Au1s1

Pragmatic code review & pre-push quality gate. Use when the user says "review code", "check code quality", "any issues with this code", "ready to push", "prepare commit", "submit code", "push to remote", or any variation requesting code inspection before merging or pushing.

Documentation drift auditor and normalizer for CLAUDE.md / AGENTS.md projects (supports Claude Code, Codex, and any project using agent instruction files at root and/or per module). Use when the user says "audit docs", "check doc freshness", "doc drift", "docs out of date", "文档审计", "文档漂移", "检查文档", "文档同步", "memory index", "MEMORY.md check", "normalize docs", "整理文档", "更新文档", "update docs", "doc health", "文档健康". Also triggers implicitly when user mentions stale docs, wrong ports in CLAUDE.md / AGENTS.md, or missing memory entries.

Playwright browser automation with two modes: (A) Record mode — automate web interactions (forms, clicks, screenshots/Canvas export) requiring JS rendering or login state. (B) Direct scripting — render Mermaid diagrams, charts, or HTML to PNG images without recording. Keywords: playwright, codegen, browser automation, web scraping, architecture diagram, render chart, Mermaid to PNG, screenshot export, canvas export. Do NOT trigger for: static page fetching (use WebFetch/curl), pure API calls, or DOM text extraction without interaction.

Scan an unfamiliar codebase, generate CLAUDE.md + OVERVIEW.md, and capture domain knowledge that code alone can't tell you. This skill should be used when the user wants to understand a new project, onboard onto a codebase, set up project documentation, create or update CLAUDE.md, or map project architecture. Common triggers include "what does this project do", "help me get started", "I just joined this repo", "document this codebase", "set up dev docs". Also applies when working on any unfamiliar or inherited project, even without an explicit onboarding request — if the user seems lost in a new codebase, this skill can help.