectRoot, runId, branch, task, status, steps, gaps, _capabilityEvidence) {
+    const artifactPath = validationArtifactPath(task);
+    const stageEvidenceRef = toBranchStageEvidenceRef(branch, 'test', artifactPath);
     const resultStatus = status === 'PASS' ? 'PASS' : status === 'FAIL' ? 'FAIL' : 'BLOCKED';
-    const content = `# 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) => `  - ${gap}`).join('\n') : '  - none'}\n\n## Acceptance Evidence\n\n${renderEvidenceBlocks(task, status, runRelativePath, steps)}\n`;
-    return writeArtifact(projectRoot, runId, artifactPath, content);
-}
-async function writeTestSyncBackProposal(projectRoot, runId, taskId, artifacts, acceptanceCoverage) {
-    const content = `# Sync-back Proposal\n\n## ${taskId}\n\n- status: verified\n- summary: /sdd:test passed with complete command and acceptance evidence.\n- artifacts:\n${artifacts.length > 0 ? artifacts.map((artifact) => `  - ${artifact}`).join('\n') : '  - none'}\n- acceptance_coverage:\n${acceptanceCoverage.length > 0 ? acceptanceCoverage.map((coverage) => `  - [${coverage.status}] ${coverage.acceptanceRef}`).join('\n') : '  - none'}\n- gaps:\n  - none\n\n## Boundaries\n\n- Proposal only; tasks.md/spec.md/plan.md were not modified by runtime.\n- /sdd:test executed declared validation commands and mapped evidence to acceptance refs.\n`;
-    const written = await writeArtifact(projectRoot, runId, 'sync-back-proposal.md', content);
-    return { path: written.runRelativePath, digest: hashDocumentContent(content) };
+    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  - ${stageEvidenceRef}\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) => `  - ${gap}`).join('\n') : '  - none'}\n\n## Acceptance Evidence\n\n${renderEvidenceBlocks(task, status, stageEvidenceRef, steps)}\n`;
+    const written = await recordStageEvidenceArtifact(projectRoot, runId, stageEvidenceRef, content, { taskId: task.id, artifactRole: 'test-validation' });
+    return { ...written, runRelativePath: stageEvidenceRef };
 }
-async function writeIndexArtifact(projectRoot, runId, payload) {
-    return writeArtifact(projectRoot, runId, `test-index-${payload.taskId}.json`, `${JSON.stringify({ contract: 'sdd-test-runtime-v1', runId, ...payload }, null, 2)}\n`);
+function validationArtifactPath(task) {
+    return `test-validation-${task.id}.md`;
 }
-async function persistTestRunState(projectRoot, state, taskId, status, commands, evidence, validationArtifact, syncBackProposal) {
+async function persistTestRunState(projectRoot, state, taskId, validationStatus, commands, evidence, validationArtifact) {
     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: artifactPath.includes('validation-') ? 'validation' : 'test', task: taskId, agent: artifactPath === validationArtifact ? 'validator' : 'test-runtime', createdAt: now }));
+        .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 ?? 'not_started',
+                verificationStatus: verificationStatusFromTest(validationStatus),
+                testStatus: validationStatus,
+                evidence
+            }
+        },
+        artifacts: [...latest.artifacts, ...nextArtifacts],
+        validation: {
+            status: validationStatus === 'PASS' ? 'pass' : validationStatus === 'FAIL' ? 'fail' : 'blocked',
+            commands,
+            evidence
+        }
+    });
+}
+async function persistTestGateOutcome(projectRoot, runId, taskId, status, validationStatus, commands, evidence, validationArtifact, gateDecision) {
+    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',
@@ -416,28 +651,63 @@ async function persistTestRunState(projectRoot, state, taskId, status, commands,
         tasks: {
             ...latest.tasks,
             [taskId]: {
-                status: status === 'PASS' ? 'tested' : 'blocked',
+                ...baseRuntimeTaskState(existingTaskState),
+                status: runtimeTaskStatusAfterGate(status, validationStatus),
+                implementationStatus: existingTaskState?.implementationStatus ?? 'not_started',
+                verificationStatus: verificationStatusFromTest(validationStatus),
                 testStatus: status,
+                validationStatus,
+                workflowGateStatus: gateDecision.status,
+                workflowGateDecisionId: gateDecision.decisionId,
                 evidence
             }
         },
         artifacts: [...latest.artifacts, ...nextArtifacts],
         validation: {
-            status: status === 'PASS' ? 'pass' : status === 'FAIL' ? 'fail' : 'blocked',
+            status: validationStatus === 'PASS' ? 'pass' : validationStatus === 'FAIL' ? 'fail' : 'blocked',
             commands,
             evidence
-        },
-        syncBack: syncBackProposal
-            ? {
-                mode: 'proposal',
-                proposalPath: syncBackProposal.path,
-                proposalDigest: syncBackProposal.digest,
-                sourceVerifyStatus: status,
-                status: 'proposed'
-            }
-            : latest.syncBack
+        }
     });
 }
+function testArtifactKind(artifactPath) {
+    const fileName = artifactPath.split('/').pop() ?? artifactPath;
+    if (fileName.startsWith('test-validation-')) {
+        return 'test-validation';
+    }
+    return 'test';
+}
+function baseRuntimeTaskState(existing) {
+    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) {
+    if (validationStatus === 'PASS') {
+        return 'pass';
+    }
+    return validationStatus === 'FAIL' ? 'failed' : 'blocked';
+}
+function runtimeTaskStatusAfterGate(status, validationStatus) {
+    if (status === 'PASS') {
+        return 'implemented_verified';
+    }
+    if (validationStatus === 'PASS') {
+        return 'workflow_gate_blocked';
+    }
+    return validationStatus === 'FAIL' ? 'validation_failed' : 'validation_blocked';
+}
 function deriveCommandStatus(gaps, steps) {
     if (gaps.length > 0 || steps.some((step) => step.status === 'blocked')) {
         return 'BLOCKED';
@@ -447,8 +717,8 @@ function deriveCommandStatus(gaps, steps) {
     }
     return 'PASS';
 }
-function buildAcceptanceCoverage(task, steps, commandStatus) {
-    const acceptanceRefs = task ? taskAcceptanceRefs(task) : [];
+function buildAcceptanceCoverage(task, steps, commandStatus, acceptanceRefsOverride) {
+    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');
@@ -461,11 +731,20 @@ function buildAcceptanceCoverage(task, steps, commandStatus) {
         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.`]
+            evidenceRefs: mappedSteps.map(commandStepRuntimeRef),
+            gaps: status === 'complete' ? [] : [`Acceptance ${acceptanceRef} has no complete non-stale evidence from sdd test task.`]
         };
     });
 }
+function runtimeEvidenceRefs(validationArtifact, steps) {
+    return [validationArtifact, ...steps.map((step) => step.outputArtifact)].filter((item) => Boolean(item));
+}
+function commandStepRuntimeRef(step) {
+    return { kind: 'command', ref: step.stepId };
+}
+function artifactRuntimeRefs(ref) {
+    return ref ? [{ kind: 'artifact', ref }] : [];
+}
 function summarizeEvidenceCoverage(acceptanceCoverage) {
     if (acceptanceCoverage.length === 0) {
         return 'missing';
@@ -487,7 +766,51 @@ function derivePolicyJudgment(commandStatus, evidenceCoverage) {
     }
     return 'PASS';
 }
-function buildUnifiedTestEvidenceRun(id, branch, runId, taskId, commandStatus, evidenceCoverage, policyJudgment, steps, acceptanceCoverage, syncBackReady, gaps) {
+async function resolveTestRuntimeDurableGap(projectRoot, branch, runId, taskId, validationStatus, gaps) {
+    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, branch, runId, taskId, status, validationStatus, runtimeJudgment, gaps, evidenceRefs) {
+    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 ${testValidationUnitCommand(branch, taskId)}.`,
+        evidenceRefs,
+        proposalRefs: [],
+        sourceRefs: [],
+        payload: { status, validationStatus, runtimeJudgment, gaps }
+    });
+}
+function testRuntimeGapId(branch, _runId, taskId) {
+    return runtimeScopedId(branch, taskId, 'test-runtime-gap');
+}
+function finalStatusForTest(validationStatus, runtimeJudgment) {
+    if (validationStatus !== 'PASS') {
+        return validationStatus;
+    }
+    return runtimeJudgment === 'PASS' || runtimeJudgment === 'WARN' ? 'PASS' : 'BLOCKED';
+}
+function buildUnifiedTestEvidenceRun(id, branch, runId, taskId, commandStatus, evidenceCoverage, policyJudgment, status, runtimeJudgment, steps, acceptanceCoverage, capabilityEvidence, gaps, gateNextAction, gateDecision) {
     return {
         contract: TEST_EVIDENCE_RUN_CONTRACT_VERSION,
         id,
@@ -498,31 +821,102 @@ function buildUnifiedTestEvidenceRun(id, branch, runId, taskId, commandStatus, e
         commands: steps.map((step) => ({
             command: step.command,
             status: step.status === 'pass' ? 'PASS' : step.status === 'fail' ? 'FAIL' : 'BLOCKED',
-            outputRef: { kind: 'artifact', ref: step.outputArtifact },
-            evidenceRefs: [{ kind: 'artifact', ref: step.outputArtifact }],
+            outputRef: step.outputArtifact ? { kind: 'artifact', ref: step.outputArtifact } : undefined,
+            evidenceRefs: [commandStepRuntimeRef(step), ...artifactRuntimeRefs(step.outputArtifact)],
             acceptanceRefs: step.acceptanceRefs,
             startedAt: new Date(Date.now() - step.durationMs).toISOString(),
             completedAt: new Date().toISOString()
         })),
         acceptanceCoverage,
-        syncBackReady,
+        capabilityEvidence,
         gaps: [...gaps, ...acceptanceCoverage.flatMap((coverage) => coverage.gaps)],
-        next: nextForTestResult(policyJudgment, branch, taskId, runId, `artifacts/test-index-${taskId}.json`, null),
+        next: nextForTestResult(status, runtimeJudgment, branch, taskId, gateNextAction, gateDecision),
         generatedAt: new Date().toISOString()
     };
 }
-function nextForTestResult(status, branch, taskId, runId, indexArtifact, gateNextAction) {
+function buildCapabilityEvidenceClassification(decision, steps) {
+    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',
+        source: 'professional_capability',
+        domainOrSourceId: domain.domain,
+        evidenceRefs: [],
+        acceptanceRefs,
+        provenanceRefs: [{ kind: 'projection', 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),
+        source: 'external_source',
+        domainOrSourceId: source.sourceId,
+        evidenceRefs: [],
+        acceptanceRefs: [],
+        provenanceRefs: [{ kind: 'external', ref: source.sourceId }],
+        reason: source.reason
+    }));
+    return [...professionalEvidence, ...externalEvidence];
+}
+function capabilityClassForRejectedSource(status) {
+    if (status === 'denied') {
+        return 'blocked';
+    }
+    if (status === 'required' || status === 'quarantined') {
+        return 'quarantined';
+    }
+    return 'diagnostic';
+}
+function capabilityEvidenceSummary(items) {
+    if (items.length === 0) {
+        return 'none';
+    }
+    const counts = new Map();
+    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, runtimeJudgment, branch, taskId, gateNextAction, gateDecision) {
+    const inspectHint = `Inspect runtime test read model for ${taskId}`;
     if (status === 'PASS') {
-        return `sdd sync-back inspect ${runId} --branch ${branch} --task ${taskId}`;
+        return `sdd test close --branch ${branch} --target goal-verify --compact-json`;
+    }
+    if (gateNextAction) {
+        return gateNextAction;
+    }
+    if (runtimeJudgment === 'HUMAN_REQUIRED') {
+        return gateDecision ? `Create a decision card for workflow gate ${gateDecision.decisionId}, then rerun ${testValidationUnitCommand(branch, taskId)}.` : `Create a decision card, then rerun ${testValidationUnitCommand(branch, taskId)}.`;
+    }
+    if (runtimeJudgment === 'WARN') {
+        return gateDecision ? `Review workflow gate ${gateDecision.decisionId} warnings, then rerun ${testValidationUnitCommand(branch, taskId)} or proceed only with explicit review.` : `Review workflow gate warnings, then rerun ${testValidationUnitCommand(branch, taskId)}.`;
+    }
+    if (runtimeJudgment === 'ADVISORY_ONLY') {
+        return `Inspect advisor assessments for ${taskId}; advisory output cannot satisfy the test gate.`;
+    }
+    if (status === 'FAIL') {
+        return `${inspectHint}, fix failing validation commands, then rerun ${testValidationUnitCommand(branch, taskId)}.`;
     }
-    return gateNextAction ?? `Inspect ${indexArtifact}, fix command/evidence gaps, then rerun sdd test task ${taskId} --branch ${branch}.`;
+    return gateDecision ? `${inspectHint} and workflow gate ${gateDecision.decisionId}, resolve blockers, then rerun ${testValidationUnitCommand(branch, taskId)}.` : `${inspectHint}, fix command/evidence gaps, then rerun ${testValidationUnitCommand(branch, taskId)}.`;
+}
+function testValidationUnitCommand(branch, taskId) {
+    return `sdd test task-${taskId} --branch ${branch}`;
 }
 function resultSentenceForTest(result) {
-    if (result.status === 'PASS' && result.lifecycleGate === 'direct') {
-        return 'Validation passed and sync-back is ready.';
-    }
     if (result.status === 'PASS') {
-        return 'Validation passed; sync-back needs review.';
+        return 'Validation and workflow gate passed; proceed to goal verification and truthAlignment before 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.';
@@ -542,7 +936,7 @@ function renderEvidenceBlocks(task, status, sourceArtifact, steps) {
     }
     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\`\`\``;
+        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(renderStepEvidenceRefs).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, commands, taskValidation) {
@@ -564,7 +958,10 @@ function normalizeTestCommandInput(input) {
     }
     throw new Error('Command input must include command or argv.');
 }
-function acceptanceRefsForCommand(task, command) {
+function acceptanceRefsForCommand(task, command, acceptanceRefsOverride) {
+    if (acceptanceRefsOverride && acceptanceRefsOverride.length > 0) {
+        return [...new Set(acceptanceRefsOverride)];
+    }
     return [...new Set((task?.validationCommands ?? [])
             .filter((entry) => entry.command === command)
             .flatMap((entry) => entry.acceptanceRefs))];
@@ -582,8 +979,32 @@ function evidenceStatusForMappedSteps(status, steps) {
     }
     return 'PASS';
 }
-function renderCommandOutput(commandInput, status, executed, durationMs) {
-    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 renderCommandOutput(commandInput, status, executed, durationMs, cacheStatus = 'unsafe', cacheKey = null, cacheSourceTestRunId = null, cacheUnsafeReasons = []) {
+    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- cache_status: ${cacheStatus}\n- cache_key: ${cacheKey ?? 'none'}\n- cache_source_test_run: ${cacheSourceTestRunId ?? 'none'}\n- cache_unsafe_reasons: ${cacheUnsafeReasons.join(',') || 'none'}\n\n## stdout\n\n\`\`\`text\n${executed.stdout}\n\`\`\`\n\n## stderr\n\n\`\`\`text\n${executed.stderr}\n\`\`\`\n`;
+}
+function shouldPersistCommandOutputArtifact(status, executed, cacheStatus, cacheUnsafeReasons) {
+    return status !== 'pass' || executed.truncated || executed.timedOut || Boolean(executed.error) || cacheStatus === 'unsafe' || cacheUnsafeReasons.length > 0;
+}
+function renderStepEvidenceRefs(step) {
+    const refs = [`  - command:${step.stepId}`];
+    if (step.outputArtifact) {
+        refs.push(`  - artifact:${step.outputArtifact}`);
+    }
+    return refs.join('\n');
+}
+function summarizeCommandOutput(executed) {
+    const parts = [
+        `exit=${executed.exitCode ?? 'none'}`,
+        `signal=${executed.signal ?? 'none'}`,
+        `stdout_bytes=${executed.stdoutBytes}`,
+        `stderr_bytes=${executed.stderrBytes}`,
+        `truncated=${executed.truncated}`,
+        `timed_out=${executed.timedOut}`
+    ];
+    if (executed.error) {
+        parts.push(`error=${executed.error}`);
+    }
+    return parts.join(' ');
 }
 function hashDocumentContent(raw) {
     return createHash('sha256').update(raw.replace(/\r\n/g, '\n'), 'utf8').digest('hex');