qe-quality-assessment

QE 质量评估

用途

指导使用 v3 的质量评估功能，包括自动化质量门禁、指标汇总、趋势分析和部署就绪情况评估。

激活时机

• 评估代码质量时
• 设置质量门禁时
• 评估部署就绪情况时
• 跟踪质量指标时
• 生成质量报告时

快速入门

# Run quality assessment
aqe quality assess --scope src/ --gates all

# Check deployment readiness
aqe quality deploy-ready --environment production

# Generate quality report
aqe quality report --format dashboard --period 30d

# Compare quality between releases
aqe quality compare --from v1.0 --to v2.0

Agent 工作流

// Comprehensive quality assessment
Task("Assess code quality", `
  Evaluate quality for src/:
  - Code complexity (cyclomatic, cognitive)
  - Test coverage and mutation score
  - Security vulnerabilities
  - Code smells and technical debt
  - Documentation coverage
  Generate quality score and recommendations.
`, "qe-quality-analyzer")

// Deployment readiness check
Task("Check deployment readiness", `
  Evaluate if release v2.1.0 is ready for production:
  - All tests passing
  - Coverage thresholds met
  - No critical vulnerabilities
  - Performance benchmarks passed
  - Documentation updated
  Provide go/no-go recommendation.
`, "qe-deployment-advisor")

质量维度

1. 代码质量指标

await qualityAnalyzer.assessCode({
  scope: 'src/**/*.ts',
  metrics: {
    complexity: {
      cyclomatic: { max: 15, warn: 10 },
      cognitive: { max: 20, warn: 15 }
    },
    maintainability: {
      index: { min: 65 },
      duplication: { max: 3 }  // percent
    },
    documentation: {
      publicAPIs: { min: 80 },
      complexity: { min: 70 }
    }
  }
});

2. 质量门禁

await qualityGate.evaluate({
  gates: {
    coverage: { min: 80, blocking: true },
    // Fault detection — coverage is necessary but NOT sufficient (ADR-113).
    // A suite can hit 90% coverage and catch no bugs; mutation score does not lie.
    mutationScore: { min: 0.6, blocking: false },   // warn-by-default; opt-in blocking
    regenerability: { min: 0.5, blocking: false },  // Deletion Test: durable oracle backing per module
    complexity: { max: 15, blocking: false },
    vulnerabilities: { critical: 0, high: 0, blocking: true },
    duplications: { max: 3, blocking: false },
    techDebt: { maxRatio: 5, blocking: false }
  },
  action: {
    onPass: 'proceed',
    onFail: 'block-merge',
    onWarn: 'notify'
  }
});

2a. 可再生成性门禁（ADR-113）

覆盖率衡量执行过的代码行；变异分数衡量测试能否发现错误，而可再生成性回答删除测试：“如果删除并重新生成此模块，错误的重建结果能被发现吗？”可再生成性 = 根据支撑该模块的耐久性层级（durable > live > ephemeral > none）折算后的变异分数；只有 ephemeral 测试的得分较低，因为它们无法经受重新实现。

import { evaluateRegenerabilityGate } from '../../../src/feedback/regenerability-gate.js';

const verdict = evaluateRegenerabilityGate(moduleProfiles, {
  mutationScoreMin: 0.6,
  regenerabilityMin: 0.5,
  mode: 'warn',   // 'block' to fail CI; warn-by-default so adoption never breaks pipelines
});
// verdict.passed (thresholds met) · verdict.blocking (should fail CI) · verdict.failures[]

在报告中将其与覆盖率并列展示：Coverage 88% ✅ / Mutation 41% ⚠️ / Regenerability: ephemeral-only ⚠️。

3. 部署就绪情况

await deploymentAdvisor.assess({
  release: 'v2.1.0',
  criteria: {
    testing: {
      unitTests: 'all-pass',
      integrationTests: 'all-pass',
      e2eTests: 'critical-pass',
      performanceTests: 'baseline-met'
    },
    quality: {
      coverage: 80,
      noNewVulnerabilities: true,
      noRegressions: true
    },
    documentation: {
      changelog: true,
      apiDocs: true,
      releaseNotes: true
    }
  }
});

质量分数计算

quality_score:
  components:
    test_coverage:
      weight: 0.25
      metrics: [statement, branch, function]

    code_quality:
      weight: 0.20
      metrics: [complexity, maintainability, duplication]

    security:
      weight: 0.25
      metrics: [vulnerabilities, dependencies]

    reliability:
      weight: 0.20
      metrics: [bug_density, flaky_tests, error_rate]

    documentation:
      weight: 0.10
      metrics: [api_coverage, readme, changelog]

  scoring:
    A: 90-100
    B: 80-89
    C: 70-79
    D: 60-69
    F: 0-59

质量仪表板

interface QualityDashboard {
  overallScore: number;  // 0-100
  grade: 'A' | 'B' | 'C' | 'D' | 'F';
  dimensions: {
    name: string;
    score: number;
    trend: 'improving' | 'stable' | 'declining';
    issues: Issue[];
  }[];
  gates: {
    name: string;
    status: 'pass' | 'fail' | 'warn';
    value: number;
    threshold: number;
  }[];
  trends: {
    period: string;
    scores: number[];
    alerts: Alert[];
  };
  recommendations: Recommendation[];
}

CI/CD 集成

# Quality gate in pipeline
quality_check:
  stage: verify
  script:
    - aqe quality assess --gates all --output report.json
  rules:
    - if: $CI_PIPELINE_SOURCE == "merge_request_event"
  artifacts:
    reports:
      quality: report.json
  allow_failure:
    exit_codes:
      - 1  # Warnings only

运行历史

每次质量评估后，将结果追加到此技能目录中的 run-history.json：

node -e "
const fs = require('fs');
const h = JSON.parse(fs.readFileSync('.claude/skills/qe-quality-assessment/run-history.json'));
h.runs.push({date: new Date().toISOString().split('T')[0], gate_result: 'PASS_OR_FAIL', failed_checks: []});
fs.writeFileSync('.claude/skills/qe-quality-assessment/run-history.json', JSON.stringify(h, null, 2));
"

每次运行前读取 run-history.json——如果最近 5 次运行中有 3 次未通过质量门禁，则发出警报。

技能组合

• 评估前 → 先运行 /qe-coverage-analysis 和 /mutation-testing
• 发现问题时 → 使用 /test-failure-investigator 诊断失败
• 审查 PR 时 → 结合 /code-review-quality 进行全面审查

注意事项

• 切勿相信 Agent 报告的通过/失败状态——曾发现 12 个测试失败，但 Agent 声称全部通过（Nagual 模式，奖励 0.92）
• 完成表演：Agent 将版本硬编码为“3.0.0”，而不是从 package.json 读取——请核实输出中的实际值
• 按优先级分批修复问题（P0 → P1 → P2），并在每批之间进行验证——不要并行修复所有问题
• quality-assessment 领域的成功率为 53.7%——请预期可能失败，并准备好备用方案
• 如果 HybridMemoryBackend 初始化失败，运行 aqe health 进行诊断，或运行 aqe init 重新初始化

协作

主要 Agent：qe-quality-analyzer、qe-deployment-advisor、qe-metrics-collector
协调器：qe-quality-coordinator
相关技能：qe-coverage-analysis、security-testing