skills

把报告编排成读者向文章：只重组排序起标题，不增删改内容。解决标题夸张、子标题冗长、逻辑弱，支持超长报告信息不丢

Engineering discipline for AI coding agents. Apply when writing, refactoring, or reviewing code that will be maintained. Not a tutorial on SOLID/DRY (you already know those) — it calibrates the trade-off calls and hard limits where agents predictably go wrong, and forces you to verify instead of guess.

面向全国所有省份的通用高考志愿填报建议器：用省份快速学习协议现查现学任意省当年规则，出默认保守的冲稳保三档+可直接录入本省志愿系统的草表。位次锁定推演、官方源+历年+多源交叉验证、零杜撰。Use for 高考志愿填报/选校选专业/冲稳保梯度/志愿草表/各省志愿规则适配/录取概率推演。

把技术博客长文/选题改写成小红书笔记(标题5选1+正文+配图清单+标签),走「五岳AI虚拟公司」过程纪实人设、强制去AI味与护城河自检。用于小红书选题/写笔记/内容改造/博客拆稿。

把 xhs-content-forge 的配图清单渲染成小红书封面/配图(4x高清、文字清晰)。固定五岳AI公司视觉IP,5套版式:封面×3(大字报/对比/看板)+内页×2(要点图/金句卡)。用于小红书出封面/做配图/渲染笔记图。

系统化投研能力中枢。覆盖宏观/行业/政策/个股/投资可行性/风险/投资规划七类研究，内置9条研究红线、深度研究5步法、7专题框架、5研究模板、8重点行业一手源清单，以及多源数据采集脚本（SEC EDGAR/HKEX/AKShare/FRED/政策门户等）。Use for: 任何投资研究、行业分析、公司研究、个股深度研究、政策解读、可行性评估、风险分析、投资规划，以及研究所需的真实一手数据采集。纯研究，不做盯盘/交易信号/交易执行。

Create professional, publication-quality technical architecture diagrams using pure SVG in HTML, then screenshot via Playwright. Produces crisp, pixel-perfect diagrams with precise connection lines, color-coded modules, and clear text at any resolution. Use when: (1) user asks for a system architecture diagram, (2) user wants a technical component diagram or flow chart, (3) user needs a data flow or pipeline visualization, (4) any diagram requiring accurate text labels and precise connecting lines. Triggers: "architecture diagram", "架构图", "技术架构", "system diagram", "component diagram", "flow diagram", "数据流图", "模块图", "draw architecture", "画架构图", "technical diagram". Prefer this over AI image generation for any diagram with text labels.

Generate ultra-high-resolution screenshots from HTML content using Playwright. Default 4x device scale factor produces crisp, pixel-perfect output ideal for UI mockups, dashboards, data visualizations, infographics, website previews, and any content requiring accurate text rendering. Use when: (1) user wants a website/webpage screenshot or mockup image, (2) content has text, numbers, tables, charts that must be legible, (3) user wants a UI design rendered as an image, (4) HTML Canvas or data-driven visuals need to be exported as PNG/JPEG. Triggers: "website screenshot", "网页截图", "网站首页图", "UI mockup image", "dashboard image", "数据可视化图片", "render HTML to image". Prefer this over AI image generation (stable-image-ultra) when the content has structured text, data, or UI elements. AI image generation cannot produce readable text.

Generate the highest quality photorealistic images using Stability AI's Stable Image Ultra and Stable Diffusion 3.5 Large models via AWS Bedrock. The most powerful text-to-image models on Bedrock. Supports multiple AWS auth methods: environment variables, credentials file, named profiles, IAM instance roles, SSO, or direct access keys. Use when the user asks for ultra-high-quality, photorealistic, or premium image generation. Triggers: "ultra quality image", "photorealistic image", "best quality image", "stable image ultra", "SD3.5", "stability ai", "ultra画质", "超高清图片", "照片级图片", "最高画质", "generate image", "生成图片", "画一张", "做配图", "生图". This is the DEFAULT image generation skill for ALL agents. Always use this unless the user explicitly asks for text-heavy diagrams (use HTML Canvas instead). Requires AWS Bedrock access with Stability AI models enabled in us-west-2. NOT for: editing existing images, generating video, or HTML/CSS canvas rendering.

Behavioral guidelines to reduce common LLM coding pitfalls, derived from Andrej Karpathy's observations. Apply these four principles when writing, editing, or reviewing code — especially for non-trivial changes. Triggers on coding tasks, code reviews, refactoring, bug fixes, feature implementation, or when the user asks for careful/disciplined coding behavior.

Generate images using Amazon Nova Canvas via AWS Bedrock. Supports multiple AWS auth methods: environment variables, credentials file, named profiles, IAM instance roles, SSO, or direct access keys. Use when the user asks to generate, create, draw, or produce an image, picture, illustration, photo, artwork, poster, icon, banner, or any visual content. Triggers: "generate an image", "draw me", "create a picture", "make an illustration", "画一张", "帮我画", "生成图片", "做一张图", "AI绘图". Requires AWS Bedrock access with Nova Canvas model enabled. NOT for: editing existing images, generating video, or HTML/CSS canvas rendering.

OpenClaw 系统诊断和性能分析工具。分析 agent 推理耗时、Token 用量、工具调用统计、 Run 时间线、Gateway 重启历史。支持多种模式：批量分析（默认）、实时跟踪（-f）、 摘要统计（-s）、高级诊断（--advanced）。支持多 Agent 过滤。 使用场景：当用户询问 OpenClaw 运行状态、性能瓶颈、推理延迟、Token 消耗、 工具执行统计、错误排查、agent 活动分析时触发。 触发词：诊断、diagnostics、性能分析、推理耗时、token统计、运行状态、 agent分析、工具调用统计、Run详情、Gateway重启。 指令触发：/diag — 直接执行诊断，支持参数透传。

搜索和获取中国软考（计算机技术与软件专业技术资格考试）历年真题。 覆盖9个科目：系统架构设计师、信息系统项目管理师、系统分析师、网络规划设计师、 软件设计师、网络工程师、信息安全工程师、数据库系统工程师、系统集成项目管理工程师。 支持按科目、知识领域、时间范围（2009-2025）筛选。支持综合知识、案例分析、论文三科。 触发词：软考真题、架构师真题、历年真题、刷题、ruankao questions、搜索真题。

生成中国软考系统架构设计师论文。根据论题和要求，按照万能模板结构生成高分论文。 论文包含摘要（300-330字）和正文（2000-2500字）五段式结构：项目背景、理论过渡、技术实践、总结展望。 使用场景：当用户提供软考论文题目、系统架构设计师考试论文、软考论文练习时触发。 触发词：软考论文、架构设计师论文、系统架构师论文、ruankao essay、软考作文。