sdd-agent-platform 0.4.0 → 0.4.1

package/node_modules/@sdd-agent-platform/core/src/lifecycle/ship.ts

@@ -3,6 +3,7 @@ import path from 'node:path';
 import { doctor } from '../doctor/doctor.js';
 import type { DoctorLevel, DoctorReport } from '../doctor/model.js';
 import { getProjectStatus, type ProjectStatus } from '../status/project-status.js';
+import { evaluateAndRecordWorkflowGateDecision } from '../workflow-gate/evidence-packet.js';
 
 export interface ShipReadinessCheck {
   name: string;
@@ -32,6 +33,8 @@ export interface ShipResult {
     dependencyBlockers: number;
     tokenProjection: ProjectStatus['tokenProjection'];
     lifecycleRisk: ProjectStatus['lifecycleRisk'];
+    capabilityHealth: ProjectStatus['capabilityHealth'];
+    workflowGateHealth: ProjectStatus['workflowGateHealth'];
   };
   doctor: {
     status: DoctorLevel;
@@ -46,12 +49,16 @@ export async function runShip(projectRoot: string, options: { branch?: string |
   const checkedAt = new Date().toISOString();
   const projectStatus = await getProjectStatus(projectRoot, { branch: options.branch ?? undefined, branchSource: options.branch ? 'cli_option' : undefined });
   const doctorReport = await doctor(projectRoot, { latestOnly: true, branch: projectStatus.branch });
-  const checks = buildShipChecks(projectStatus, doctorReport);
+  const gateDecision = projectStatus.latestRun?.taskId
+    ? (await evaluateAndRecordWorkflowGateDecision(projectRoot, { branch: projectStatus.branch, taskId: projectStatus.latestRun.taskId, runId: projectStatus.latestRun.runId, decisionKind: 'ship' })).decision
+    : null;
+  const projectStatusWithGate = gateDecision ? await getProjectStatus(projectRoot, { branch: options.branch ?? undefined, branchSource: options.branch ? 'cli_option' : undefined }) : projectStatus;
+  const checks = buildShipChecks(projectStatusWithGate, doctorReport, gateDecision);
   const blocked = checks.some((check) => check.status === 'BLOCKED');
   const status = blocked ? 'BLOCKED' : 'PASS';
   const releasePath = path.join(projectStatus.context.specDir, 'release.md');
   const nextActions = shipNextActions(projectStatus.branch, checks);
-  const releaseDocument = renderReleaseDocument({ checkedAt, projectStatus, doctorReport, checks, nextActions, status });
+  const releaseDocument = renderReleaseDocument({ checkedAt, projectStatus: projectStatusWithGate, doctorReport, checks, nextActions, status });
   const dryRun = options.dryRun === true;
   if (!dryRun) {
     const absoluteReleasePath = path.join(projectRoot, releasePath);
@@ -69,29 +76,31 @@ export async function runShip(projectRoot: string, options: { branch?: string |
     checks,
     nextActions,
     projectStatus: {
-      workflowStatus: projectStatus.workflowStatus,
-      tasks: projectStatus.tasks,
-      taskRisk: projectStatus.taskRisk,
-      latestRun: projectStatus.latestRun,
-      latestRunEvidence: projectStatus.latestRunEvidence,
-      staleReasons: projectStatus.latestRunStaleReasons,
-      affectedFileConflicts: projectStatus.affectedFileConflicts.length,
-      dependencyBlockers: projectStatus.dependencyBlockers.length,
-      tokenProjection: projectStatus.tokenProjection,
-      lifecycleRisk: projectStatus.lifecycleRisk,
+      workflowStatus: projectStatusWithGate.workflowStatus,
+      tasks: projectStatusWithGate.tasks,
+      taskRisk: projectStatusWithGate.taskRisk,
+      latestRun: projectStatusWithGate.latestRun,
+      latestRunEvidence: projectStatusWithGate.latestRunEvidence,
+      staleReasons: projectStatusWithGate.latestRunStaleReasons,
+      affectedFileConflicts: projectStatusWithGate.affectedFileConflicts.length,
+      dependencyBlockers: projectStatusWithGate.dependencyBlockers.length,
+      tokenProjection: projectStatusWithGate.tokenProjection,
+      lifecycleRisk: projectStatusWithGate.lifecycleRisk,
+      capabilityHealth: projectStatusWithGate.capabilityHealth,
+      workflowGateHealth: projectStatusWithGate.workflowGateHealth
     },
     doctor: summarizeDoctor(doctorReport),
     releaseDocument
   };
 }
 
-function buildShipChecks(projectStatus: ProjectStatus, doctorReport: DoctorReport): ShipReadinessCheck[] {
+function buildShipChecks(projectStatus: ProjectStatus, doctorReport: DoctorReport, gateDecision: ProjectStatus['workflowGateHealth']['latestDecision']): ShipReadinessCheck[] {
   const checks: ShipReadinessCheck[] = [];
   checks.push({
     name: 'documents',
     status: projectStatus.documents.specExists && projectStatus.documents.planExists && projectStatus.documents.tasksExists && projectStatus.documents.verifyExists ? 'PASS' : 'BLOCKED',
     message: `spec=${projectStatus.documents.specExists} plan=${projectStatus.documents.planExists} tasks=${projectStatus.documents.tasksExists} verify=${projectStatus.documents.verifyExists}`,
-    nextAction: 'sdd init --branch <branch> --scaffold-docs'
+    nextAction: `Run missing document stages for ${projectStatus.branch}: sdd spec, sdd plan, then sdd tasks so verify.md is regenerated from the frozen tasks contract.`
   });
   checks.push({
     name: 'workflow_gaps',
@@ -107,7 +116,7 @@ function buildShipChecks(projectStatus: ProjectStatus, doctorReport: DoctorRepor
   });
   checks.push({
     name: 'doctor_fast',
-    status: doctorReport.status === 'PASS' ? 'PASS' : 'BLOCKED',
+    status: doctorReport.status === 'FAIL' ? 'BLOCKED' : 'PASS',
     message: `doctor_status=${doctorReport.status}`,
     nextAction: `sdd doctor fast --branch ${projectStatus.branch}`
   });
@@ -155,6 +164,18 @@ function buildShipChecks(projectStatus: ProjectStatus, doctorReport: DoctorRepor
     message: `status=${projectStatus.subagentDispatches.status} blocking_open=${projectStatus.subagentDispatches.blockingOpen} failed=${projectStatus.subagentDispatches.failed} archived=${projectStatus.subagentDispatches.archived} superseded=${projectStatus.subagentDispatches.superseded}`,
     nextAction: subagentDispatchesReady(projectStatus) ? undefined : projectStatus.subagentDispatches.reasons.join(' ')
   });
+  checks.push({
+    name: 'capability_health',
+    status: projectStatus.capabilityHealth.status === 'blocked' ? 'BLOCKED' : 'PASS',
+    message: `status=${projectStatus.capabilityHealth.status} baseline=${projectStatus.capabilityHealth.baselineDomains.length}/${projectStatus.capabilityHealth.requiredProfessionalDomains.length} release_critical_gaps=${projectStatus.capabilityHealth.releaseCriticalGaps.join(',') || 'none'} evidence_candidate=${projectStatus.capabilityHealth.evidence.candidate} evidence_quarantined=${projectStatus.capabilityHealth.evidence.quarantined} evidence_blocked=${projectStatus.capabilityHealth.evidence.blocked}` ,
+    nextAction: projectStatus.capabilityHealth.status === 'blocked' ? `Run sdd test task, sdd test batch, or sdd test wave for ${projectStatus.branch} to record accepted runtime evidence; capability output remains advisory.` : undefined
+  });
+  checks.push({
+    name: 'workflow_gate',
+    status: gateDecision && gateDecision.status !== 'PASS' ? 'BLOCKED' : 'PASS',
+    message: gateDecision ? `decision=${gateDecision.decisionId} status=${gateDecision.status} hard_blocks=${gateDecision.hardBlocks.length} warnings=${gateDecision.warnings.length} human_required=${gateDecision.humanRequired}` : 'decision=none',
+    nextAction: gateDecision && gateDecision.status !== 'PASS' ? gateDecision.allowedNextActions.join(', ') : undefined
+  });
   return checks;
 }
 
@@ -163,8 +184,7 @@ function latestRunReady(projectStatus: ProjectStatus): boolean {
     return false;
   }
   return projectStatus.latestRun.status === 'completed'
-    && projectStatus.latestRun.validationStatus === 'pass'
-    && projectStatus.latestRun.syncBackStatus === 'applied';
+    && projectStatus.latestRun.validationStatus === 'pass';
 }
 
 function lifecycleRiskReady(projectStatus: ProjectStatus): boolean {
@@ -231,6 +251,8 @@ function renderReleaseDocument(input: {
     evidence ? `- latest_run_evidence: route_preflight=${evidence.routePreflight} agent_executions=${evidence.agentExecutions} team_sessions=${evidence.teamSessions} worker_runtimes=${evidence.workerRuntimes} stale_worker_runtimes=${evidence.staleWorkerRuntimes} artifact_ingestions=${evidence.artifactIngestions}` : '- latest_run_evidence: none',
     `- token_health: ${input.projectStatus.tokenProjection.health} estimated_tokens=${input.projectStatus.tokenProjection.estimatedTokens ?? 'unknown'} context_packages=${input.projectStatus.tokenProjection.contextPackages} team_runtime_decisions=${input.projectStatus.tokenProjection.teamRuntimeDecisions}`,
     `- lifecycle_risk: status=${input.projectStatus.lifecycleRisk.status} profile=${input.projectStatus.lifecycleRisk.profile ?? 'none'} approval=${input.projectStatus.lifecycleRisk.approvalPolicy ?? 'none'} input=${input.projectStatus.lifecycleRisk.inputHash ?? 'none'} expected=${input.projectStatus.lifecycleRisk.expectedInputHash}`,
+    `- capability_health: status=${input.projectStatus.capabilityHealth.status} baseline=${input.projectStatus.capabilityHealth.baselineDomains.length}/${input.projectStatus.capabilityHealth.requiredProfessionalDomains.length} active=${input.projectStatus.capabilityHealth.activeDomains.join(',') || 'none'} release_critical_gaps=${input.projectStatus.capabilityHealth.releaseCriticalGaps.join(',') || 'none'} evidence_candidate=${input.projectStatus.capabilityHealth.evidence.candidate} evidence_quarantined=${input.projectStatus.capabilityHealth.evidence.quarantined} evidence_blocked=${input.projectStatus.capabilityHealth.evidence.blocked}`,
+    `- workflow_gate: decisions=${input.projectStatus.workflowGateHealth.decisions} latest=${input.projectStatus.workflowGateHealth.latestDecision ? `${input.projectStatus.workflowGateHealth.latestDecision.decisionKind}:${input.projectStatus.workflowGateHealth.latestDecision.status}:${input.projectStatus.workflowGateHealth.latestDecision.decisionId}` : 'none'}`,
     `- doctor_fast: status=${input.doctorReport.status} checks=${input.doctorReport.checks.length}`,
     '',
     '## Next actions',

package/node_modules/@sdd-agent-platform/core/src/phase8-contracts.test.ts

@@ -208,9 +208,10 @@ test('work units, subagents, context offload, evidence, and model artifacts stay
     policyJudgment: 'PASS',
     commands: [],
     acceptanceCoverage: [],
-    syncBackReady: true,
+    capabilityEvidence: [],
+    syncBackReady: false,
     gaps: [],
-    next: 'sdd sync-back inspect --branch master',
+    next: 'sdd ship --branch master --dry-run',
     generatedAt: now
   };
 

package/node_modules/@sdd-agent-platform/core/src/phase8-risk-kernel.test.ts

@@ -244,6 +244,11 @@ function task(): SddTask {
     title: 'Integrate lifecycle risk diagnostics',
     status: 'pending',
     wave: null,
+    implementationWave: null,
+    validationBatch: null,
+    validationTiming: 'task_end',
+    requiresVerifyBeforeNext: true,
+    changeSurface: 'unknown',
     dependsOn: [],
     affectedFiles: ['packages/core/src/status/project-status.ts'],
     validation: ['npm test'],

package/node_modules/@sdd-agent-platform/core/src/planning/task-graph.test.ts

@@ -7,6 +7,7 @@ import path from 'node:path';
 import { initProject } from '../config/init-project.js';
 import { doctor } from '../doctor/doctor.js';
 import { graphTaskMarkdown, harnessTaskMarkdown, writeBranchDocs } from '../test-support/fixtures.js';
+import { writeVerifyContract } from '../verification/verify-contract.js';
 import { inspectTaskGraph } from './task-graph.js';
 
 test('task graph planner builds dependency and file overlap graph', async () => {
@@ -14,6 +15,7 @@ test('task graph planner builds dependency and file overlap graph', async () =>
   try {
     await initProject(root);
     await writeBranchDocs(root, 'master', `# Tasks\n\n${graphTaskMarkdown('G1', [], ['src/a.ts'], [])}\n${graphTaskMarkdown('G2', ['G1'], ['./src/a.ts', 'src/b.ts'], ['security'])}`);
+    await writeVerifyContract(root, { branch: 'master', branchSource: 'cli_option', force: true });
 
     const graph = await inspectTaskGraph(root, { branch: 'master' });
 

package/node_modules/@sdd-agent-platform/core/src/planning/wave-plan.test.ts

@@ -7,6 +7,7 @@ import path from 'node:path';
 import { initProject } from '../config/init-project.js';
 import { doctor } from '../doctor/doctor.js';
 import { graphTaskMarkdown, writeBranchDocs } from '../test-support/fixtures.js';
+import { writeVerifyContract } from '../verification/verify-contract.js';
 import { inspectWavePlan } from './wave-plan.js';
 
 test('wave planner builds dependency waves and separates file overlaps', async () => {
@@ -14,6 +15,7 @@ test('wave planner builds dependency waves and separates file overlaps', async (
   try {
     await initProject(root);
     await writeBranchDocs(root, 'master', `# Tasks\n\n${graphTaskMarkdown('W1', [], ['src/a.ts'], [])}\n${graphTaskMarkdown('W2', [], ['./src/a.ts'], [])}\n${graphTaskMarkdown('W3', ['W1'], ['src/c.ts'], [])}`);
+    await writeVerifyContract(root, { branch: 'master', branchSource: 'cli_option', force: true });
 
     const plan = await inspectWavePlan(root, { branch: 'master', capabilityId: 'native-file-edit' });
 
@@ -32,6 +34,7 @@ test('wave planner routes manual gates and downstream blocked tasks', async () =
   try {
     await initProject(root);
     await writeBranchDocs(root, 'master', `# Tasks\n\n${graphTaskMarkdown('DB1', [], ['db/schema.sql'], ['database'])}\n${graphTaskMarkdown('APP1', ['DB1'], ['src/app.ts'], [])}`);
+    await writeVerifyContract(root, { branch: 'master', branchSource: 'cli_option', force: true });
 
     const plan = await inspectWavePlan(root, { branch: 'master', capabilityId: 'native-file-edit' });
 

package/node_modules/@sdd-agent-platform/core/src/registries/agent-capability-catalog.ts

@@ -3,29 +3,56 @@ import { parseProjectConfig } from '../config/project-config.js';
 import { getProjectConfigPath } from '../runtime-paths.js';
 import { readFile } from 'node:fs/promises';
 
-export type AgentCapabilityDomain = 'norm_discovery' | 'uncertainty_resolution' | 'performance_planning' | 'verification_design' | 'evidence_collection' | 'sync_back_risk_review' | 'release_summary' | 'context_curation';
+export type WorkflowCapabilityDomain = 'norm_discovery' | 'uncertainty_resolution' | 'performance_planning' | 'verification_design' | 'evidence_collection' | 'sync_back_risk_review' | 'release_summary' | 'context_curation';
+export type ProfessionalCapabilityDomain = 'solution-design' | 'architecture-design' | 'backend-engineering' | 'frontend-engineering' | 'db-data-engineering' | 'testing-quality-engineering' | 'security-engineering' | 'performance-engineering' | 'observability-engineering' | 'release-engineering' | 'ui-ux-product-design';
+export type AgentCapabilityDomain = WorkflowCapabilityDomain | ProfessionalCapabilityDomain;
+export type AgentCapabilityDomainGroup = 'workflow' | 'professional';
+export const REQUIRED_PROFESSIONAL_CAPABILITY_DOMAINS: ProfessionalCapabilityDomain[] = [
+  'solution-design',
+  'architecture-design',
+  'backend-engineering',
+  'frontend-engineering',
+  'db-data-engineering',
+  'testing-quality-engineering',
+  'security-engineering',
+  'performance-engineering',
+  'observability-engineering',
+  'release-engineering',
+  'ui-ux-product-design'
+];
 export type AgentCapabilityStage = 'spec' | 'plan' | 'tasks' | 'verifies' | 'test' | 'do' | 'verify' | 'sync-back' | 'ship';
 export type AgentCapabilityAuthority = 'advisory_only' | 'gate_evidence' | 'validation_runner';
 export type MaterialPackLoadPolicy = 'route_when_triggered' | 'summary_only' | 'never_inline';
+export type AgentCapabilityMaterialPolicy = 'metadata_only' | 'route_baseline_pack' | 'runtime_evidence_required';
+export type AgentCapabilityEvidenceExpectation = 'advisory_note' | 'candidate_evidence' | 'accepted_runtime_evidence';
+export type MaterialPackSourceKind = 'sdd_runtime_contract' | 'sdd_native_baseline';
 
 export interface AgentCapabilityMaterialPack {
   id: string;
   summary: string;
+  domains: AgentCapabilityDomain[];
   triggerStages: AgentCapabilityStage[];
   triggerKeywords: string[];
   loadPolicy: MaterialPackLoadPolicy;
   sourceId: string;
+  sourceVersion: string;
+  sourceKind: MaterialPackSourceKind;
   contextBudget: 'tiny' | 'small' | 'medium';
+  expectedOutputs: string[];
+  forbiddenUses: string[];
 }
 
 export interface AgentCapabilityCatalogEntry {
   version: typeof AGENT_CAPABILITY_CATALOG_CONTRACT_VERSION;
   id: string;
   domain: AgentCapabilityDomain;
+  domainGroup: AgentCapabilityDomainGroup;
   stages: AgentCapabilityStage[];
   inputs: string[];
   outputs: string[];
   authority: AgentCapabilityAuthority;
+  materialPolicy: AgentCapabilityMaterialPolicy;
+  evidenceExpectation: AgentCapabilityEvidenceExpectation;
   routing: {
     riskTags: string[];
     projectStackTags: string[];
@@ -62,22 +89,104 @@ export interface AgentCapabilityCatalogValidation {
 }
 
 const MATERIAL_PACKS: AgentCapabilityMaterialPack[] = [
-  { id: 'project-norms', summary: 'Project conventions, coding standards, and existing-spec alignment cues.', triggerStages: ['spec', 'plan', 'tasks'], triggerKeywords: ['规范', 'convention', 'standard', 'style', 'compatibility'], loadPolicy: 'route_when_triggered', sourceId: 'project_context_pack', contextBudget: 'small' },
-  { id: 'uncertainty-map', summary: 'Decision-boundary prompts for resolving unknowns from repo evidence before asking the user.', triggerStages: ['spec', 'plan'], triggerKeywords: ['uncertain', 'unknown', 'ambiguous', '不确定', '边界'], loadPolicy: 'summary_only', sourceId: 'harness_learning', contextBudget: 'tiny' },
-  { id: 'performance-risk', summary: 'Planning checks for token, context, IO, runtime, and validation cost risks.', triggerStages: ['plan', 'tasks'], triggerKeywords: ['performance', 'token', 'context', 'latency', 'cost', '性能'], loadPolicy: 'summary_only', sourceId: 'skill_agent_eval', contextBudget: 'tiny' },
-  { id: 'verification-design', summary: 'Acceptance-to-evidence design patterns for verify.md and runtime evidence handoff.', triggerStages: ['verifies', 'test', 'verify'], triggerKeywords: ['verify', 'evidence', 'validation', 'acceptance', '验收'], loadPolicy: 'route_when_triggered', sourceId: 'verify_contract', contextBudget: 'small' },
-  { id: 'sync-back-risk', summary: 'Risk checks for applying semantic completion back into tasks and release state.', triggerStages: ['sync-back', 'ship'], triggerKeywords: ['sync-back', 'ship', 'release', 'apply', '发布'], loadPolicy: 'summary_only', sourceId: 'governance_policy', contextBudget: 'tiny' }
+  materialPack(
+    'project-norms',
+    ['norm_discovery', 'context_curation'],
+    'Project conventions, coding standards, and existing-spec alignment cues.',
+    ['spec', 'plan', 'tasks'],
+    ['规范', 'convention', 'standard', 'style', 'compatibility'],
+    'route_when_triggered',
+    'project_context_pack',
+    'sdd_runtime_contract',
+    'small',
+    ['norm alignment note', 'coding convention cue'],
+    ['lifecycle gate authority', 'stage completion decision', 'sync-back apply decision', 'ship readiness pass']
+  ),
+  materialPack(
+    'uncertainty-map',
+    ['uncertainty_resolution', 'context_curation'],
+    'Decision-boundary prompts for resolving unknowns from repo evidence before asking the user.',
+    ['spec', 'plan'],
+    ['uncertain', 'unknown', 'ambiguous', '不确定', '边界'],
+    'summary_only',
+    'harness_learning',
+    'sdd_runtime_contract',
+    'tiny',
+    ['resolved assumption note', 'question checkpoint'],
+    ['lifecycle gate authority', 'stage completion decision', 'sync-back apply decision', 'ship readiness pass']
+  ),
+  materialPack(
+    'performance-risk',
+    ['performance_planning', 'context_curation'],
+    'Planning checks for token, context, IO, runtime, and validation cost risks.',
+    ['plan', 'tasks'],
+    ['performance', 'token', 'context', 'latency', 'cost', '性能'],
+    'summary_only',
+    'skill_agent_eval',
+    'sdd_runtime_contract',
+    'tiny',
+    ['performance risk note', 'context budget note'],
+    ['lifecycle gate authority', 'stage completion decision', 'sync-back apply decision', 'ship readiness pass']
+  ),
+  materialPack(
+    'verification-design',
+    ['verification_design', 'evidence_collection'],
+    'Acceptance-to-evidence design patterns for verify.md and runtime evidence handoff.',
+    ['verifies', 'test', 'verify'],
+    ['verify', 'evidence', 'validation', 'acceptance', '验收'],
+    'route_when_triggered',
+    'verify_contract',
+    'sdd_runtime_contract',
+    'small',
+    ['verification matrix', 'evidence requirement'],
+    ['lifecycle gate authority', 'stage completion decision', 'sync-back apply decision', 'ship readiness pass']
+  ),
+  materialPack(
+    'sync-back-risk',
+    ['sync_back_risk_review', 'release_summary'],
+    'Risk checks for applying semantic completion back into tasks and release state.',
+    ['sync-back', 'ship'],
+    ['sync-back', 'ship', 'release', 'apply', '发布'],
+    'summary_only',
+    'governance_policy',
+    'sdd_runtime_contract',
+    'tiny',
+    ['apply risk note', 'release readiness concern'],
+    ['lifecycle gate authority', 'stage completion decision', 'sync-back apply decision', 'ship readiness pass']
+  ),
+  professionalMaterialPack('baseline-solution-design', 'solution-design', 'Problem framing, option tradeoff, scope boundary, and success criteria review.', ['spec', 'plan'], ['solution', 'scope', 'tradeoff', 'goal', '方案'], ['problem framing', 'option tradeoff', 'scope boundary'], 'small'),
+  professionalMaterialPack('baseline-architecture-design', 'architecture-design', 'System boundary, component responsibility, integration, and runtime constraint review.', ['plan', 'tasks'], ['architecture', 'boundary', 'integration', 'component', '架构'], ['architecture decision note', 'component boundary note', 'handoff risk'], 'small'),
+  professionalMaterialPack('baseline-backend-engineering', 'backend-engineering', 'API, service, domain logic, error handling, and backend validation review.', ['plan', 'tasks', 'do', 'test'], ['backend', 'api', 'service', 'domain', '后端'], ['backend implementation note', 'API risk note', 'validation need'], 'small'),
+  professionalMaterialPack('baseline-frontend-engineering', 'frontend-engineering', 'Component boundary, UI state, browser behavior, accessibility, and frontend validation review.', ['plan', 'tasks', 'do', 'test'], ['frontend', 'component', 'browser', 'accessibility', '前端'], ['frontend implementation note', 'interaction state risk', 'browser validation need'], 'small'),
+  professionalMaterialPack('baseline-db-data-engineering', 'db-data-engineering', 'Schema, query, migration, data integrity, rollback, and data-safety review.', ['plan', 'tasks', 'do', 'test', 'ship'], ['database', 'schema', 'migration', 'query', '数据'], ['data safety note', 'migration risk', 'rollback consideration'], 'small'),
+  professionalMaterialPack('baseline-testing-quality-engineering', 'testing-quality-engineering', 'Acceptance coverage, test strategy, evidence quality, and regression-risk review.', ['verifies', 'test', 'verify'], ['test', 'quality', 'coverage', 'acceptance', '测试'], ['test strategy', 'coverage gap', 'evidence requirement'], 'small'),
+  professionalMaterialPack('baseline-security-engineering', 'security-engineering', 'Trust boundary, secret handling, permissions, abuse case, and security evidence review.', ['spec', 'plan', 'tasks', 'do', 'test', 'ship'], ['security', 'secret', 'permission', 'auth', '安全'], ['security risk note', 'abuse case', 'security evidence requirement'], 'small'),
+  professionalMaterialPack('baseline-performance-engineering', 'performance-engineering', 'Hot path, latency, resource use, benchmark, and capacity-risk review.', ['plan', 'tasks', 'do', 'test', 'ship'], ['performance', 'latency', 'resource', 'benchmark', '性能'], ['performance risk note', 'benchmark need', 'capacity consideration'], 'small'),
+  professionalMaterialPack('baseline-observability-engineering', 'observability-engineering', 'Logging, metrics, tracing, diagnostic signal, and operator-readiness review.', ['plan', 'tasks', 'do', 'test', 'ship'], ['observability', 'logging', 'metrics', 'tracing', '可观测'], ['logging/metrics/tracing need', 'diagnostic gap', 'operability note'], 'small'),
+  professionalMaterialPack('baseline-release-engineering', 'release-engineering', 'Release scope, CI state, deployment risk, rollback, and readiness review.', ['tasks', 'test', 'sync-back', 'ship'], ['release', 'ci', 'deploy', 'rollback', '发布'], ['release readiness note', 'deployment risk', 'ship blocker'], 'small'),
+  professionalMaterialPack('baseline-ui-ux-product-design', 'ui-ux-product-design', 'User journey, interaction state, UX acceptance, accessibility, and design evidence review.', ['spec', 'plan', 'tasks', 'do', 'test'], ['ui', 'ux', 'design', 'figma', '交互'], ['UX acceptance note', 'interaction gap', 'design evidence requirement'], 'small')
 ];
 
 const CAPABILITIES: AgentCapabilityCatalogEntry[] = [
-  capability('cap.norm-discovery', 'norm_discovery', ['spec', 'plan', 'tasks'], ['user request', 'project docs', 'existing specs', 'affected files'], ['norm alignment finding', 'scope constraint'], 'advisory_only', ['compatibility', 'style', 'architecture'], ['project-norms'], 'project_context_pack'),
-  capability('cap.uncertainty-resolution', 'uncertainty_resolution', ['spec', 'plan'], ['user request', 'repo evidence', 'open gaps'], ['resolved assumption', 'question checkpoint', 'blocked gap'], 'advisory_only', ['ambiguous', 'high_risk'], ['uncertainty-map'], 'harness_learning'),
-  capability('cap.performance-planning', 'performance_planning', ['plan', 'tasks'], ['target design', 'runtime path', 'context usage', 'validation cost'], ['performance impact note', 'token risk note', 'validation cost control'], 'advisory_only', ['performance', 'context_budget', 'token_risk'], ['performance-risk'], 'skill_agent_eval'),
-  capability('cap.verification-design', 'verification_design', ['verifies', 'test', 'verify'], ['tasks.md', 'verify.md', 'acceptance refs', 'runtime evidence'], ['verification matrix', 'evidence requirement', 'policy boundary'], 'gate_evidence', ['acceptance', 'evidence', 'policy'], ['verification-design'], 'verify_contract'),
-  capability('cap.evidence-collection', 'evidence_collection', ['test', 'do', 'verify'], ['validation command', 'artifact refs', 'run state'], ['command evidence', 'artifact ref', 'provenance ref'], 'validation_runner', ['validation', 'runtime_evidence'], ['verification-design'], 'test_runtime'),
-  capability('cap.sync-back-risk-review', 'sync_back_risk_review', ['sync-back'], ['verify decision', 'sync-back proposal', 'tasks.md'], ['apply risk note', 'approval requirement'], 'gate_evidence', ['shared_state', 'semantic_update'], ['sync-back-risk'], 'governance_policy'),
-  capability('cap.release-summary', 'release_summary', ['ship'], ['task status', 'doctor report', 'pack output', 'git state'], ['release readiness summary', 'blocked release gap'], 'advisory_only', ['release', 'external_state'], ['sync-back-risk'], 'governance_policy'),
-  capability('cap.context-curation', 'context_curation', ['spec', 'plan', 'do', 'verify', 'ship'], ['material packs', 'context budget', 'run summaries'], ['context pack selection', 'excluded material list'], 'advisory_only', ['context_budget', 'prompt_bloat'], ['project-norms', 'uncertainty-map', 'performance-risk'], 'project_context_pack')
+  workflowCapability('cap.norm-discovery', 'norm_discovery', ['spec', 'plan', 'tasks'], ['user request', 'project docs', 'existing specs', 'affected files'], ['norm alignment finding', 'scope constraint'], 'advisory_only', ['compatibility', 'style', 'architecture'], ['project-norms'], 'project_context_pack'),
+  workflowCapability('cap.uncertainty-resolution', 'uncertainty_resolution', ['spec', 'plan'], ['user request', 'repo evidence', 'open gaps'], ['resolved assumption', 'question checkpoint', 'blocked gap'], 'advisory_only', ['ambiguous', 'high_risk'], ['uncertainty-map'], 'harness_learning'),
+  workflowCapability('cap.performance-planning', 'performance_planning', ['plan', 'tasks'], ['target design', 'runtime path', 'context usage', 'validation cost'], ['performance impact note', 'token risk note', 'validation cost control'], 'advisory_only', ['performance', 'context_budget', 'token_risk'], ['performance-risk'], 'skill_agent_eval'),
+  workflowCapability('cap.verification-design', 'verification_design', ['verifies', 'test', 'verify'], ['tasks.md', 'verify.md', 'acceptance refs', 'runtime evidence'], ['verification matrix', 'evidence requirement', 'policy boundary'], 'gate_evidence', ['acceptance', 'evidence', 'policy'], ['verification-design'], 'verify_contract'),
+  workflowCapability('cap.evidence-collection', 'evidence_collection', ['test', 'do', 'verify'], ['validation command', 'artifact refs', 'run state'], ['command evidence', 'artifact ref', 'provenance ref'], 'validation_runner', ['validation', 'runtime_evidence'], ['verification-design'], 'test_runtime'),
+  workflowCapability('cap.sync-back-risk-review', 'sync_back_risk_review', ['sync-back'], ['verify decision', 'sync-back proposal', 'tasks.md'], ['apply risk note', 'approval requirement'], 'gate_evidence', ['shared_state', 'semantic_update'], ['sync-back-risk'], 'governance_policy'),
+  workflowCapability('cap.release-summary', 'release_summary', ['ship'], ['task status', 'doctor report', 'pack output', 'git state'], ['release readiness summary', 'blocked release gap'], 'advisory_only', ['release', 'external_state'], ['sync-back-risk'], 'governance_policy'),
+  workflowCapability('cap.context-curation', 'context_curation', ['spec', 'plan', 'do', 'verify', 'ship'], ['material packs', 'context budget', 'run summaries'], ['context pack selection', 'excluded material list'], 'advisory_only', ['context_budget', 'prompt_bloat'], ['project-norms', 'uncertainty-map', 'performance-risk'], 'project_context_pack'),
+  professionalCapability('cap.solution-design', 'solution-design', ['spec', 'plan'], ['user goal', 'constraints', 'success criteria'], ['problem framing', 'solution option tradeoff', 'scope boundary'], ['architecture', 'ambiguous'], ['baseline-solution-design', 'project-norms', 'uncertainty-map']),
+  professionalCapability('cap.architecture-design', 'architecture-design', ['plan', 'tasks'], ['system boundaries', 'integration points', 'runtime constraints'], ['architecture decision', 'component boundary', 'handoff risk'], ['architecture', 'source_boundary'], ['baseline-architecture-design', 'project-norms', 'performance-risk']),
+  professionalCapability('cap.backend-engineering', 'backend-engineering', ['plan', 'tasks', 'do', 'test'], ['API contracts', 'service logic', 'runtime path'], ['backend implementation note', 'API risk note', 'validation need'], ['api-schema', 'runtime_state'], ['baseline-backend-engineering', 'project-norms', 'verification-design']),
+  professionalCapability('cap.frontend-engineering', 'frontend-engineering', ['plan', 'tasks', 'do', 'test'], ['UI state', 'component boundary', 'browser behavior'], ['frontend implementation note', 'interaction state risk', 'browser validation need'], ['frontend', 'accessibility'], ['baseline-frontend-engineering', 'project-norms', 'verification-design']),
+  professionalCapability('cap.db-data-engineering', 'db-data-engineering', ['plan', 'tasks', 'do', 'test', 'ship'], ['schema change', 'query path', 'migration plan'], ['data safety note', 'migration risk', 'rollback consideration'], ['database', 'data-loss'], ['baseline-db-data-engineering', 'project-norms', 'verification-design']),
+  professionalCapability('cap.testing-quality-engineering', 'testing-quality-engineering', ['verifies', 'test', 'verify'], ['acceptance refs', 'test command', 'quality risk'], ['test strategy', 'coverage gap', 'evidence requirement'], ['validation', 'acceptance'], ['baseline-testing-quality-engineering', 'verification-design']),
+  professionalCapability('cap.security-engineering', 'security-engineering', ['spec', 'plan', 'tasks', 'do', 'test', 'ship'], ['trust boundary', 'secret handling', 'permission model'], ['security risk note', 'abuse case', 'security evidence requirement'], ['security', 'token_secret'], ['baseline-security-engineering', 'project-norms', 'verification-design']),
+  professionalCapability('cap.performance-engineering', 'performance-engineering', ['plan', 'tasks', 'do', 'test', 'ship'], ['hot path', 'resource use', 'latency constraint'], ['performance risk note', 'benchmark need', 'capacity consideration'], ['performance', 'context_budget'], ['baseline-performance-engineering', 'performance-risk', 'verification-design']),
+  professionalCapability('cap.observability-engineering', 'observability-engineering', ['plan', 'tasks', 'do', 'test', 'ship'], ['runtime signal', 'failure mode', 'operator need'], ['logging/metrics/tracing need', 'diagnostic gap', 'operability note'], ['observability', 'runtime_state'], ['baseline-observability-engineering', 'project-norms', 'verification-design']),
+  professionalCapability('cap.release-engineering', 'release-engineering', ['tasks', 'test', 'sync-back', 'ship'], ['release scope', 'CI state', 'rollback need'], ['release readiness note', 'deployment risk', 'ship blocker'], ['release', 'ci-build'], ['baseline-release-engineering', 'sync-back-risk', 'verification-design']),
+  professionalCapability('cap.ui-ux-product-design', 'ui-ux-product-design', ['spec', 'plan', 'tasks', 'do', 'test'], ['user journey', 'interaction state', 'design constraint'], ['UX acceptance note', 'interaction gap', 'design evidence requirement'], ['ui-ux', 'accessibility'], ['baseline-ui-ux-product-design', 'project-norms', 'uncertainty-map'])
 ];
 
 const COMMAND_MAPPINGS: CapabilityCommandMapping[] = [
@@ -111,18 +220,29 @@ export async function validateAgentCapabilityCatalog(projectRoot: string): Promi
   };
 }
 
-function capability(id: string, domain: AgentCapabilityDomain, stages: AgentCapabilityStage[], inputs: string[], outputs: string[], authority: AgentCapabilityAuthority, riskTags: string[], materialPackIds: string[], sourceId: string): AgentCapabilityCatalogEntry {
+function workflowCapability(id: string, domain: WorkflowCapabilityDomain, stages: AgentCapabilityStage[], inputs: string[], outputs: string[], authority: AgentCapabilityAuthority, riskTags: string[], materialPackIds: string[], sourceId: string): AgentCapabilityCatalogEntry {
+  return capability(id, domain, 'workflow', stages, inputs, outputs, authority, riskTags, materialPackIds, sourceId);
+}
+
+function professionalCapability(id: string, domain: ProfessionalCapabilityDomain, stages: AgentCapabilityStage[], inputs: string[], outputs: string[], riskTags: string[], materialPackIds: string[]): AgentCapabilityCatalogEntry {
+  return capability(id, domain, 'professional', stages, inputs, outputs, 'advisory_only', riskTags, materialPackIds, 'sdd_professional_baseline');
+}
+
+function capability(id: string, domain: AgentCapabilityDomain, domainGroup: AgentCapabilityDomainGroup, stages: AgentCapabilityStage[], inputs: string[], outputs: string[], authority: AgentCapabilityAuthority, riskTags: string[], materialPackIds: string[], sourceId: string): AgentCapabilityCatalogEntry {
   return {
     version: AGENT_CAPABILITY_CATALOG_CONTRACT_VERSION,
     id,
     domain,
+    domainGroup,
     stages,
     inputs,
     outputs,
     authority,
+    materialPolicy: capabilityMaterialPolicy(authority),
+    evidenceExpectation: capabilityEvidenceExpectation(authority),
     routing: {
       riskTags,
-      projectStackTags: ['generic'],
+      projectStackTags: projectStackTagsForDomain(domain),
       confidenceThreshold: 0.65,
       materialPackIds
     },
@@ -134,10 +254,112 @@ function capability(id: string, domain: AgentCapabilityDomain, stages: AgentCapa
   };
 }
 
+function capabilityMaterialPolicy(authority: AgentCapabilityAuthority): AgentCapabilityMaterialPolicy {
+  return authority === 'validation_runner' ? 'runtime_evidence_required' : 'route_baseline_pack';
+}
+
+function capabilityEvidenceExpectation(authority: AgentCapabilityAuthority): AgentCapabilityEvidenceExpectation {
+  if (authority === 'validation_runner' || authority === 'gate_evidence') {
+    return 'accepted_runtime_evidence';
+  }
+  return 'candidate_evidence';
+}
+
+function projectStackTagsForDomain(domain: AgentCapabilityDomain): string[] {
+  if (domain === 'frontend-engineering' || domain === 'ui-ux-product-design') {
+    return ['frontend', 'browser', 'component', 'tsx', 'jsx', 'css'];
+  }
+  if (domain === 'backend-engineering') {
+    return ['backend', 'api', 'controller', 'service', 'java', 'go', 'python'];
+  }
+  if (domain === 'db-data-engineering') {
+    return ['sql', 'database', 'migration', 'prisma'];
+  }
+  if (domain === 'testing-quality-engineering') {
+    return ['test', 'spec', 'playwright', 'vitest', 'jest'];
+  }
+  if (domain === 'security-engineering') {
+    return ['security', 'auth', 'secret', 'permission'];
+  }
+  if (domain === 'performance-engineering') {
+    return ['performance', 'latency', 'benchmark', 'resource'];
+  }
+  if (domain === 'observability-engineering') {
+    return ['observability', 'logging', 'metrics', 'tracing'];
+  }
+  if (domain === 'release-engineering') {
+    return ['release', 'ci', 'deploy', 'rollback'];
+  }
+  return ['generic'];
+}
+
+function materialPack(id: string, domains: AgentCapabilityDomain[], summary: string, triggerStages: AgentCapabilityStage[], triggerKeywords: string[], loadPolicy: MaterialPackLoadPolicy, sourceId: string, sourceKind: MaterialPackSourceKind, contextBudget: AgentCapabilityMaterialPack['contextBudget'], expectedOutputs: string[], forbiddenUses: string[]): AgentCapabilityMaterialPack {
+  return {
+    id,
+    summary,
+    domains,
+    triggerStages,
+    triggerKeywords,
+    loadPolicy,
+    sourceId,
+    sourceVersion: '0.4.1',
+    sourceKind,
+    contextBudget,
+    expectedOutputs,
+    forbiddenUses
+  };
+}
+
+function professionalMaterialPack(id: string, domain: ProfessionalCapabilityDomain, summary: string, triggerStages: AgentCapabilityStage[], triggerKeywords: string[], expectedOutputs: string[], contextBudget: AgentCapabilityMaterialPack['contextBudget']): AgentCapabilityMaterialPack {
+  return materialPack(
+    id,
+    [domain],
+    summary,
+    triggerStages,
+    triggerKeywords,
+    'route_when_triggered',
+    'sdd_professional_baseline',
+    'sdd_native_baseline',
+    contextBudget,
+    expectedOutputs,
+    ['lifecycle gate authority', 'stage completion decision', 'sync-back apply decision', 'ship readiness pass', 'host-specific skill projection']
+  );
+}
+
+
 function validateCatalog(catalog: AgentCapabilityCatalog): string[] {
   const issues: string[] = [];
-  const packIds = new Set(catalog.materialPacks.map((pack) => pack.id));
+  const packById = new Map(catalog.materialPacks.map((pack) => [pack.id, pack]));
+  const packIds = new Set(packById.keys());
   const domains = new Set(catalog.capabilities.map((entry) => entry.domain));
+  const professionalDomains = new Set(catalog.capabilities.filter((entry) => entry.domainGroup === 'professional').map((entry) => entry.domain));
+  const professionalBaselineDomains = new Set<ProfessionalCapabilityDomain>();
+  for (const pack of catalog.materialPacks) {
+    if (pack.domains.length === 0) {
+      issues.push(`${pack.id}: domains are required.`);
+    }
+    if (pack.triggerStages.length === 0 || pack.triggerKeywords.length === 0) {
+      issues.push(`${pack.id}: trigger stages and keywords are required.`);
+    }
+    if (!pack.sourceId || !pack.sourceVersion) {
+      issues.push(`${pack.id}: source provenance is required.`);
+    }
+    if (pack.expectedOutputs.length === 0 || pack.forbiddenUses.length === 0) {
+      issues.push(`${pack.id}: expected outputs and forbidden uses are required.`);
+    }
+    if (pack.sourceKind === 'sdd_native_baseline') {
+      if (!pack.forbiddenUses.includes('host-specific skill projection')) {
+        issues.push(`${pack.id}: SDD-native baseline packs must forbid host-specific skill projection.`);
+      }
+      for (const domain of pack.domains) {
+        if (isRequiredProfessionalDomain(domain)) {
+          professionalBaselineDomains.add(domain);
+        } else {
+          issues.push(`${pack.id}: SDD-native baseline packs must target professional domains.`);
+        }
+      }
+    }
+  }
   for (const capability of catalog.capabilities) {
     if (capability.stages.length === 0) {
       issues.push(`${capability.id}: stages are required.`);
@@ -145,6 +367,12 @@ function validateCatalog(catalog: AgentCapabilityCatalog): string[] {
     if (capability.inputs.length === 0 || capability.outputs.length === 0) {
       issues.push(`${capability.id}: inputs and outputs are required.`);
     }
+    if (!capability.materialPolicy || !capability.evidenceExpectation) {
+      issues.push(`${capability.id}: material policy and evidence expectation are required.`);
+    }
+    if (capability.routing.projectStackTags.length === 0) {
+      issues.push(`${capability.id}: project stack tags are required.`);
+    }
     if (capability.routing.confidenceThreshold <= 0 || capability.routing.confidenceThreshold > 1) {
       issues.push(`${capability.id}: confidence threshold must be within (0, 1].`);
     }
@@ -153,6 +381,26 @@ function validateCatalog(catalog: AgentCapabilityCatalog): string[] {
         issues.push(`${capability.id}: unknown material pack ${packId}.`);
       }
     }
+    if (capability.domainGroup === 'professional') {
+      if (capability.authority !== 'advisory_only') {
+        issues.push(`${capability.id}: professional capabilities must remain advisory_only.`);
+      }
+      const hasDomainBaselinePack = capability.routing.materialPackIds.some((packId) => {
+        const pack = packById.get(packId);
+        return pack?.sourceKind === 'sdd_native_baseline' && pack.domains.includes(capability.domain);
+      });
+      if (!hasDomainBaselinePack) {
+        issues.push(`${capability.id}: professional capability must reference an SDD-native baseline pack for ${capability.domain}.`);
+      }
+    }
+  }
+  for (const domain of REQUIRED_PROFESSIONAL_CAPABILITY_DOMAINS) {
+    if (!professionalDomains.has(domain)) {
+      issues.push(`required professional domain ${domain} is missing.`);
+    }
+    if (!professionalBaselineDomains.has(domain)) {
+      issues.push(`required professional domain ${domain} has no SDD-native baseline pack.`);
+    }
   }
   for (const mapping of catalog.commandMappings) {
     for (const domain of mapping.requiredDomains) {
@@ -167,6 +415,10 @@ function validateCatalog(catalog: AgentCapabilityCatalog): string[] {
   return issues;
 }
 
+function isRequiredProfessionalDomain(domain: AgentCapabilityDomain): domain is ProfessionalCapabilityDomain {
+  return REQUIRED_PROFESSIONAL_CAPABILITY_DOMAINS.includes(domain as ProfessionalCapabilityDomain);
+}
+
 async function assertProjectConfigReadable(projectRoot: string): Promise<void> {
   const configPath = getProjectConfigPath(projectRoot);
   const raw = await readFile(configPath, 'utf8');

package/node_modules/@sdd-agent-platform/core/src/registries/agent-registry.ts

@@ -115,14 +115,14 @@ const BUILT_IN_AGENT_REGISTRY: AgentRegistryEntry[] = [
     id: 'validator',
     role: 'Map acceptance criteria to review, diff, and command evidence.',
     allowedStages: ['verify'],
-    capabilities: ['run declared validation', 'acceptance mapping', 'PASS/PASS_WITH_GAPS/FAIL/BLOCKED decision'],
+    capabilities: ['run declared validation', 'acceptance mapping', 'evidence sufficiency summary'],
     readBoundary: ['task acceptance', 'review artifact', 'validation outputs'],
     writeBoundary: ['artifacts/validation-<task>.md'],
     toolAllowlist: ['read', 'validation command'],
     requiredArtifact: 'artifacts/validation-<task>.md',
     verificationExpectation: 'PASS requires acceptance evidence, not just command success.',
     autonomyCeiling: 'validation_only',
-    stopCondition: 'Stop when acceptance mapping and sync-back recommendation are clear.'
+    stopCondition: 'Stop when acceptance evidence mapping and unresolved gaps are clear.'
   }
 ];