sdd-agent-platform 0.4.1 → 0.5.0

package/packages/core/dist/verification/test-runtime.js

@@ -1,14 +1,15 @@
 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 { inspectSddTask } from '../sdd-docs/task-inspection.js';
-import { recordRuntimeAcceptanceEvidenceMap, recordRuntimeDurableGap, recordRuntimeProjection, recordRuntimeTestRun, recordRuntimeTestStep, recordRuntimeValidationEnvironmentSession, recordRuntimeValidationWaveRun, runtimeScopedId, updateRuntimeDurableGapStatus } from '../storage/runtime-store.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 } from './verify-contract.js';
 import { ensureTaskOrchestration, inspectOrchestrationGate } from '../orchestration/runtime.js';
@@ -16,8 +17,11 @@ import { recordStageRunProjection, recordWorkflowHandoffProjection, validateWork
 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) {
@@ -39,6 +43,7 @@ export async function runSddTest(projectRoot, options) {
     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,
@@ -69,12 +74,15 @@ export async function runSddTest(projectRoot, options) {
         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(...inspected.gaps.filter((gap) => gap.severity === 'blocking').map((gap) => `${gap.field}: ${gap.message}`));
-        gaps.push(...dependencyBlockingReasonsForTask(model, options.taskId));
+        gaps.push(...dependencyBlockingReasonsForTask(model, options.taskId, { runtimeByTask }));
     }
     if (verifyContract.action === 'blocked') {
         gaps.push(verifyContractBlocker(verifyInspection));
@@ -87,14 +95,14 @@ export async function runSddTest(projectRoot, options) {
         runId: state.runId,
         taskId: options.taskId,
         agent: 'validator',
-        stage: 'test',
+        stage: 'execute',
         status: 'active'
     });
     const orchestrationGate = await inspectOrchestrationGate(projectRoot, {
         branch: context.partition,
         runId: state.runId,
         taskId: options.taskId,
-        target: 'test',
+        target: 'execution',
         riskDecision: orchestration.riskDecision,
         stageRun: orchestration.stageRun,
         contextLoadSignal: orchestration.contextLoadSignal,
@@ -121,7 +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, validationWaveAcceptanceRefs), 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',
@@ -137,12 +145,11 @@ export async function runSddTest(projectRoot, options) {
     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) => Boolean(item));
-    await persistTestRunState(projectRoot, state, options.taskId, validationStatus, commands, evidenceBeforeIndex, validationArtifact?.runRelativePath ?? null);
+    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,
@@ -152,39 +159,10 @@ export async function runSddTest(projectRoot, options) {
     })).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
-    });
+    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 evidence = runtimeEvidenceRefs(validationArtifact?.runRelativePath ?? null, steps);
     await recordRuntimeTestRun(projectRoot, {
         testRunId,
         runId: state.runId,
@@ -193,7 +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, validationStatus, workflowGateStatus: gateDecision.status, runtimeJudgment, workflowGateDecision: gateDecision, acceptanceCoverage, capabilityEvidence, syncBackReady, commands, commandInputs, evidence, gaps }
+        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',
@@ -275,13 +253,12 @@ export async function runSddTest(projectRoot, options) {
         policyJudgment,
         acceptanceCoverage,
         capabilityEvidence,
-        syncBackReady,
         commands,
         steps,
         validationArtifact: validationArtifact?.runRelativePath ?? null,
-        indexArtifact: indexArtifact.runRelativePath,
+        indexArtifact: null,
         gaps,
-        next: nextForTestResult(status, runtimeJudgment, context.partition, options.taskId, state.runId, indexArtifact.runRelativePath, workflowGate.nextAction, gateDecision)
+        next: nextForTestResult(status, runtimeJudgment, context.partition, options.taskId, workflowGate.nextAction, gateDecision)
     };
 }
 async function recordAcceptanceEvidenceMaps(projectRoot, waveRunId, testRunId, partition, runId, taskId, acceptanceCoverage) {
@@ -315,12 +292,12 @@ async function recordTestWorkflowProjection(projectRoot, input) {
     await recordStageRunProjection(projectRoot, completedStage);
     const handoff = {
         contract: WORKFLOW_HANDOFF_CONTRACT_VERSION,
-        id: `${completedStage.id}:handoff:goal-verify`,
+        id: `${completedStage.id}:handoff:ship`,
         scope: completedStage.scope,
-        fromStage: 'test',
-        toStage: 'goal-verify',
+        fromStage: 'execute',
+        toStage: 'ship',
         fromAgent: 'validator',
-        toAgent: 'verifier',
+        toAgent: 'ship-manager',
         status: input.status === 'PASS' ? 'proposed' : 'blocked',
         outputRefs,
         requiredInputRefs: [{ kind: 'task', ref: input.taskId }],
@@ -343,12 +320,17 @@ function verifyContractBlocker(inspection) {
     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, 'execute', 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') {
@@ -376,14 +358,39 @@ export function renderSddTestResult(result) {
         `- ${result.next}`
     ].join('\n');
 }
-async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, timeoutMs) {
+async function runCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, timeoutMs, model, task, validationScopeTaskIds, validationScopeAcceptanceRefs) {
+    const cachePlan = buildValidationCachePlan({ branch, model, task, command: commandInput.command, argv: commandInput.argv, shell: commandInput.shell, validationScopeTaskIds, acceptanceRefs: validationScopeAcceptanceRefs });
+    const cached = cachePlan.eligible ? await readRuntimeValidationCacheEntry(projectRoot, { branchSlug: branch, cacheKey: cachePlan.cacheKey }) : null;
+    if (cached) {
+        return recordCachedCommandStep(projectRoot, runId, branch, taskId, testRunId, sequence, commandInput, acceptanceRefs, cachePlan.cacheKey, cached);
+    }
     const started = Date.now();
+    const startedAt = new Date(started).toISOString();
     const executed = await executeCommand(projectRoot, commandInput, timeoutMs);
-    const durationMs = Date.now() - started;
+    const ended = Date.now();
+    const endedAt = new Date(ended).toISOString();
+    const durationMs = ended - started;
     const status = executed.timedOut || executed.error ? 'blocked' : executed.exitCode === 0 ? 'pass' : 'fail';
     const stepId = `${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, 'execute', 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,
@@ -391,10 +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,
@@ -403,7 +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 = {
@@ -416,10 +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,
@@ -430,9 +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) {
@@ -483,15 +592,16 @@ 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, capabilityEvidence) {
-    const artifactPath = `test-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, 'execute', 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);
+    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, validationStatus, commands, evidence, validationArtifact) {
     const latest = await readRunState(projectRoot, state.runId);
@@ -511,7 +621,7 @@ async function persistTestRunState(projectRoot, state, taskId, validationStatus,
             [taskId]: {
                 ...baseRuntimeTaskState(existingTaskState),
                 status: validationStatus === 'PASS' ? 'validation_passed_pending_gate' : validationStatus === 'FAIL' ? 'validation_failed' : 'validation_blocked',
-                implementationStatus: existingTaskState?.implementationStatus ?? 'implemented',
+                implementationStatus: existingTaskState?.implementationStatus ?? 'not_started',
                 verificationStatus: verificationStatusFromTest(validationStatus),
                 testStatus: validationStatus,
                 evidence
@@ -522,8 +632,7 @@ async function persistTestRunState(projectRoot, state, taskId, validationStatus,
             status: validationStatus === 'PASS' ? 'pass' : validationStatus === 'FAIL' ? 'fail' : 'blocked',
             commands,
             evidence
-        },
-        syncBack: latest.syncBack
+        }
     });
 }
 async function persistTestGateOutcome(projectRoot, runId, taskId, status, validationStatus, commands, evidence, validationArtifact, gateDecision) {
@@ -544,7 +653,7 @@ async function persistTestGateOutcome(projectRoot, runId, taskId, status, valida
             [taskId]: {
                 ...baseRuntimeTaskState(existingTaskState),
                 status: runtimeTaskStatusAfterGate(status, validationStatus),
-                implementationStatus: existingTaskState?.implementationStatus ?? 'implemented',
+                implementationStatus: existingTaskState?.implementationStatus ?? 'not_started',
                 verificationStatus: verificationStatusFromTest(validationStatus),
                 testStatus: status,
                 validationStatus,
@@ -558,8 +667,7 @@ async function persistTestGateOutcome(projectRoot, runId, taskId, status, valida
             status: validationStatus === 'PASS' ? 'pass' : validationStatus === 'FAIL' ? 'fail' : 'blocked',
             commands,
             evidence
-        },
-        syncBack: latest.syncBack
+        }
     });
 }
 function testArtifactKind(artifactPath) {
@@ -567,9 +675,6 @@ function testArtifactKind(artifactPath) {
     if (fileName.startsWith('test-validation-')) {
         return 'test-validation';
     }
-    if (fileName.startsWith('test-index-')) {
-        return 'test-index';
-    }
     return 'test';
 }
 function baseRuntimeTaskState(existing) {
@@ -626,11 +731,20 @@ function buildAcceptanceCoverage(task, steps, commandStatus, acceptanceRefsOverr
         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.`]
+            evidenceRefs: mappedSteps.map(commandStepRuntimeRef),
+            gaps: status === 'complete' ? [] : [`Acceptance ${acceptanceRef} has no complete non-stale execute validation evidence.`]
         };
     });
 }
+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';
@@ -673,14 +787,14 @@ async function recordTestRuntimeDurableGap(projectRoot, branch, runId, taskId, s
         partition: branch,
         taskId,
         runId,
-        stage: 'test',
-        gate: 'test',
+        stage: 'execute',
+        gate: 'execute',
         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}.`,
+        recommendation: `Resolve execute runtime gaps for ${taskId}, then rerun ${testValidationUnitCommand(branch)}.`,
         evidenceRefs,
         proposalRefs: [],
         sourceRefs: [],
@@ -694,9 +808,9 @@ function finalStatusForTest(validationStatus, runtimeJudgment) {
     if (validationStatus !== 'PASS') {
         return validationStatus;
     }
-    return runtimeJudgment === 'PASS' ? 'PASS' : 'BLOCKED';
+    return runtimeJudgment === 'PASS' || runtimeJudgment === 'WARN' ? 'PASS' : 'BLOCKED';
 }
-function buildUnifiedTestEvidenceRun(id, branch, runId, taskId, commandStatus, evidenceCoverage, policyJudgment, status, runtimeJudgment, steps, acceptanceCoverage, capabilityEvidence, syncBackReady, gaps, gateNextAction, gateDecision) {
+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,
@@ -707,17 +821,16 @@ 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,
         capabilityEvidence,
-        syncBackReady,
         gaps: [...gaps, ...acceptanceCoverage.flatMap((coverage) => coverage.gaps)],
-        next: nextForTestResult(status, runtimeJudgment, branch, taskId, runId, `artifacts/test-index-${taskId}.json`, gateNextAction, gateDecision),
+        next: nextForTestResult(status, runtimeJudgment, branch, taskId, gateNextAction, gateDecision),
         generatedAt: new Date().toISOString()
     };
 }
@@ -763,12 +876,6 @@ function capabilityClassForRejectedSource(status) {
     }
     return 'diagnostic';
 }
-function renderCapabilityEvidenceClassification(items) {
-    if (items.length === 0) {
-        return '- none';
-    }
-    return items.map((item) => `- [${item.class}] ${item.source}:${item.domainOrSourceId} — ${item.reason}`).join('\n');
-}
 function capabilityEvidenceSummary(items) {
     if (items.length === 0) {
         return 'none';
@@ -779,30 +886,34 @@ function capabilityEvidenceSummary(items) {
     }
     return [...counts.entries()].map(([kind, count]) => `${kind}:${count}`).join(',');
 }
-function nextForTestResult(status, runtimeJudgment, branch, taskId, _runId, indexArtifact, gateNextAction, gateDecision) {
+function nextForTestResult(status, runtimeJudgment, branch, taskId, gateNextAction, gateDecision) {
+    const inspectHint = `Inspect runtime test read model for ${taskId}`;
     if (status === 'PASS') {
-        return `sdd ship --branch ${branch} --dry-run`;
+        return `sdd execute close --branch ${branch} --compact-json`;
     }
     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}.`;
+        return gateDecision ? `Create a decision card for workflow gate ${gateDecision.decisionId}, then rerun ${testValidationUnitCommand(branch)}.` : `Create a decision card, then rerun ${testValidationUnitCommand(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}.`;
+        return gateDecision ? `Review workflow gate ${gateDecision.decisionId} warnings, then rerun ${testValidationUnitCommand(branch)} or proceed only with explicit review.` : `Review workflow gate warnings, then rerun ${testValidationUnitCommand(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 ${branch}.`;
+        return `${inspectHint}, fix failing validation commands, then rerun ${testValidationUnitCommand(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 ${branch}.`;
+    return gateDecision ? `${inspectHint} and workflow gate ${gateDecision.decisionId}, resolve blockers, then rerun ${testValidationUnitCommand(branch)}.` : `${inspectHint}, fix command/evidence gaps, then rerun ${testValidationUnitCommand(branch)}.`;
+}
+function testValidationUnitCommand(branch) {
+    return `sdd execute --branch ${branch} --json`;
 }
 function resultSentenceForTest(result) {
     if (result.status === 'PASS') {
-        return 'Validation and workflow gate passed; proceed to release readiness.';
+        return 'Validation and workflow gate passed; proceed to execute evidence judgment and truthAlignment before release readiness.';
     }
     if (result.validationStatus === 'PASS' && result.workflowGateStatus !== 'PASS') {
         return `Validation passed, but workflow gate returned ${result.workflowGateStatus}.`;
@@ -825,7 +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) {
@@ -868,8 +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');