JochenYang

Implement distinctive, production-grade frontend UI code with strong visual direction, motion systems, local media asset generation, conversion-aware copy, and polished frontend execution. Use when building landing pages, marketing sites, product pages, dashboards, motion-heavy interfaces, or frontend experiences that need real assets and compelling copy.

Project context engine for repo orientation, cached handoff, and task-focused code retrieval.

Create and resume structured manual session handoffs for long-running development work. Use when approaching context limits, before manual reset, before switching models or IDEs, after a milestone, or when automatic compact would lose important implementation state.

Review current conversation, analyze tasks, errors, and user feedback, extract learning opportunities for skill improvement. Use when user says "reflect", "review session", "what did we learn", "session summary", or after completing a complex task.

Rapid prototyping tool for React interactive demos and single-file HTML artifacts. Uses shadcn/ui component library with 40+ components. Use when user asks to create React demo, build HTML artifact, prototype UI, or create product demo. Perfect for design validation and shareable browser previews.

Audit all .claude/*.md files for redundant instructions, verbose phrasing, and content that could move to memory. Use when user wants to optimize their Claude configuration rules.

Database architecture design, schema optimization, query performance tuning, and data migration. Handles PostgreSQL, MySQL, MongoDB, Redis. Use when user asks to design database, optimize queries, create migrations, or improve database performance. Do NOT use for frontend or application logic.

Comprehensive full-stack development for web, mobile, and game projects. Handles frontend (React/Vue/Angular), backend (Node.js/Python/Go/Java), mobile (Flutter/React Native/Swift/Kotlin), and game development (Unity/Unreal/Godot). Use when user asks to build, create, develop, implement, debug, or fix any web, mobile, or game project. Do NOT use for design-only tasks (use frontend-design or ui-ux-pro-max instead).

Use when executing plans through specialized subagents with evidence-gated review loops

Use when work needs an explicit delivery contract, iterative repair, bounded autonomy, or resume-safe state

Use when a user asks to continue, resume, pick up prior loop work, or inspect previous loop state

Use when work needs codebase orientation, impact analysis, existing-pattern discovery, or risk scouting before planning or implementation

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

Use when implementation direction depends on unresolved product/design choices, multiple viable approaches, or owner approval on a design.

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

Use when you have a written implementation plan to execute in a separate session with review checkpoints

将一个开发需求转化为可执行、可验证的最小实现计划，明确范围、影响、依赖、风险和验收标准。

按复现、隔离、可证伪假设、验证和最小修复流程处理开发问题，避免凭猜测直接改代码。

对当前代码改动进行只读、发现优先的工程审查，覆盖正确性、回归、安全、性能、数据一致性和测试缺口。

根据当前代码改动选择最小但足够的测试范围，执行并解释测试结果与覆盖缺口。

在提交前检查改动范围、验证证据、敏感信息和 Git 规范，并生成可供确认的分支与提交信息提案。

在发布前核对版本、变更记录、构建、测试、发布物、兼容性、安全和回滚条件，给出明确放行结论。

Project context engine for repo orientation, cached handoff, and task-focused code retrieval.

Local-first dual-layer personal knowledge base (per-project topic + shared) with hybrid retrieval (BM25 + hierarchy + RRF) and optional LLM answers. Use when recalling or saving project docs, cross-project patterns, decisions, runbooks, or postmortems. Match by intent (search/save/browse), not keywords. CLI: one shell call per command; JSON default for search/show/tree. Prefer project default_topic first, then --topic shared.