Produce self-contained HTML artifacts instead of markdown when the request benefits from spatial layout, color, diagrams, interactivity, sharing, or a round-trip editor. Use this skill for substantial docs, writeups, plans, specs, reports, explainers, summaries, comparisons, reviews, PR descriptions, mockups, diagrams, flowcharts, decks, status updates, post-mortems, incident reports, playgrounds, one-off editors, and document-bound editors. Also use it when the user wants to edit an existing markdown, JSON, YAML, prompt, config, or planning document through an HTML tool and export a patch/diff/instruction for an agent to apply. Stay in markdown only for short conversational replies, code-only outputs, terminal-style command answers, and content that is genuinely just a few sentences.
Maintain a personal skills index repository such as L-yifan/skills. Use when the user discovers a useful external skill, wants to add an external skill reference to the repo README, creates a new custom skill and wants it backed up into the repo, or asks to sync, catalog, publish, or preserve skills in their personal skills repository.
Analyze and optimize any system, workflow, plan, design, or problem by identifying its underlying structural patterns and applying proven optimization principles. Use when the user wants to find bottlenecks, streamline a process, restructure an organization, debug a workflow, or asks "why is this slow", "how to improve this", "what's wrong with this structure", "optimize this workflow". Stackable — works alongside any other skill or standalone. Also use when user describes friction, inefficiency, or complexity in any system — technical, organizational, or creative.
Find and recommend useful agent skills from a personal skills index repository such as L-yifan/skills. Use when the user asks which skill to use, wants to discover a skill for a task, asks whether an existing backed-up skill fits, wants external skill suggestions, or needs install and backup recommendations for skills.
Use when the user wants to write, improve, or evaluate prompts for Claude Code agent teams, teammates, multi-agent collaboration, parallel reviewers, competing hypotheses, complex refactors, coordinated investigations, PR reviews, or tasks that need independent Claude Code sessions to discuss and coordinate.
Access AI-generated documentation and insights for GitHub repositories via DeepWiki. This skill should be used when exploring unfamiliar codebases, understanding repository architecture, finding implementation patterns, or asking questions about how a GitHub project works. Supports any public GitHub repository.
去除文本中的 AI 生成痕迹。适用于编辑或审阅文本，使其听起来更自然、更像人类书写。 基于维基百科的"AI 写作特征"综合指南。检测并修复以下模式：夸大的象征意义、 宣传性语言、以 -ing 结尾的肤浅分析、模糊的归因、破折号过度使用、三段式法则、 AI 词汇、否定式排比、过多的连接性短语。