By suriyel
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
You are a codebase structure locator. You perform a breadth-first scan of a project to identify and catalog key structural positions — module boundaries, entry points, API endpoints, data models, configuration surfaces, test directories, and external integrations. Your output is a structured location inventory that downstream agents (codebase-analyzer, codebase-pattern-finder) use as their analysis target list.
You are a codebase pattern finder and health measurer. Given a location inventory from the codebase-locator agent, you analyze dependency structures, internal coupling, complexity hotspots, test coverage landscape, and technical debt markers. Your output is a metrics-driven analysis document with evidence tables.
**LANGUAGE RULE**: You MUST respond in Chinese (Simplified). All generated documents, reports, and user-facing output must be written in Chinese. Code identifiers and JSON field names remain in English.
You are a usage example generator. After System Testing passes with a Go verdict, you produce a concise set of **scenario-based** runnable examples that demonstrate how external developers and AI Code Agents can use this project.
You are a skill system reflection analyst. When user feedback during a Worker session indicates a skill produced wrong or suboptimal output, you analyze WHY and write a structured improvement record.
Use when design doc exists but no ATS doc and no feature-list.json - generate a global Acceptance Test Strategy mapping every requirement to acceptance scenarios with category constraints
Use when SRS doc exists but no design doc and no feature-list.json - take the approved SRS as input and produce an architecture/design document focused on HOW to build it
Use for on-demand deep exploration of an existing codebase - analyzes architecture, data flow, domain model, API surface, dependencies, and code health
Use before TDD in a long-task project — produce feature-level detailed design with interface contracts, algorithm pseudocode, diagrams, and test inventory
Use after quality gates pass in a long-task project — independently manages test environment lifecycle (start/cleanup), executes black-box acceptance testing per feature, generates ISO/IEC/IEEE 29119 compliant test case documents
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimnpx claudepluginhub suriyel/longtaskforagent --plugin long-taskBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
在 Claude Code 中,首先注册市场:
/plugin marketplace add suriyel/longtaskforagent
然后从该市场安装插件:
/plugin install long-task@longtaskforagent
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/suriyel/longtaskforagent/main/claude-code/install.sh | bash
Windows(PowerShell):
irm https://raw.githubusercontent.com/suriyel/longtaskforagent/main/claude-code/install.ps1 | iex
脚本会自动:
~/.claude/plugins/marketplaces/longtaskforagent/known_marketplaces.json 注册信息安装完成后,使用 Claude Code 安装插件:
/plugin install long-task@longtaskforagent
如果您使用 OpenCode:
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/suriyel/longtaskforagent/main/install.sh | bash
Windows(PowerShell,需开发者模式或管理员权限):
irm https://raw.githubusercontent.com/suriyel/longtaskforagent/main/install.ps1 | iex
安装完成后重启 OpenCode 即可激活。完整说明请参阅 OpenCode 安装指南。
启动 Claude Code 后,只需告诉它您想构建什么:
> 我想构建一个GitHub 热门项目周报系统。使用 long task skill。
系统将自动进入需求阶段,通过结构化提问帮助您完善需求,最终生成标准化的 SRS 文档。后续工作流程完全自动化:
需求 → UCD (如有UI) → 设计 → ATS (验收测试策略) → 初始化 → 工作循环 → 系统测试

一款 Claude Code 技能插件,将单会话 AI 编码转变为严谨的多会话软件工程工作流。
大多数 AI 编程助手在一次对话后会丢失上下文。Long-Task Agent 通过实现七阶段架构和持久状态桥接解决了这个问题——使 Claude Code 能够以专业工程团队的纪律,跨无限会话构建复杂项目。

| 问题 | Long-Task Agent 如何解决 |
|---|---|
AI 在 /clear 后忘记所有内容 | 持久化产物(feature-list.json、task-progress.md、git 历史)自动桥接会话 |
| AI 不理解需求就生成代码 | 符合 ISO/IEC/IEEE 29148 的需求收集在编写任何代码前产生经批准的 SRS |
| AI 跳过测试或编写浅层测试 | 严格的 TDD(红→绿→重构)配合覆盖率门禁(≥90% 行覆盖,≥80% 分支覆盖)和变异测试(≥80% 得分) |
| AI 产生不一致的 UI | 带令牌化设计系统的 UCD 风格指南确保所有功能的视觉一致性 |
| AI 验收测试覆盖不全 | ATS(验收测试策略)在设计后前置规划每个需求的测试类别,独立 subagent 审核确保无覆盖盲区 |
| AI 偏离批准的设计 | 设计接口覆盖门 + 每个功能后内联合规检查 |
| 无法安全地向现有项目添加功能 | 增量技能执行影响分析,就地更新 SRS/设计/UCD,用波次跟踪变更 |
| "在我机器上能跑"综合症 | 系统测试阶段(IEEE 829)包含回归、集成、端到端和 NFR 验证 |

每个项目都从结构化的需求收集开始——而不是编码。SRS 捕获做什么,UCD 捕获外观,设计文档捕获怎么做。三者全部批准后才会编写代码。
十多个持久化产物确保会话间零知识丢失:
| 产物 | 用途 |
|---|---|
feature-list.json | 带状态跟踪的结构化任务清单(JSON 防止模型损坏) |
task-progress.md | 逐会话日志,带当前状态标题 |
docs/plans/*-srs.md | 已批准的软件需求规格说明书 |
docs/plans/*-design.md | 已批准的技术设计文档 |
docs/plans/*-ats.md | 已批准的验收测试策略(需求→场景映射,独立 subagent 审核) |
docs/plans/*-ucd.md | 已批准的 UCD 风格指南(UI 项目) |
long-task-guide.md | 工作会话指南,含环境激活 + 工具命令 |
docs/test-cases/feature-*.md | 每功能的 ST 测试用例文档(ISO/IEC/IEEE 29119) |
docs/plans/*-st-plan.md | 带 RTM 的系统测试计划 |
docs/plans/*-st-report.md | 带 Go/No-Go 结论的系统测试报告 |
RELEASE_NOTES.md | Keep a Changelog 格式的活态变更日志 |
| Git 历史 | 带描述性提交的完整变更历史 |
每个功能都要通过一系列自动化质量门禁——无例外,无捷径:
每个工作会话专注于恰好一个功能。这防止上下文耗尽,确保干净的提交,并使每个功能独立可验证。


docs/plans/*-srs.md)docs/plans/*-ucd.md)docs/plans/*-design.md)docs/plans/*-ats.md)init.sh / init.ps1)每个循环遵循严格纪律:
定位 → 引导 → 配置门禁 → 开发工具门禁 → 计划
→ TDD 红 → TDD 绿 → 覆盖率门禁
→ TDD 重构 → 变异门禁
→ 功能 ST(黑盒) → 内联合规检查
→ 持久化 → 下一个功能
increment-request.json 信号文件 → 技能自动检测
Long-Task Agent 使用按需技能加载模式——只有引导路由器在会话开始时加载;阶段技能按需加载,保持上下文精简。
TODO駆動で大きな実装タスクを段階的に進め、サブエージェントがレビューするスキル。ディレクトリベースのタスク管理で長時間・大規模な変更を着実に完遂。
Software engineering workflows with skills for planning, implementation, quality review, and structured thinking, plus a suite of specialist agents
Persona-driven AI development team: orchestrator, team agents, review agents, skills, slash commands, and advisory hooks for Claude Code
Evidence-Driven Development Protocol — 24 command-equivalent skill surfaces plus a skill-native deep-work entry alias (cross-platform: Claude Code skills + Codex / Copilot CLI / Gemini CLI / SDK), 3-layer architecture with Skill-based phase dispatch, computational enforcement (worktree guard + phase transition injection), TDD enforcement, and receipt validation
Requirements-driven development workflow with quality gates for practical feature implementation
Superpowers Plus core skills library for Claude Code: planning, execution routing, TDD, debugging, and collaboration workflows