@winspan/claude-forge 8.53.2 → 8.54.4

package/scripts/e2e-user-methodology.ts

@@ -1,326 +0,0 @@
-#!/usr/bin/env tsx
-/**
- * E2E 验证：~/.claude-forge/methodologies.yaml → AgentRegistry
- *
- * 测试策略：
- *   A. 纯 loader 测试（不写磁盘）
- *      - 用 configPath 参数指向临时文件，覆盖各种合法/非法场景
- *      - 跨方法论组合、覆盖 official preset 的 warning、异常 yaml 隔离
- *
- *   B. AgentRegistry 集成（写入真实 ~/.claude-forge/methodologies.yaml，做备份/恢复）
- *      - 因 AgentRegistry 使用 homedir() 固化在 user-config-loader 内
- *      - 执行前备份旧文件，执行后无论成败都恢复
- */
-
-import fs from 'node:fs';
-import { mkdtempSync, rmSync, writeFileSync, mkdirSync, readFileSync, existsSync } from 'node:fs';
-import { tmpdir, homedir } from 'node:os';
-import { join } from 'node:path';
-
-import { loadUserMethodologyPresets } from '../src/agents/methodologies/user-config-loader.js';
-import { buildMethodologyAgent } from '../src/agents/methodologies/agent-builder.js';
-import { AgentRegistry } from '../src/agents/registry.js';
-
-type CheckResult = { name: string; pass: boolean; detail?: string };
-const results: CheckResult[] = [];
-
-function check(name: string, pass: boolean, detail?: string): void {
-  results.push({ name, pass, detail });
-  const mark = pass ? 'PASS' : 'FAIL';
-  console.log(`[${mark}] ${name}${detail ? ` — ${detail}` : ''}`);
-}
-
-console.log('\n=== User Methodology E2E ===\n');
-
-// ── Part A: loader 直接测试 ───────────────────────────────
-const tmp = mkdtempSync(join(tmpdir(), 'forge-e2e-methodology-'));
-
-try {
-  console.log('--- Part A: loader (with configPath override) ---\n');
-
-  // Case 1: 合法 presets（纯 BMAD、纯 Harness、跨方法论、覆盖 official）
-  const goodYaml = `
-schemaVersion: '1.0'
-agents:
-  - id: e2e-test-quick
-    description: 'E2E 测试 - 快速版'
-    phases:
-      - bmad/implement
-      - bmad/test
-    tools: [Read, Write, Bash]
-    tags: [e2e, test]
-
-  - id: e2e-test-mixed
-    description: 'E2E 测试 - 混合方法论'
-    phases:
-      - harness/safety-net
-      - bmad/implement
-      - bmad/test
-    tools: [Read, Write, Edit, Bash]
-
-  # 覆盖 official bmad-feature-quick
-  - id: bmad-feature-quick
-    description: 'E2E 测试 - 覆盖 official'
-    phases:
-      - bmad/implement
-`;
-  const goodPath = join(tmp, 'good.yaml');
-  writeFileSync(goodPath, goodYaml);
-  const good = loadUserMethodologyPresets(goodPath);
-  console.log(`loaded ${good.presets.length} presets, ${good.warnings.length} warnings`);
-
-  check('A1. all 3 valid presets loaded', good.presets.length === 3, `got=${good.presets.length}`);
-  check('A1b. zero warnings for good yaml', good.warnings.length === 0, good.warnings.join(' / '));
-
-  const quick = good.presets.find(p => p.id === 'e2e-test-quick');
-  check('A2. e2e-test-quick loaded', !!quick, quick?.phases.join(', '));
-  check(
-    'A2b. tools passthrough correct',
-    JSON.stringify(quick?.tools) === JSON.stringify(['Read', 'Write', 'Bash']),
-  );
-  check(
-    'A2c. tags passthrough correct',
-    JSON.stringify(quick?.tags) === JSON.stringify(['e2e', 'test']),
-  );
-
-  const mixed = good.presets.find(p => p.id === 'e2e-test-mixed');
-  check('A3. cross-methodology preset loaded', !!mixed);
-  check(
-    'A3b. phases preserve user order across workflows',
-    JSON.stringify(mixed?.phases) === JSON.stringify(['harness/safety-net', 'bmad/implement', 'bmad/test']),
-  );
-
-  const override = good.presets.find(p => p.id === 'bmad-feature-quick');
-  check('A4. override of official preset loaded (loader level)', !!override);
-  check(
-    'A4b. override carries user-provided description',
-    override?.description === 'E2E 测试 - 覆盖 official',
-  );
-
-  // Case 2: buildMethodologyAgent 能生成 prompt
-  if (mixed) {
-    const built = buildMethodologyAgent(mixed);
-    check('A5. buildMethodologyAgent produces content', built.content.length > 100);
-    check(
-      'A5b. built.name matches preset.id',
-      built.name === 'e2e-test-mixed',
-    );
-    check('A5c. skills aggregated from phases', Array.isArray(built.skills) && built.skills.length > 0);
-    check('A5d. mixed workflow labeled as MIXED', built.content.includes('MIXED'));
-    check(
-      'A5e. phases enumerated in content',
-      built.content.includes('safety-net') && built.content.includes('implement'),
-    );
-  }
-
-  // Case 3: 异常 yaml 不影响其他 preset
-  const mixedBadYaml = `
-schemaVersion: '1.0'
-agents:
-  # 缺 id
-  - description: 'no id'
-    phases: [bmad/implement]
-
-  # phases 非数组
-  - id: bad-phases
-    phases: "not-an-array"
-
-  # 未知 phase
-  - id: bad-phase-id
-    phases: [bmad/nonexistent-phase]
-
-  # 大写 id 不合法
-  - id: BadCaseId
-    phases: [bmad/implement]
-
-  # duplicate id（与后续同名）
-  - id: dup-agent
-    phases: [bmad/implement]
-  - id: dup-agent
-    phases: [bmad/test]
-
-  # 合法 preset，应该保留
-  - id: survivor
-    description: '异常邻居里的幸存者'
-    phases:
-      - bmad/review
-`;
-  const badPath = join(tmp, 'mixed-bad.yaml');
-  writeFileSync(badPath, mixedBadYaml);
-  const bad = loadUserMethodologyPresets(badPath);
-
-  console.log(`\nmixed-bad: ${bad.presets.length} presets, ${bad.warnings.length} warnings`);
-  for (const w of bad.warnings) console.log(`  [warn] ${w}`);
-
-  check(
-    'A6. invalid entries skipped, valid ones survive',
-    bad.presets.length === 2 && bad.presets.some(p => p.id === 'survivor') && bad.presets.some(p => p.id === 'dup-agent'),
-    `ids=${bad.presets.map(p => p.id).join(',')}`,
-  );
-  check('A6b. multiple warnings generated', bad.warnings.length >= 4, `warnings=${bad.warnings.length}`);
-  check(
-    'A6c. duplicate-id warning present',
-    bad.warnings.some(w => w.includes('duplicate id')),
-  );
-  check(
-    'A6d. unknown phase warning present',
-    bad.warnings.some(w => w.includes('phase') && w.includes('not found')),
-  );
-
-  // Case 4: 缺失文件返回空
-  const emptyResult = loadUserMethodologyPresets(join(tmp, 'nonexistent.yaml'));
-  check(
-    'A7. missing file returns empty result without warnings',
-    emptyResult.presets.length === 0 && emptyResult.warnings.length === 0,
-  );
-
-  // Case 5: 解析错误
-  const corruptYaml = `
-schemaVersion: '1.0'
-agents:
-  - id: test
-    phases: [bmad/implement
-  # unclosed bracket
-`;
-  const corruptPath = join(tmp, 'corrupt.yaml');
-  writeFileSync(corruptPath, corruptYaml);
-  const corrupt = loadUserMethodologyPresets(corruptPath);
-  check(
-    'A8. parse error produces warning, not crash',
-    corrupt.presets.length === 0 && corrupt.warnings.length >= 1,
-    `warnings=${corrupt.warnings.length}`,
-  );
-} finally {
-  rmSync(tmp, { recursive: true, force: true });
-}
-
-// ── Part B: AgentRegistry 集成 ────────────────────────────
-console.log('\n--- Part B: AgentRegistry integration (writes to real ~/.claude-forge) ---\n');
-
-const realDir = join(homedir(), '.claude-forge');
-const realPath = join(realDir, 'methodologies.yaml');
-const backupPath = `${realPath}.e2e-backup-${Date.now()}`;
-
-let backupCreated = false;
-let dirCreated = false;
-try {
-  if (!existsSync(realDir)) {
-    mkdirSync(realDir, { recursive: true });
-    dirCreated = true;
-  }
-  if (existsSync(realPath)) {
-    const original = readFileSync(realPath, 'utf-8');
-    writeFileSync(backupPath, original);
-    backupCreated = true;
-    console.log(`(backed up existing methodologies.yaml → ${backupPath})`);
-  }
-
-  const registryYaml = `
-schemaVersion: '1.0'
-agents:
-  - id: e2e-registry-quick
-    description: 'AgentRegistry 集成测试 - 快速版'
-    phases: [bmad/implement, bmad/test]
-    tools: [Read, Write, Bash]
-
-  - id: e2e-registry-mixed
-    description: 'AgentRegistry 集成测试 - 跨方法论'
-    phases:
-      - harness/safety-net
-      - bmad/implement
-      - bmad/test
-
-  # 覆盖 official
-  - id: bmad-feature-quick
-    description: '[e2e override] bmad-feature-quick'
-    phases: [bmad/implement]
-`;
-  writeFileSync(realPath, registryYaml);
-
-  const registry = new AgentRegistry();
-  const all = registry.getAll();
-
-  console.log(`AgentRegistry loaded ${all.length} agents`);
-
-  const quick = registry.get('e2e-registry-quick');
-  check('B1. custom agent registered in AgentRegistry', !!quick);
-  check(
-    'B1b. custom agent has source=user-methodology',
-    quick?.source === 'user-methodology',
-    `source=${quick?.source}`,
-  );
-  // buildPromptContent 使用 phase.template.id（单段），而不是 `bmad/implement` 全称。
-  // 内容检查：有 "BMAD" 标签 + phase 的 title/id。
-  check(
-    'B1c. custom agent has methodology content injected',
-    !!quick &&
-      quick.content.includes('BMAD') &&
-      quick.content.includes('implement') &&
-      quick.content.includes('test'),
-  );
-
-  const mixed = registry.get('e2e-registry-mixed');
-  check('B2. cross-methodology agent registered', !!mixed);
-  check(
-    'B2b. mixed-workflow content labeled MIXED',
-    !!mixed && mixed.content.includes('MIXED'),
-  );
-
-  const override = registry.get('bmad-feature-quick');
-  check(
-    'B3. official preset overridden by user',
-    override?.source === 'user-methodology',
-    `source=${override?.source}`,
-  );
-  check(
-    'B3b. overridden description matches user entry',
-    override?.description === '[e2e override] bmad-feature-quick',
-  );
-
-  // tools default expansion
-  check(
-    'B4. tools default applied when omitted',
-    Array.isArray(mixed?.tools) && mixed!.tools.length >= 5,
-    `tools=${mixed?.tools?.join(',')}`,
-  );
-
-  // Other official agents still present
-  const bmadFull = registry.get('bmad-feature-full');
-  check('B5. un-overridden official preset still present', !!bmadFull);
-  check(
-    'B5b. un-overridden agent has source=builtin',
-    bmadFull?.source === 'builtin',
-    `source=${bmadFull?.source}`,
-  );
-} finally {
-  // 恢复
-  if (backupCreated) {
-    const original = readFileSync(backupPath, 'utf-8');
-    writeFileSync(realPath, original);
-    rmSync(backupPath, { force: true });
-    console.log('(restored original methodologies.yaml)');
-  } else if (existsSync(realPath)) {
-    rmSync(realPath, { force: true });
-    console.log('(removed test methodologies.yaml)');
-  }
-  // 目录只清理我们刚建的那层
-  if (dirCreated && existsSync(realDir)) {
-    try {
-      fs.rmdirSync(realDir);
-    } catch {
-      // 已有别的文件，留着
-    }
-  }
-}
-
-const passed = results.filter(r => r.pass).length;
-const failed = results.filter(r => !r.pass).length;
-console.log(`\n=== Methodology E2E Summary: ${passed} passed, ${failed} failed ===`);
-if (failed > 0) {
-  console.log('\nFailures:');
-  for (const r of results.filter(x => !x.pass)) {
-    console.log(`  - ${r.name}${r.detail ? ` | ${r.detail}` : ''}`);
-  }
-  process.exit(1);
-}
-console.log('\n=== Result: PASS ===');

package/scripts/e2e-web-workflows.ts

@@ -1,299 +0,0 @@
-#!/usr/bin/env tsx
-/**
- * E2E 验证：Web UI 工作流 API 全链路
- *
- * 步骤：
- *   1. 启动独立 WebServer（端口 0 自动分配，临时 SQLite）
- *   2. 通过 storage 写入测试数据（模拟完整跑完 BMAD + Harness 流程）
- *   3. 调用 /api/workflows 验证列表聚合
- *   4. 调用 /api/workflows/:id 验证详情
- *   5. 验证过滤、排序、404、鉴权（GET 不需要 token）
- *
- * 注意：GET 接口在 server.ts 中被显式 next()（见 setupRoutes），
- * 所以本脚本不需要写 daemon.token。
- */
-
-import { mkdtempSync, rmSync } from 'node:fs';
-import { tmpdir } from 'node:os';
-import { join } from 'node:path';
-import { randomUUID } from 'node:crypto';
-import type { AddressInfo } from 'node:net';
-import type express from 'express';
-
-import { SQLiteStorage } from '../src/core/storage/sqlite.js';
-import { RuleEngine } from '../src/engine/rule-engine.js';
-import { WebServer } from '../src/web/server.js';
-
-type CheckResult = { name: string; pass: boolean; detail?: string };
-const results: CheckResult[] = [];
-
-function check(name: string, pass: boolean, detail?: string): void {
-  results.push({ name, pass, detail });
-  const mark = pass ? 'PASS' : 'FAIL';
-  console.log(`[${mark}] ${name}${detail ? ` — ${detail}` : ''}`);
-}
-
-const tmp = mkdtempSync(join(tmpdir(), 'forge-e2e-web-'));
-const dbPath = join(tmp, 'data.db');
-
-const storage = new SQLiteStorage(dbPath);
-const ruleEngine = new RuleEngine();
-
-const server = new WebServer({ port: 0, storage, ruleEngine }) as unknown as {
-  app: express.Application;
-  server: ReturnType<express.Application['listen']> | null;
-  stop(): Promise<void>;
-};
-
-let port = 0;
-await new Promise<void>((resolve, reject) => {
-  const app = server.app;
-  const handle = app.listen(0, () => {
-    const addr = handle.address() as AddressInfo;
-    port = addr.port;
-    server.server = handle;
-    resolve();
-  });
-  handle.on('error', reject);
-});
-
-// ── 种子数据 ────────────────────────────────────────────────
-const bmadSession = 'e2e-bmad-session';
-const bmadRoute = 'e2e-bmad-route';
-const bmadSlug = 'e2e-test-feature';
-const bmadPhases = ['analyze', 'design', 'implement', 'test', 'review'] as const;
-const bmadSkill: Record<string, string> = {
-  analyze: 'official-spec-driven-design',
-  design: 'official-architecture-decision',
-  implement: 'official-tdd',
-  test: 'official-tdd',
-  review: 'official-pr-review',
-};
-const bmadArtifact: Record<string, string | null> = {
-  analyze: `docs/specs/${bmadSlug}.md`,
-  design: `docs/design/${bmadSlug}.md`,
-  implement: null,
-  test: null,
-  review: `docs/reviews/${bmadSlug}.md`,
-};
-
-const nowStart = Date.now();
-bmadPhases.forEach((phase, idx) => {
-  storage.writeSkillInvocation({
-    id: randomUUID(),
-    route_request_id: bmadRoute,
-    session_id: bmadSession,
-    agent_id: 'bmad-feature-full',
-    skill_id: bmadSkill[phase],
-    invocation_type: 'dynamic',
-    reason: `bmad/${phase}:${bmadSlug}`,
-    workflow: 'bmad',
-    phase,
-    feature_slug: bmadSlug,
-    artifact_path: bmadArtifact[phase],
-    depth: 0,
-    success: 1,
-    error: null,
-    timestamp: nowStart + idx * 1000,
-  });
-});
-
-// Harness 流程
-const harnessSession = 'e2e-harness-session';
-const harnessRoute = 'e2e-harness-route';
-const harnessPhases = ['reproduce', 'safety-net', 'root-cause'] as const;
-const harnessBase = nowStart + 10_000;
-harnessPhases.forEach((phase, idx) => {
-  storage.writeSkillInvocation({
-    id: randomUUID(),
-    route_request_id: harnessRoute,
-    session_id: harnessSession,
-    agent_id: 'harness-debug-full',
-    skill_id: 'official-debug',
-    invocation_type: 'dynamic',
-    reason: `harness/${phase}:e2e-bug`,
-    workflow: 'harness',
-    phase,
-    feature_slug: 'e2e-bug',
-    artifact_path: null,
-    depth: 0,
-    success: 1,
-    error: null,
-    timestamp: harnessBase + idx * 500,
-  });
-});
-
-// 非工作流型调用（workflow=null）— 验证聚合会忽略
-storage.writeSkillInvocation({
-  id: randomUUID(),
-  route_request_id: 'plain-route-ignored',
-  session_id: bmadSession,
-  agent_id: null,
-  skill_id: 'official-debug',
-  invocation_type: 'dynamic',
-  reason: null,
-  workflow: null,
-  phase: null,
-  feature_slug: null,
-  artifact_path: null,
-  depth: 0,
-  success: 1,
-  error: null,
-  timestamp: nowStart - 5000,
-});
-
-// ── 工具函数 ────────────────────────────────────────────────
-async function fetchJson(path: string, init?: RequestInit): Promise<{ status: number; body: unknown }> {
-  const r = await fetch(`http://127.0.0.1:${port}${path}`, init);
-  let body: unknown = null;
-  try {
-    body = await r.json();
-  } catch {
-    body = null;
-  }
-  return { status: r.status, body };
-}
-
-console.log('\n=== Web API E2E ===\n');
-console.log(`Temp DB: ${dbPath}`);
-console.log(`Server port: ${port}\n`);
-
-try {
-  // ── Test 1: 列表 API 基础 ─────────────────────────────────
-  const list = (await fetchJson('/api/workflows')).body as {
-    total: number;
-    items: Array<{
-      route_request_id: string;
-      workflow: string;
-      phases_completed: string[];
-      feature_slug: string | null;
-      total_invocations: number;
-      success_count: number;
-      first_at: number;
-      last_at: number;
-    }>;
-  };
-
-  console.log(`/api/workflows returned total=${list.total}`);
-  console.log(`  items: ${list.items.map(i => `${i.workflow}/${i.feature_slug ?? '(null)'}`).join(', ')}`);
-
-  check('1. /api/workflows returns list', list.total === 2, `total=${list.total} expected 2`);
-  check(
-    '2. BMAD and Harness both aggregated',
-    !!list.items.find(i => i.workflow === 'bmad') && !!list.items.find(i => i.workflow === 'harness'),
-  );
-
-  const bmad = list.items.find(i => i.workflow === 'bmad');
-  const harness = list.items.find(i => i.workflow === 'harness');
-
-  check(
-    '3. phases_completed is time-sorted, deduped (BMAD 5 phases)',
-    JSON.stringify(bmad?.phases_completed) === JSON.stringify(['analyze', 'design', 'implement', 'test', 'review']),
-    JSON.stringify(bmad?.phases_completed),
-  );
-  check('4. feature_slug parsed correctly', bmad?.feature_slug === bmadSlug, `got=${bmad?.feature_slug}`);
-  check('5. total_invocations counts only workflow rows', bmad?.total_invocations === 5, `got=${bmad?.total_invocations}`);
-  check('6. success_count aggregates correctly', bmad?.success_count === 5, `got=${bmad?.success_count}`);
-  check(
-    '7. non-workflow row (workflow=null) is excluded',
-    !list.items.find(i => i.route_request_id === 'plain-route-ignored'),
-  );
-
-  // ── Test 2: 详情 API ────────────────────────────────────
-  const detail = (await fetchJson(`/api/workflows/${encodeURIComponent(bmadRoute)}`)).body as {
-    route_request_id: string;
-    workflow: string;
-    feature_slug: string | null;
-    invocations: Array<{
-      phase: string | null;
-      timestamp: number;
-      success: boolean;
-      artifact_path: string | null;
-      skill_id: string;
-    }>;
-  };
-
-  console.log(`\n/api/workflows/${bmadRoute} returned workflow=${detail.workflow}, invocations=${detail.invocations.length}`);
-
-  check('8. detail returns the workflow', detail.workflow === 'bmad' && detail.feature_slug === bmadSlug);
-  check('9. detail.invocations matches expected length', detail.invocations.length === 5, `got=${detail.invocations.length}`);
-
-  const timestamps = detail.invocations.map(i => i.timestamp);
-  const sortedTs = [...timestamps].sort((a, b) => a - b);
-  check(
-    '10. invocations sorted by timestamp ascending',
-    JSON.stringify(timestamps) === JSON.stringify(sortedTs),
-  );
-
-  // artifact_path: analyze → design → implement(null) → test(null) → review
-  const byPhase = new Map(detail.invocations.map(i => [i.phase, i]));
-  check(
-    '11. artifact_path preserved (analyze)',
-    byPhase.get('analyze')?.artifact_path === `docs/specs/${bmadSlug}.md`,
-    byPhase.get('analyze')?.artifact_path ?? '(null)',
-  );
-  check(
-    '12. artifact_path null preserved (implement)',
-    byPhase.get('implement')?.artifact_path === null,
-  );
-  check('13. success converted to boolean', typeof detail.invocations[0].success === 'boolean');
-
-  // ── Test 3: workflow query 过滤 ─────────────────────────
-  const bmadOnly = (await fetchJson('/api/workflows?workflow=bmad')).body as { total: number; items: Array<{ workflow: string }> };
-  check(
-    '14. workflow=bmad filter returns only bmad',
-    bmadOnly.total === 1 && bmadOnly.items.every(i => i.workflow === 'bmad'),
-    `total=${bmadOnly.total}`,
-  );
-
-  const harnessOnly = (await fetchJson('/api/workflows?workflow=harness')).body as { total: number; items: Array<{ workflow: string }> };
-  check('15. workflow=harness filter', harnessOnly.total === 1, `total=${harnessOnly.total}`);
-
-  // ── Test 4: session_id query 过滤 ───────────────────────
-  const bySession = (await fetchJson(`/api/workflows?session_id=${bmadSession}`)).body as { total: number };
-  check('16. session_id filter works', bySession.total === 1, `total=${bySession.total}`);
-
-  // ── Test 5: 404 ──────────────────────────────────────────
-  const notFound = await fetchJson('/api/workflows/nonexistent-route-id-xyz');
-  check('17. Unknown route returns 404', notFound.status === 404, `status=${notFound.status}`);
-  check(
-    '18. 404 body has error field',
-    typeof (notFound.body as { error?: string })?.error === 'string',
-  );
-
-  // ── Test 6: GET 不需要 token ────────────────────────────
-  const noAuth = await fetchJson('/api/workflows');
-  check('19. GET /api/workflows works without Authorization header', noAuth.status === 200);
-
-  // ── Test 7: 非 GET 请求（如果有写接口）应该受限 ──────────
-  // /api/workflows 目前没有写接口，改为验证 /api/routing POST 等需要 token。
-  // 这里仅做 sanity：在没有 daemon.token 文件时，require-auth 会 next() 放行（早期启动保护），
-  // 所以不强求 401。仅打印信息。
-  console.log('\n(auth middleware note: POST-type workflows endpoints do not exist; skipping 401 check)');
-
-  // ── Test 8: 空列表（用新 storage）───────────────────────
-  // 由于主 storage 已有数据，验证 limit=0 不会炸。
-  const limitZero = (await fetchJson('/api/workflows?limit=0')).body as { total: number };
-  check(
-    '20. invalid limit (0) falls back to default (no crash)',
-    typeof limitZero.total === 'number',
-  );
-} finally {
-  if (server.server) {
-    await new Promise<void>(res => server.server!.close(() => res()));
-  }
-  storage.close();
-  rmSync(tmp, { recursive: true, force: true });
-}
-
-const passed = results.filter(r => r.pass).length;
-const failed = results.filter(r => !r.pass).length;
-console.log(`\n=== Web E2E Summary: ${passed} passed, ${failed} failed ===`);
-if (failed > 0) {
-  console.log('\nFailures:');
-  for (const r of results.filter(x => !x.pass)) {
-    console.log(`  - ${r.name}${r.detail ? ` | ${r.detail}` : ''}`);
-  }
-  process.exit(1);
-}
-console.log('\n=== Result: PASS ===');