Token导航 LogoToken导航

Agent Skill · 开发

fault-injection-testing(故障注入测试)

用于辅助测试设计、自动化测试、用例整理和回归验证。它适合让 Agent 编写单元测试、端到端测试、测试计划或根据失败日志定位问题。使用时需要确认项目测试框架、运行命令和夹具数据,避免为了通过测试而改坏真实逻辑;涉及浏览器或外部服务时,应区分本地模拟、测试环境和生产环境。

fault-injection-testing 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为调用外部服务;许可证信息还不明确,第三方扫描已通过,按安装量在开发类的 21,746 个 Skill 里排第 18,175。

总安装 353周安装 15GitHub Stars 5下载 124核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:fault-injection-testing(故障注入测试)
来源仓库:https://github.com/apankov1/quality-engineering
仓库路径:skills/fault-injection-testing
安装命令:
npx skills add https://github.com/apankov1/quality-engineering --skill fault-injection-testing
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/apankov1/quality-engineering --skill fault-injection-testing

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

用于辅助测试设计、自动化测试和回归验证,帮助编写单元测试或端到端测试。适合处理故障路径测试、重试逻辑和外部服务交互的容错行为验证。使用时需确认项目测试框架、运行命令及夹具数据,避免误改真实逻辑。涉及浏览器或外部服务时,应区分本地模拟、测试环境与生产环境。

fault-injection-testing 属于开发类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/fault-injection-testing

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/apankov1/quality-engineering --skill fault-inj…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要调用外部服务,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

fault-injection-testing 怎么安装?

在终端执行 npx skills add https://github.com/apankov1/quality-engineering --skill fault-injection-tes…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

fault-injection-testing 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

fault-injection-testing 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

fault-injection-testing 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

fault-injection-testing 用的人多吗?

公开数据是累计安装 353 次,近一周安装 15 次,下载 124 次,GitHub 5 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

fault-injection-testing 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
fault-injection-testing

故障注入测试 · 本页

3535调用外部服务未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Fault Injection Testing

Test failure paths, not just happy paths.

Production systems fail. Databases time out. Networks reset. Services degrade. If you only test happy paths, your first encounter with failure will be in production. This skill teaches you to systematically inject faults and verify resilience behavior.

When to use: Testing recovery paths, retry logic, circuit breakers, transaction rollback, queue preservation on failure, any code that interacts with external services.

When not to use: Happy-path-only tests, pure functions, UI components, code with no external dependencies.

Rationalizations (Do Not Skip)

RationalizationWhy It's WrongRequired Action
"The happy path works"Production failures aren't happyTest each fault scenario explicitly
"Retry logic is simple"Exponential backoff has edge casesVerify delay calculations mathematically
"Circuit breakers are overkill"Cascading failures take down systemsImplement and test all 3 states
"Queue data is safe"Transient failures can lose queue itemsAssert queue preservation on every failure

What To Protect (Start Here)

Before generating fault tests, identify which resilience decisions apply to your code:

DecisionQuestion to AnswerIf Yes → Use
Service degradation must not cascadeDoes this code call an external service that other callers also depend on?CircuitBreaker
Retries must not overwhelm the targetCan multiple clients retry simultaneously after an outage?RetryPolicy with jitter
Failed operations must not lose queued workIs there a queue or buffer that persists across retries?assertQueuePreserved
Faults must be tested at boundaries, not inside logicWhere does your code cross a system boundary (DB, network, external API)?createFaultInjector

Do not generate tests for decisions the human hasn't confirmed. A circuit breaker test that verifies state transitions without connecting to a real failure scenario (cascading outage, thundering herd) is testing the mechanism, not the decision.


Included Utilities

import {
  createFaultScenario,
  CircuitBreaker,
  RetryPolicy,
  createFaultInjector,
  assertQueuePreserved,
  assertQueueTrimmed,
} from './fault-injection.ts';

Core Workflow

Step 1: Define Fault Scenarios

List all failure modes for your external dependencies:

const faultScenarios = [
  createFaultScenario('timeout', 'ETIMEDOUT', 'retry_scheduled'),
  createFaultScenario('connection_reset', 'ECONNRESET', 'circuit_half_open'),
  createFaultScenario('rate_limited', '429', 'backoff_extended'),
  createFaultScenario('partial_write', 'SQLITE_CONSTRAINT', 'idempotent_retry'),
];

Step 2: Implement Circuit Breaker

Circuit breakers prevent cascading failures by failing fast when a service is down:

const cb = new CircuitBreaker({
  failureThreshold: 3,    // Open after 3 failures
  resetTimeout: 30000,    // Try half-open after 30s
  successThreshold: 2,    // Require 2 successes to close
}, Date.now);

// Usage
if (!cb.canExecute()) {
  throw new Error('Circuit open - service unavailable');
}

try {
  await callExternalService();
  cb.recordSuccess();
} catch (err) {
  cb.recordFailure();
  throw err;
}

Step 3: Test Circuit Breaker State Machine

The circuit breaker is a state machine: closed → open → half-open → closed. Test all transitions:

describe('circuit breaker', () => {
  it('starts closed', () => {
    const cb = new CircuitBreaker({ failureThreshold: 3, resetTimeout: 1000 });
    assert.equal(cb.getState(), 'closed');
  });

  it('opens after failure threshold', () => {
    const cb = new CircuitBreaker({ failureThreshold: 3, resetTimeout: 1000 });
    cb.recordFailure();
    cb.recordFailure();
    cb.recordFailure();
    assert.equal(cb.getState(), 'open');
  });

  // Use fake clock (inject nowFn) — no setTimeout, no flake
  it('transitions to half-open after timeout', () => {
    let now = 0;
    const cb = new CircuitBreaker({ failureThreshold: 1, resetTimeout: 50 }, () => now);
    cb.recordFailure();
    now = 60;
    assert.equal(cb.getState(), 'half-open');
  });

  it('closes on success in half-open', () => {
    let now = 0;
    const cb = new CircuitBreaker({ failureThreshold: 1, resetTimeout: 10 }, () => now);
    cb.recordFailure();
    now = 20;
    cb.recordSuccess();
    assert.equal(cb.getState(), 'closed');
  });

  it('reopens on failure in half-open', () => {
    let now = 0;
    const cb = new CircuitBreaker({ failureThreshold: 1, resetTimeout: 10 }, () => now);
    cb.recordFailure();
    now = 20;
    cb.recordFailure();
    assert.equal(cb.getState(), 'open');
  });
});

Step 4: Implement Retry Policy with Backoff

Exponential backoff with jitter prevents thundering herd:

const policy = new RetryPolicy({
  maxRetries: 5,
  baseDelay: 100,
  maxDelay: 30000,     // Hard cap: final jittered delay never exceeds this
  jitterFactor: 0.1,  // ±10% randomization
}, Math.random);

// Get delay for attempt N (0-indexed)
const delay = policy.getDelay(2);  // ~400ms ± jitter

Inject deterministic RNG for stable tests:

const low = new RetryPolicy({ maxRetries: 3, baseDelay: 1000, jitterFactor: 0.1 }, () => 0);
const high = new RetryPolicy({ maxRetries: 3, baseDelay: 1000, jitterFactor: 0.1 }, () => 1);

RetryPolicy and CircuitBreaker validate numeric config and throw on invalid values (negative delays, non-integer thresholds/retry counts, jitter outside [0, 1]).

Step 5: Test Backoff Calculations

Verify mathematical correctness of backoff:

describe('retry policy', () => {
  it('doubles delay each attempt', () => {
    const policy = new RetryPolicy({ maxRetries: 5, baseDelay: 100 });
    assert.equal(policy.getDelayWithoutJitter(0), 100);
    assert.equal(policy.getDelayWithoutJitter(1), 200);
    assert.equal(policy.getDelayWithoutJitter(2), 400);
    assert.equal(policy.getDelayWithoutJitter(3), 800);
  });

  it('caps at maxDelay', () => {
    const policy = new RetryPolicy({ maxRetries: 10, baseDelay: 100, maxDelay: 500 });
    assert.equal(policy.getDelayWithoutJitter(10), 500);
  });

  it('jitter stays within bounds', () => {
    const policy = new RetryPolicy({ maxRetries: 5, baseDelay: 1000, jitterFactor: 0.1 });
    for (let i = 0; i < 20; i++) {
      const delay = policy.getDelay(0);
      assert.ok(delay >= 900 && delay <= 1100);
    }
  });
});

Step 6: Inject Faults at Boundaries

Use createFaultInjector to wrap functions with controlled failure triggers:

const realDbWrite = async (data) => { /* ... */ };

const faultMap = {
  timeout: new Error('ETIMEDOUT'),
  constraint: new Error('SQLITE_CONSTRAINT'),
  reset: new Error('ECONNRESET'),
};

const dbWrite = createFaultInjector(realDbWrite, faultMap);

// In tests
await dbWrite(null, data);       // Normal execution
await dbWrite('timeout', data);  // Throws ETIMEDOUT (synchronously)
Note: createFaultInjector throws synchronously even when wrapping async functions. This is intentional — fault injection simulates failures at the call boundary, not inside the async pipeline. Use assert.throws(), not assert.rejects().

Step 7: Assert Queue Preservation

Transient failures must NOT lose queued data:

it('preserves queue on transient failure', async () => {
  const queueBefore = [{ sequenceNumber: 1 }, { sequenceNumber: 2 }];

  // Simulate failure during processing
  try {
    await processWithFault(queueBefore, 'timeout');
  } catch {}

  const queueAfter = getQueue();
  assertQueuePreserved(queueBefore, queueAfter);
});

it('trims queue after successful processing', async () => {
  const queueBefore = [{ sequenceNumber: 1 }, { sequenceNumber: 2 }, { sequenceNumber: 3 }];

  await processSuccessfully(queueBefore, /* maxSeq */ 2);

  const queueAfter = getQueue();
  assertQueueTrimmed(queueAfter, 2);  // Only seq 3 should remain
});

Violation Rules

missing_circuit_breaker_test

Circuit breakers MUST have tests for all 3 state transitions: closed→open, open→half-open, half-open→closed/open. Severity: must-fail

missing_backoff_verification

Retry policies MUST verify exponential calculation, max delay cap, and jitter bounds. Severity: must-fail

missing_queue_preservation_test

Operations that can fail MUST have tests verifying no data loss on transient failure. Severity: must-fail

fault_injection_inside_unit

Fault injection should only happen at system BOUNDARIES (database, network, external services), not inside business logic. Severity: should-fail


Companion Skills

This skill provides testing utilities for resilience patterns, not resilience architecture guidance. For broader methodology:

  • Search resilience on skills.sh for bulkhead isolation, health checks, graceful degradation
  • The circuit breaker is a state machine — use model-based-testing for systematic transition matrix coverage

Quick Reference

ComponentStates/BehaviorKey Tests
Circuit Breakerclosed → open → half-open → closedAll 4 transitions, threshold counts, timeout timing
Retry PolicyExponential backoff with capDelay doubling, max cap, jitter bounds
Fault InjectorNamed faults → specific errorsEach fault triggers correct error
Queue PreservationNo data loss on failureBefore/after sequence comparison

See patterns.md for failure matrix methodology, boundary-only injection rules, and integration with real infrastructure.

平台分布

Codex

40.03%
按下载量换算50

Claude

28.11%
按下载量换算35

Cursor

18.29%
按下载量换算23

Gemini CLI

10.23%
按下载量换算13

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills