Task Master (Grade A)针对 Claude AI 的安全测试开发技能。A等级。将高层次请求分解为有序的可操作任务。使用时，当用户要求将任务分解为步骤、创建计划或列出步骤时。

Why You Should Use This?

Individual: 30%+ token savings, stop repeating standards, security guardrails, instant scaffolding. Teams: Consistent AI behavior, day-one onboarding, codified patterns, compounding cost savings. Organizations: Always-on compliance, same standards across 1,000 projects, central governance, measurable ROI.

Asked by Ximenez Alvarado · Mar 14, 2026

What You Can Customize?

Project: Name, description, repo URL. Tech Stack: Language, framework, tools (e.g., TypeScript + Express or Python + FastAPI). Paths: Directory structure for handlers, services, and common code. Domain: Business entities like Order, Product, Customer with lifecycle states. Patterns: Code patterns such as a 7-step handler flow or error handling strategy. Testing: Quality gates like 90% coverage, test/lint/type-check commands. Database: Conventions like soft delete field, timestamp columns, naming. Packages: Internal packages settings (e.g., @your-org scope, registry URL). Conventions: Git and workflow rules like branch prefixes, commit format, PR templates.

Asked by Tunde Balogun · Mar 10, 2026

What it detects?

Language: Detects tsconfig.json, go.mod, Cargo.toml, requirements.txt, pom.xml, file extensions. Framework: Detects dependencies in package.json or requirements.txt (React, Express, Django, Spring, etc.). Database: Detects ORM configs (prisma/, sequelize, typeorm), .sql files, migration folders. Testing: Detects jest.config., vitest, pytest, cypress/, playwright.config. Infrastructure: Detects Dockerfile, terraform/, cdk.json, serverless.yml, cloud SDK dependencies. CI/CD: Detects .github/workflows/, .gitlab-ci.yml, Jenkinsfile.

Asked by Yelena Popova · Jan 18, 2026

Task Master (Grade A) 内容有哪些?

规则（47）：每次 AI 交互都会执行Coding Standard，总是开启。Agent（62）：为复杂任务提供的专属助手，通过 /agent-name 提供即时服务。技能（50）：带有任务清单的逐步步骤流程，按上下文触发。命令（37）：轻量化的TOKEN快捷操作，通过 /command 提供即时服务。钩子（12）：在 AI 循环中触发的自动化脚本，在事件发生前、后执行。模板（9）：提供给 handler、组件、测试用例等的骨架，通过技能和 Agent 引用。

Asked by Mireille Dupont · Jan 3, 2026

它是如何工作的?

第1层——预处理：钩子注入项目上下文和防止危险命令到达AI之前的项目内容；第2层——规则引擎：47种规则始终有效，实施令牌效率、安全性、架构、代码标准、数据库约定和测试门槛；第3层——专门处理：基于提示的某个组件激活，即使代理、技能或命令；第4层——后处理：钩子验证输出、自动格式化、扫描敏感信息并确认覆盖。

Asked by Ulrik Madsen · Dec 31, 2025