create-byan-agent 2.19.2 → 2.20.1

package/install/templates/.claude/workflows/testarch-nfr.js

@@ -0,0 +1,316 @@
+export const meta = {
+  name: 'testarch-nfr',
+  description: 'Native port of the BYAN testarch-nfr workflow (create mode): assess non-functional requirements before release with evidence-based PASS/CONCERNS/FAIL outcomes. Load context+knowledge, define NFR categories+thresholds, gather evidence, fan out 4 parallel domain subprocesses (security/performance/reliability/scalability), aggregate into an executive summary, generate the report, then return a structured verdict for the orchestrating skill to present at the human gate.',
+  phases: [
+    { title: 'CONTEXT', detail: 'step-01: load NFR requirements, evidence sources, knowledge base' },
+    { title: 'THRESHOLDS', detail: 'step-02: select NFR categories (ADR 8) + extract/UNKNOWN thresholds' },
+    { title: 'EVIDENCE', detail: 'step-03: gather measurable evidence per category; gaps -> CONCERNS' },
+    { title: 'ASSESS', detail: 'step-04: parallel fan-out of 4 NFR domain subprocesses (security/performance/reliability/scalability)' },
+    { title: 'AGGREGATE', detail: 'step-04e: read 4 outputs, overall risk, compliance, cross-domain risks, exec summary' },
+    { title: 'REPORT', detail: 'step-05: produce nfr-assessment.md, validate vs checklist, completion summary' },
+  ],
+}
+
+// ---------------------------------------------------------------------------
+// FD / STRICT STATE CONTRACT  (re-asserted inline — enforcement-bridge).
+//
+// The in-CLI Workflow tool runs this script OUTSIDE the conversation turn, so
+// BYAN's main-thread hooks (fd-phase-guard, strict-scope-guard, strict-stop-
+// guard, mantra-validate) DO NOT fire here. This script therefore:
+//   - NEVER imports/requires _byan/.../lib/fd-state.js and NEVER writes
+//     fd-state.json directly  (enforced by byan-lint-workflows.js).
+//   - uses NO wall-clock (wall-clock/wall-clock) and NO randomness (RNG):
+//     those break Workflow resume. The unique run id used by the source for
+//     the /tmp/tea-nfr-*-{timestamp}.json subprocess outputs is passed IN via
+//     args.runId (the source generated it with an injected timestamp).
+//   - returns DATA only. The orchestrating skill is the human-gated conductor;
+//     IT records FD/strict state via the byan_fd_* / byan_strict_* MCP tools
+//     AT the gate. The nfr-assessment.md report is the workflow's product,
+//     written by the REPORT leaf — that is the artifact, not platform state.
+// ---------------------------------------------------------------------------
+
+// Risk hierarchy mirrors steps-c/step-04e-aggregate-nfr.md (HIGH>MEDIUM>LOW>NONE).
+// Inlined (sandbox forbids import) so the prose rule becomes a real reduction.
+const RISK_RANK = { HIGH: 3, MEDIUM: 2, LOW: 1, NONE: 0 }
+function overallRisk(levels) {
+  const max = levels.reduce((m, r) => Math.max(m, RISK_RANK[r] || 0), 0)
+  return Object.keys(RISK_RANK).find((k) => RISK_RANK[k] === max) || 'NONE'
+}
+
+// Per-domain subprocess output schema — mirrors the OUTPUT FORMAT JSON in
+// steps-c/step-04a..04d (domain, risk_level, findings[], compliance, priority_actions, summary).
+const DOMAIN_SCHEMA = {
+  type: 'object',
+  required: ['domain', 'risk_level', 'findings', 'summary'],
+  properties: {
+    domain: { type: 'string' },
+    risk_level: { type: 'string', enum: ['HIGH', 'MEDIUM', 'LOW', 'NONE'] },
+    findings: {
+      type: 'array',
+      items: {
+        type: 'object',
+        required: ['category', 'status'],
+        properties: {
+          category: { type: 'string' },
+          status: { type: 'string', enum: ['PASS', 'CONCERN', 'FAIL', 'N/A'] },
+          description: { type: 'string' },
+          evidence: { type: 'array', items: { type: 'string' } },
+          recommendations: { type: 'array', items: { type: 'string' } },
+        },
+      },
+    },
+    compliance: { type: 'object' },
+    priority_actions: { type: 'array', items: { type: 'string' } },
+    summary: { type: 'string' },
+  },
+}
+
+const story = (args && args.story) || (args && args.target) || 'the implementation under release review'
+const runId = (args && args.runId) || (args && args.timestamp) || 'nfr-run'
+const customCategories = (args && args.customNfrCategories) || ''
+
+// === Step 1: Load Context & Knowledge Base (steps-c/step-01) ===
+phase('CONTEXT')
+const context = await agent(
+  `You are the Master Test Architect running testarch-nfr (create mode). Target: ${JSON.stringify(story)}. ` +
+    `Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-01-load-context.md and follow it.\n` +
+    `1) Verify prerequisites: implementation accessible for evaluation AND evidence sources available (test results, metrics, logs). If either is missing, set halt=true and name the missing inputs (do NOT invent evidence).\n` +
+    `2) Load knowledge fragments from the tea index (adr-quality-readiness-checklist, ci-burn-in, test-quality, playwright-config, error-handling).\n` +
+    `3) If available, read tech-spec.md (primary NFRs), PRD.md (product NFRs), story/test-design docs (feature NFRs).\n` +
+    `4) Summarize the loaded NFR sources and evidence availability.`,
+  {
+    label: 'load-context',
+    phase: 'CONTEXT',
+    schema: {
+      type: 'object',
+      required: ['halt'],
+      properties: {
+        halt: { type: 'boolean', description: 'true if prerequisites (implementation or evidence) are missing' },
+        missingInputs: { type: 'array', items: { type: 'string' } },
+        nfrSources: { type: 'array', items: { type: 'string' } },
+        evidenceAvailable: { type: 'array', items: { type: 'string' } },
+        summary: { type: 'string' },
+      },
+    },
+  }
+)
+
+if (context && context.halt) {
+  // step-01 mandatory rule: HALT if implementation or evidence is unavailable.
+  // Surface as a verdict (gap, not silent cut); the gate decides.
+  return {
+    workflow: 'testarch-nfr',
+    target: story,
+    status: 'halted-missing-inputs',
+    summary: 'NFR assessment halted at CONTEXT: prerequisites (implementation/evidence) missing.',
+    missingInputs: (context && context.missingInputs) || [],
+    steps: 6,
+    needsHumanGate: true,
+  }
+}
+
+// === Step 2: Define NFR Categories & Thresholds (steps-c/step-02) ===
+phase('THRESHOLDS')
+const thresholds = await agent(
+  `testarch-nfr step 2 (define thresholds). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-02-define-thresholds.md.\n` +
+    `Context summary: ${JSON.stringify((context && context.summary) || '')}.\n` +
+    `1) Select the 8 ADR Quality Readiness categories (Testability & Automation, Test Data Strategy, Scalability & Availability, Disaster Recovery, Security, Monitorability/Debuggability/Manageability, QoS/QoE, Deployability). Add any custom categories: ${JSON.stringify(customCategories)}.\n` +
+    `2) For each category extract thresholds from tech-spec (primary), PRD (secondary), story/test-design (feature-specific). NEVER guess a threshold: if unknown, mark it UNKNOWN and plan to report CONCERNS for it.\n` +
+    `3) Produce the NFR matrix: each category with its threshold or UNKNOWN status.`,
+  {
+    label: 'define-thresholds',
+    phase: 'THRESHOLDS',
+    schema: {
+      type: 'object',
+      required: ['matrix'],
+      properties: {
+        matrix: {
+          type: 'array',
+          items: {
+            type: 'object',
+            required: ['category', 'threshold'],
+            properties: {
+              category: { type: 'string' },
+              threshold: { type: 'string', description: 'concrete threshold or "UNKNOWN"' },
+              unknown: { type: 'boolean' },
+            },
+          },
+        },
+        notes: { type: 'string' },
+      },
+    },
+  }
+)
+
+// === Step 3: Gather Evidence (steps-c/step-03) ===
+phase('EVIDENCE')
+const evidence = await agent(
+  `testarch-nfr step 3 (gather evidence). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-03-gather-evidence.md.\n` +
+    `NFR matrix: ${JSON.stringify((thresholds && thresholds.matrix) || [])}.\n` +
+    `1) Collect MEASURABLE evidence per category: Performance (load tests, metrics, response-time data); Security (scans, auth tests, vuln reports); Reliability (error rates, burn-in runs, failover tests); Maintainability (test quality, code-health signals); Other (logs, monitoring, DR drills, deployability checks).\n` +
+    `2) Where evidence is MISSING for a category, mark that category CONCERNS (do not fabricate evidence).`,
+  {
+    label: 'gather-evidence',
+    phase: 'EVIDENCE',
+    schema: {
+      type: 'object',
+      required: ['evidenceByCategory'],
+      properties: {
+        evidenceByCategory: {
+          type: 'array',
+          items: {
+            type: 'object',
+            required: ['category'],
+            properties: {
+              category: { type: 'string' },
+              evidence: { type: 'array', items: { type: 'string' } },
+              missing: { type: 'boolean', description: 'true -> category marked CONCERNS for evidence gap' },
+            },
+          },
+        },
+        gaps: { type: 'array', items: { type: 'string' } },
+        notes: { type: 'string' },
+      },
+    },
+  }
+)
+
+// === Step 4: Orchestrate parallel NFR assessment — 4 subprocesses (steps-c/step-04 + 04a..04d) ===
+// Source step-04 launches FOUR subprocesses in PARALLEL and waits for ALL.
+// parallel() mirrors that fan-out exactly; each subprocess assesses ONE domain
+// and writes /tmp/tea-nfr-<domain>-<runId>.json. runId is passed via args (no clock).
+phase('ASSESS')
+const sharedContext =
+  `Shared context — system: ${JSON.stringify((context && context.summary) || '')}; ` +
+  `thresholds: ${JSON.stringify((thresholds && thresholds.matrix) || [])}; ` +
+  `evidence: ${JSON.stringify((evidence && evidence.evidenceByCategory) || [])}. ` +
+  `Run id (use for the temp output file name): ${JSON.stringify(runId)}.`
+
+const [security, performance, reliability, scalability] = await parallel([
+  () =>
+    agent(
+      `testarch-nfr subprocess 4A (SECURITY only). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-04a-subprocess-security.md.\n` +
+        `${sharedContext}\n` +
+        `Assess SECURITY ONLY (auth/authz, data protection, input validation, API security, secrets management). ` +
+        `Set per-finding status PASS/CONCERN/FAIL/N/A, a domain risk_level (HIGH/MEDIUM/LOW/NONE), compliance (SOC2/GDPR/HIPAA/PCI-DSS/ISO27001), priority_actions and a summary. ` +
+        `Write the JSON to /tmp/tea-nfr-security-${runId}.json. Do NOT assess other domains.`,
+      { label: 'sub-security', phase: 'ASSESS', schema: DOMAIN_SCHEMA }
+    ),
+  () =>
+    agent(
+      `testarch-nfr subprocess 4B (PERFORMANCE only). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-04b-subprocess-performance.md.\n` +
+        `${sharedContext}\n` +
+        `Assess PERFORMANCE ONLY (response times, throughput, resource usage, optimization/caching/CDN/indexing). ` +
+        `Set per-finding status, domain risk_level, compliance (SLA tiers), priority_actions and a summary. ` +
+        `Write the JSON to /tmp/tea-nfr-performance-${runId}.json. Do NOT assess other domains.`,
+      { label: 'sub-performance', phase: 'ASSESS', schema: DOMAIN_SCHEMA }
+    ),
+  () =>
+    agent(
+      `testarch-nfr subprocess 4C (RELIABILITY only). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-04c-subprocess-reliability.md.\n` +
+        `${sharedContext}\n` +
+        `Assess RELIABILITY ONLY (error handling/circuit breakers/retries, monitoring & observability, fault tolerance/failover/backup/DR, uptime & availability). ` +
+        `Set per-finding status, domain risk_level, compliance, priority_actions and a summary. ` +
+        `Write the JSON to /tmp/tea-nfr-reliability-${runId}.json. Do NOT assess other domains.`,
+      { label: 'sub-reliability', phase: 'ASSESS', schema: DOMAIN_SCHEMA }
+    ),
+  () =>
+    agent(
+      `testarch-nfr subprocess 4D (SCALABILITY only). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-04d-subprocess-scalability.md.\n` +
+        `${sharedContext}\n` +
+        `Assess SCALABILITY ONLY (horizontal scaling, vertical scaling, data scaling/sharding/replicas, traffic handling/CDN/queues). ` +
+        `Set per-finding status, domain risk_level, compliance (user-scale tiers), priority_actions and a summary. ` +
+        `Write the JSON to /tmp/tea-nfr-scalability-${runId}.json. Do NOT assess other domains.`,
+      { label: 'sub-scalability', phase: 'ASSESS', schema: DOMAIN_SCHEMA }
+    ),
+])
+
+// === Step 4E: Aggregate the 4 domain assessments (steps-c/step-04e) ===
+phase('AGGREGATE')
+const domainRisks = [security, performance, reliability, scalability].map((d) => (d && d.risk_level) || 'NONE')
+const computedOverall = overallRisk(domainRisks)
+log(`domain risks: security=${domainRisks[0]} performance=${domainRisks[1]} reliability=${domainRisks[2]} scalability=${domainRisks[3]} -> overall=${computedOverall}`)
+
+const aggregate = await agent(
+  `testarch-nfr step 4E (aggregate). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-04e-aggregate-nfr.md.\n` +
+    `Four domain outputs: security=${JSON.stringify(security)}; performance=${JSON.stringify(performance)}; reliability=${JSON.stringify(reliability)}; scalability=${JSON.stringify(scalability)}.\n` +
+    `Computed overall risk (HIGH>MEDIUM>LOW>NONE) is ${JSON.stringify(computedOverall)} — use it.\n` +
+    `Do NOT re-assess any domain (use the subprocess outputs). 1) Aggregate compliance per standard: FAIL if any domain FAILs, else PARTIAL if any PARTIAL/CONCERN, else PASS. ` +
+    `2) Identify cross-domain risks (e.g. performance+scalability concerns; security FAIL + reliability concern). ` +
+    `3) Aggregate priority actions, marking URGENT those from a HIGH-risk domain. ` +
+    `4) Produce an executive summary (overall risk, per-domain breakdown, compliance summary, cross-domain risks, prioritized actions).`,
+  {
+    label: 'aggregate',
+    phase: 'AGGREGATE',
+    schema: {
+      type: 'object',
+      required: ['overallRisk', 'riskBreakdown'],
+      properties: {
+        overallRisk: { type: 'string', enum: ['HIGH', 'MEDIUM', 'LOW', 'NONE'] },
+        riskBreakdown: {
+          type: 'object',
+          properties: {
+            security: { type: 'string' },
+            performance: { type: 'string' },
+            reliability: { type: 'string' },
+            scalability: { type: 'string' },
+          },
+        },
+        complianceSummary: { type: 'object' },
+        crossDomainRisks: { type: 'array', items: { type: 'object' } },
+        priorityActions: { type: 'array', items: { type: 'object' } },
+        executiveSummary: { type: 'string' },
+      },
+    },
+  }
+)
+
+// === Step 5: Generate Report & Validate (steps-c/step-05) ===
+phase('REPORT')
+const report = await agent(
+  `testarch-nfr step 5 (generate report & validate). Read _byan/workflow/simple/testarch/nfr-assess/steps-c/step-05-generate-report.md, the nfr-report-template.md, and the workflow checklist.md.\n` +
+    `Executive aggregate: ${JSON.stringify(aggregate)}.\n` +
+    `1) Using nfr-report-template.md, write the report to the configured output folder as nfr-assessment.md: per-category results (PASS/CONCERNS/FAIL), evidence summary, remediation actions, and a gate-ready YAML snippet if applicable. ` +
+    `2) Validate the report against checklist.md and fix any gaps. ` +
+    `3) Completion summary: overall NFR status, critical blockers or waivers needed, and the next recommended workflow (trace or release gate).`,
+  {
+    label: 'generate-report',
+    phase: 'REPORT',
+    schema: {
+      type: 'object',
+      required: ['reportPath', 'overallStatus', 'valid'],
+      properties: {
+        reportPath: { type: 'string' },
+        overallStatus: { type: 'string', enum: ['PASS', 'CONCERNS', 'FAIL'] },
+        valid: { type: 'boolean', description: 'true only if the report satisfies checklist.md' },
+        blockers: { type: 'array', items: { type: 'string' } },
+        waiversNeeded: { type: 'array', items: { type: 'string' } },
+        nextWorkflow: { type: 'string' },
+        gaps: { type: 'array', items: { type: 'string' } },
+        summary: { type: 'string' },
+      },
+    },
+  }
+)
+
+// Return DATA only. The orchestrating skill presents this at the human gate
+// and records FD/strict state via MCP. No platform state is written here.
+return {
+  workflow: 'testarch-nfr',
+  target: story,
+  status: report && report.valid ? 'nfr-assessed' : 'gaps-found',
+  overallRisk: (aggregate && aggregate.overallRisk) || computedOverall,
+  overallStatus: (report && report.overallStatus) || 'CONCERNS',
+  riskBreakdown: (aggregate && aggregate.riskBreakdown) || {
+    security: domainRisks[0],
+    performance: domainRisks[1],
+    reliability: domainRisks[2],
+    scalability: domainRisks[3],
+  },
+  reportPath: (report && report.reportPath) || '',
+  blockers: (report && report.blockers) || [],
+  nextWorkflow: (report && report.nextWorkflow) || 'trace',
+  steps: 6,
+  needsHumanGate: true,
+  result: report,
+}

package/install/templates/.claude/workflows/testarch-test-design.js

@@ -0,0 +1,293 @@
+export const meta = {
+  name: 'testarch-test-design',
+  description: 'Native port of the BYAN testarch-test-design workflow (create mode): a dual-mode (system-level vs epic-level) test-design engine that detects mode, loads context + TEA knowledge fragments, performs testability + risk assessment, designs a coverage/priority/execution plan, then generates the output document(s) and validates them against the checklist. Returns a structured verdict for the orchestrating skill to present at the human gate.',
+  phases: [
+    { title: 'DETECT-MODE', detail: 'system-level vs epic-level + prerequisite check' },
+    { title: 'LOAD-CONTEXT', detail: 'load artifacts, config flags, TEA knowledge fragments' },
+    { title: 'RISK-AND-TESTABILITY', detail: 'testability review (system-level) + risk matrix (all modes)' },
+    { title: 'COVERAGE-PLAN', detail: 'coverage matrix, priorities, execution strategy, estimates, gates' },
+    { title: 'GENERATE-OUTPUT', detail: 'write template(s) and validate against checklist' },
+  ],
+}
+
+// ---------------------------------------------------------------------------
+// FD / STRICT STATE CONTRACT  (re-asserted inline — enforcement-bridge).
+//
+// The in-CLI Workflow tool runs this script OUTSIDE the conversation turn, so
+// BYAN's main-thread hooks (fd-phase-guard, strict-scope-guard, strict-stop-
+// guard, mantra-validate) DO NOT fire here. This script therefore:
+//   - NEVER imports/requires _byan/.../lib/fd-state.js and NEVER writes
+//     fd-state.json directly  (enforced by byan-lint-workflows.js).
+//   - uses NO wall-clock and NO randomness primitive (wall-clock/RNG
+//     break resume); any date/id arrives via `args`.
+//   - returns DATA only. The orchestrating skill is the human-gated conductor;
+//     IT records FD/strict state via the byan_fd_* / byan_strict_* MCP tools
+//     AT the gate. The mode-detection question and the team-review gate stay
+//     OUT of this script — the script surfaces them as verdict fields.
+// The test-design document(s) ARE the workflow's product, written by the
+// generate leaf — that is the artifact, not BYAN platform state.
+// ---------------------------------------------------------------------------
+
+// Inputs arrive via args (sandbox: no fs, no clock, no RNG). `mode` may force
+// 'system-level' or 'epic-level'; default 'auto-detect' lets the detect leaf
+// resolve it per the source priority order. `date` is passed in for stamping.
+const requestedMode = (args && args.mode) || 'auto-detect'
+const epicNum = (args && args.epicNum) || null
+const designLevel = (args && args.designLevel) || 'full'
+const date = (args && args.date) || null
+
+const SOURCE = '_byan/workflow/simple/testarch/test-design'
+
+// ---- STEP 1 : Detect Mode & Prerequisites (steps-c/step-01-detect-mode.md) --
+phase('DETECT-MODE')
+const DETECT_SCHEMA = {
+  type: 'object',
+  required: ['mode', 'prerequisitesMet'],
+  properties: {
+    mode: { type: 'string', enum: ['system-level', 'epic-level', 'ambiguous'], description: 'resolved mode; "ambiguous" if it cannot be decided without the user' },
+    rationale: { type: 'string', description: 'why this mode (user intent > file-based detection > ask)' },
+    prerequisitesMet: { type: 'boolean', description: 'true only if the mode-specific required inputs are present' },
+    missingInputs: { type: 'array', items: { type: 'string' }, description: 'required inputs that are missing (HALT condition)' },
+    needsUserDecision: { type: 'boolean', description: 'true if mode is ambiguous and the user must choose A/B' },
+  },
+}
+const detect = await agent(
+  `You are the Master Test Architect running step 1 of the BYAN testarch-test-design workflow. ` +
+    `Read the COMPLETE source step ${SOURCE}/steps-c/step-01-detect-mode.md and the workflow.yaml in that folder. ` +
+    `Requested mode: ${JSON.stringify(requestedMode)} (auto-detect means resolve it yourself).\n` +
+    `Mode detection priority order (mirror the source exactly): ` +
+    `(A) explicit user intent — PRD+ADR with no epic/stories => system-level; epic+stories with no PRD/ADR => epic-level; both present => prefer system-level first. ` +
+    `(B) file-based — if {implementation_artifacts}/sprint-status.yaml exists => epic-level, otherwise system-level. ` +
+    `(C) still unclear => mark mode "ambiguous" and needsUserDecision true (do NOT guess).\n` +
+    `Then run the mode-specific prerequisite check: system-level requires PRD (FR+NFR) + ADR/architecture decisions + architecture/tech-spec; ` +
+    `epic-level requires epic/story requirements with acceptance criteria (+ architecture context if available). ` +
+    `If required inputs are missing, set prerequisitesMet false and list them in missingInputs (HALT condition — do not fabricate inputs). ` +
+    `State which mode you will use and why.`,
+  { label: 'detect-mode', phase: 'DETECT-MODE', schema: DETECT_SCHEMA }
+)
+
+// HALT conditions are surfaced to the orchestrating skill, not decided here.
+const blocked = !detect.prerequisitesMet || detect.mode === 'ambiguous'
+if (blocked) {
+  log('detect-mode requires a human decision or missing inputs; halting before context load')
+  return {
+    workflow: 'testarch-test-design',
+    summary: detect.mode === 'ambiguous'
+      ? 'mode ambiguous — user must choose system-level (A) or epic-level (B)'
+      : `prerequisites missing for ${detect.mode} mode`,
+    mode: detect.mode,
+    designLevel,
+    steps: 1,
+    halted: true,
+    needsHumanGate: true,
+    needsUserDecision: detect.mode === 'ambiguous',
+    missingInputs: detect.missingInputs || [],
+    detect,
+  }
+}
+
+const mode = detect.mode
+log(`mode resolved: ${mode} (${detect.rationale || 'no rationale'})`)
+
+// ---- STEP 2 : Load Context & Knowledge Base (steps-c/step-02-load-context) --
+phase('LOAD-CONTEXT')
+const LOAD_SCHEMA = {
+  type: 'object',
+  required: ['loaded'],
+  properties: {
+    loaded: { type: 'array', items: { type: 'string' }, description: 'artifacts + knowledge fragments actually loaded' },
+    teaUsePlaywrightUtils: { type: 'boolean', description: 'value of tea_use_playwright_utils from config' },
+    techStack: { type: 'array', items: { type: 'string' } },
+    integrationPoints: { type: 'array', items: { type: 'string' } },
+    nfrs: { type: 'array', items: { type: 'string' }, description: 'performance/security/reliability/compliance NFRs (system-level)' },
+    existingCoverageGaps: { type: 'array', items: { type: 'string' }, description: 'gaps + flaky areas found by scanning tests/ spec e2e api (epic-level)' },
+    missing: { type: 'array', items: { type: 'string' }, description: 'inputs the user still needs to provide' },
+  },
+}
+const context = await agent(
+  `Master Test Architect, step 2 of testarch-test-design (mode: ${mode}). ` +
+    `Read the COMPLETE source step ${SOURCE}/steps-c/step-02-load-context.md.\n` +
+    `1. Load config from _byan/tea/config.yaml: read tea_use_playwright_utils, note output_folder.\n` +
+    `2. Load mode-specific artifacts. system-level: PRD (FR+NFR), ADRs/architecture decisions, architecture/tech-spec, epics for scope; ` +
+    `extract tech stack & dependencies, integration points, NFRs (perf/security/reliability/compliance). ` +
+    `epic-level: epic + story docs with acceptance criteria, PRD if available, architecture/tech-spec if available, prior system-level test-design outputs if available; ` +
+    `extract testable requirements, integration points, known coverage gaps.\n` +
+    `3. epic-level only: scan the repo for existing tests (tests/, spec, e2e, api folders), identify coverage gaps and flaky areas, note existing fixture/test patterns.\n` +
+    `4. Load ONLY the relevant TEA knowledge fragments via {project-root}/_byan/connaissance/testarch/tea-index.csv. ` +
+    `system-level required: adr-quality-readiness-checklist.md, test-levels-framework.md, risk-governance.md, test-quality.md. ` +
+    `epic-level required: risk-governance.md, probability-impact.md, test-levels-framework.md, test-priorities-matrix.md.\n` +
+    `5. Summarize what was loaded; flag anything missing in "missing" (do not invent content).`,
+  { label: 'load-context', phase: 'LOAD-CONTEXT', schema: LOAD_SCHEMA }
+)
+log(`loaded ${context.loaded.length} artifacts/fragments`)
+
+// ---- STEP 3 : Testability & Risk Assessment (steps-c/step-03-...) ----------
+phase('RISK-AND-TESTABILITY')
+const RISK_SCHEMA = {
+  type: 'object',
+  required: ['risks'],
+  properties: {
+    testabilityConcerns: { type: 'array', items: { type: 'string' }, description: 'system-level: actionable controllability/observability/reliability concerns' },
+    testabilityStrengths: { type: 'array', items: { type: 'string' }, description: 'system-level: what is already strong' },
+    asrs: {
+      type: 'array',
+      description: 'Architecturally Significant Requirements (system-level), each ACTIONABLE or FYI',
+      items: {
+        type: 'object',
+        required: ['requirement', 'classification'],
+        properties: {
+          requirement: { type: 'string' },
+          classification: { type: 'string', enum: ['ACTIONABLE', 'FYI'] },
+        },
+      },
+    },
+    risks: {
+      type: 'array',
+      description: 'risk matrix — genuine risks, not features',
+      items: {
+        type: 'object',
+        required: ['id', 'category', 'probability', 'impact', 'score'],
+        properties: {
+          id: { type: 'string', description: 'unique id R-001, R-002, ...' },
+          category: { type: 'string', enum: ['TECH', 'SEC', 'PERF', 'DATA', 'BUS', 'OPS'] },
+          description: { type: 'string' },
+          probability: { type: 'integer', minimum: 1, maximum: 3 },
+          impact: { type: 'integer', minimum: 1, maximum: 3 },
+          score: { type: 'integer', minimum: 1, maximum: 9, description: 'probability * impact' },
+          mitigation: { type: 'string' },
+          owner: { type: 'string' },
+          timeline: { type: 'string' },
+        },
+      },
+    },
+    highRisks: { type: 'array', items: { type: 'string' }, description: 'ids of risks with score >= 6' },
+  },
+}
+const risk = await agent(
+  `Master Test Architect, step 3 of testarch-test-design (mode: ${mode}). ` +
+    `Read the COMPLETE source step ${SOURCE}/steps-c/step-03-risk-and-testability.md. Base every conclusion on evidence from the loaded artifacts.\n` +
+    `1. system-level ONLY: evaluate architecture for Controllability (state seeding, mockability, fault injection), Observability (logs, metrics, traces, deterministic assertions), Reliability (isolation, reproducibility, parallel safety). ` +
+    `Structure: actionable Testability Concerns FIRST, then a Testability Assessment Summary of strengths. ` +
+    `Identify ASRs and mark each ACTIONABLE or FYI.\n` +
+    `2. ALL modes: risk assessment using risk-governance.md + probability-impact.md. Identify GENUINE risks (not features). ` +
+    `Classify by category TECH/SEC/PERF/DATA/BUS/OPS. Score Probability (1-3) and Impact (1-3). Compute Risk Score = P*I. ` +
+    `Flag high risks (score >= 6). Define mitigation, owner, timeline for high risks.\n` +
+    `3. Summarize the highest risks and their mitigation priorities.`,
+  { label: 'risk-and-testability', phase: 'RISK-AND-TESTABILITY', schema: RISK_SCHEMA }
+)
+log(`${risk.risks.length} risks; ${(risk.highRisks || []).length} high (>=6)`)
+
+// ---- STEP 4 : Coverage Plan & Execution Strategy (steps-c/step-04-...) ------
+phase('COVERAGE-PLAN')
+const COVERAGE_SCHEMA = {
+  type: 'object',
+  required: ['scenarios', 'qualityGates'],
+  properties: {
+    scenarios: {
+      type: 'array',
+      description: 'coverage matrix — atomic scenarios, one test level each, no duplicate coverage',
+      items: {
+        type: 'object',
+        required: ['testId', 'testLevel', 'priority'],
+        properties: {
+          testId: { type: 'string' },
+          requirement: { type: 'string' },
+          testLevel: { type: 'string', enum: ['E2E', 'API', 'Component', 'Unit'] },
+          priority: { type: 'string', enum: ['P0', 'P1', 'P2', 'P3'] },
+          riskLink: { type: 'string', description: 'risk id this scenario covers' },
+          notes: { type: 'string' },
+        },
+      },
+    },
+    executionStrategy: {
+      type: 'object',
+      description: 'simple PR / Nightly / Weekly model — no smoke/P0/P1/P2 tier structure, no re-listing tests',
+      properties: {
+        pr: { type: 'string' },
+        nightly: { type: 'string' },
+        weekly: { type: 'string' },
+      },
+    },
+    resourceEstimates: {
+      type: 'object',
+      description: 'interval ranges only, no false precision',
+      properties: {
+        p0: { type: 'string' },
+        p1: { type: 'string' },
+        p2: { type: 'string' },
+        p3: { type: 'string' },
+        total: { type: 'string' },
+        timeline: { type: 'string' },
+      },
+    },
+    qualityGates: {
+      type: 'object',
+      required: ['p0PassRate'],
+      properties: {
+        p0PassRate: { type: 'string', description: '100%' },
+        p1PassRate: { type: 'string', description: '>= 95%' },
+        highRiskMitigation: { type: 'string', description: 'complete before release' },
+        coverageTarget: { type: 'string', description: '>= 80% unless justified' },
+      },
+    },
+  },
+}
+const coverage = await agent(
+  `Master Test Architect, step 4 of testarch-test-design (mode: ${mode}, design level: ${designLevel}). ` +
+    `Read the COMPLETE source step ${SOURCE}/steps-c/step-04-coverage-plan.md. Avoid redundant coverage across test levels.\n` +
+    `1. Coverage matrix: for each requirement or risk-driven scenario, decompose into atomic scenarios; select ONE test level (E2E/API/Component/Unit) via test-levels-framework.md; no duplicate coverage; assign priority P0-P3 via test-priorities-matrix.md. ` +
+    `Priority rules: P0 = blocks core functionality + high risk + no workaround; P1 = critical paths + medium/high risk; P2 = secondary flows + low/medium risk; P3 = nice-to-have/exploratory/benchmarks. Link scenarios to risk ids from step 3.\n` +
+    `2. Execution strategy = simple PR / Nightly / Weekly. PR = all functional tests if <15 min; Nightly/Weekly = long-running/expensive (perf, chaos, large datasets). Do NOT re-list all tests; do NOT build a smoke/P0/P1/P2 tier structure.\n` +
+    `3. Resource estimates as INTERVAL RANGES only (e.g. P0 ~25-40h) — no false precision; total + timeline as ranges.\n` +
+    `4. Quality gates: P0 pass rate = 100%, P1 pass rate >= 95%, high-risk mitigations complete before release, coverage target >= 80% (adjust only if justified).`,
+  { label: 'coverage-plan', phase: 'COVERAGE-PLAN', schema: COVERAGE_SCHEMA }
+)
+log(`${coverage.scenarios.length} scenarios planned`)
+
+// ---- STEP 5 : Generate Outputs & Validate (steps-c/step-05-...) ------------
+phase('GENERATE-OUTPUT')
+const GENERATE_SCHEMA = {
+  type: 'object',
+  required: ['outputs', 'checklistPass'],
+  properties: {
+    outputs: { type: 'array', items: { type: 'string' }, description: 'output file path(s) written' },
+    templateUsed: { type: 'array', items: { type: 'string' }, description: 'template file(s) used' },
+    checklistPass: { type: 'boolean', description: 'true only if all checklist.md criteria are satisfied' },
+    checklistFailures: { type: 'array', items: { type: 'string' }, description: 'checklist items not satisfied (fix before completion)' },
+    openAssumptions: { type: 'array', items: { type: 'string' } },
+    needsEpicNum: { type: 'boolean', description: 'epic-level only: true if epic_num was unclear and must be asked' },
+  },
+}
+const generate = await agent(
+  `Master Test Architect, step 5 of testarch-test-design (mode: ${mode}). ` +
+    `Read the COMPLETE source step ${SOURCE}/steps-c/step-05-generate-output.md and the checklist ${SOURCE}/checklist.md.\n` +
+    `1. Select template(s) by mode. system-level (Phase 3) => TWO documents: {output_folder}/test-design-architecture.md from test-design-architecture-template.md, and {output_folder}/test-design-qa.md from test-design-qa-template.md. ` +
+    `epic-level (Phase 4) => ONE document: {output_folder}/test-design-epic-${epicNum || '{epic_num}'}.md from test-design-template.md` +
+    `${mode === 'epic-level' && !epicNum ? ' — epic_num is NOT provided, set needsEpicNum true and do NOT invent it' : ''}.\n` +
+    `2. Populate the template(s) using the step-3 risk matrix and step-4 coverage plan: risk assessment matrix, coverage matrix + priorities, execution strategy, resource estimates (ranges), quality-gate criteria, and any mode-specific sections the template requires. ` +
+    `tea_use_playwright_utils=${context.teaUsePlaywrightUtils === undefined ? 'unknown' : context.teaUsePlaywrightUtils} (drive the QA-doc code-example imports accordingly).\n` +
+    `3. Validate the output(s) against checklist.md item-by-item. If any criterion is unmet, FIX it before reporting; list any residual failures in checklistFailures and set checklistPass false.\n` +
+    `4. Completion report: mode used, output file path(s), key risks + gate thresholds, open assumptions.`,
+  { label: 'generate-output', phase: 'GENERATE-OUTPUT', schema: GENERATE_SCHEMA }
+)
+log(`outputs: ${(generate.outputs || []).join(', ') || 'none'} | checklist ${generate.checklistPass ? 'PASS' : 'FAIL'}`)
+
+// Single top-level return — DATA only. The team-review gate (checklist
+// Post-Workflow Actions) is the human decision and stays with the skill.
+return {
+  workflow: 'testarch-test-design',
+  summary: `test design (${mode}) — ${coverage.scenarios.length} scenarios, ${risk.risks.length} risks, checklist ${generate.checklistPass ? 'PASS' : 'FAIL'}`,
+  mode,
+  designLevel,
+  epicNum,
+  date,
+  steps: 5,
+  halted: false,
+  needsHumanGate: true,
+  checklistPass: generate.checklistPass,
+  outputs: generate.outputs || [],
+  highRisks: risk.highRisks || [],
+  needsEpicNum: !!generate.needsEpicNum,
+  checklistFailures: generate.checklistFailures || [],
+  result: { detect, context, risk, coverage, generate },
+}