Apply this skill after every code change. Reviews and cleans up comments ONLY in the code that was just modified — never touches pre-existing comments outside the changed scope. Enforces lean, context-aware commenting: structural markers, genuine clarification, and non-obvious reasoning only. Strips conversation artifacts, redundant labels, and anything a fresh reader would find confusing or meaningless.
Apply this skill when creating or updating any project documentation that agents read — CLAUDE.md, AGENTS.md, README, architecture docs, agent_docs/, or any markdown that shapes how an agent understands the project. Covers what to write, what kills agent performance, how to structure for context efficiency, and how to keep docs honest and current — not artificial. Based on Anthropic official docs and practitioner patterns from 2025/2026.
Apply this skill whenever writing, reviewing, or generating TypeScript code — especially for backend/Node services, WebSocket clients, REST/WS API integrations, data normalization pipelines, and real-time aggregators. Covers 2026 patterns, strict typing, comment conventions, import order, file structure, and what to avoid.
Apply this skill when building or reviewing Vite + React + TypeScript projects — dashboards, SPAs, real-time UIs, data-heavy tools, and internal apps. Covers official Vite 8 + React 19 patterns, project structure, component conventions, state management split (server vs client), hooks, env vars, build config, and what to never do. Tailored for data-heavy, auth-gated, real-time applications like trading dashboards — not SEO-first marketing sites.
Apply this skill when writing, reviewing, or generating tests with Vitest. Covers Vitest v4 setup with Vite, config, file structure, test patterns, mocking (vi.fn, vi.mock, MSW for HTTP/WS), async testing, coverage with v8, what to never do, and patterns specific to TypeScript + real-time/WebSocket codebases. Based on official vitest.dev docs.