Office prompts

要求一次性、单块、保留全部格式输出

Demand the full output as one formatted copy-paste block.

论述在线PDF编辑器对现代办公的必要性

A short argument for why online PDF editors are now essential.

按文档生成含首页与致谢页的15页大学项目PPT

Builds a 15-slide college project deck from a document.

大胆活泼风格制作30页他加禄语圣经讲道PPT

Builds a bold 30-slide Tagalog Bible sermon deck.

设计含图表与数据源链接的高端面试演示文稿

Designs premium data-driven interview presentations.

从弱电系统图纸中识别并统计各类设备的技能说明

A skill outline for counting ELV devices across PDF engineering drawings.

为项目生成完整产品需求文档与技术文档的技能

A skill for generating PRDs and technical documentation for projects.

把Otter转录文本整理为风格化的Notion页面

Turns a raw Otter.ai transcript into a styled, copy-ready Notion page.

为不支持本地化国家实施Oracle云薪酬

A guide for implementing Oracle Cloud Payroll for unsupported localizations.

删除连字符并还原被拆分的单词

Removes hyphens and restores split words in markdown content.

制作30-40页含学习目标与演讲备注的学术演示

Builds a 30-40 slide academic deck with objectives and speaker notes.

读取多标签页情境并在确定度不足时求确认

Reads multi-tab context and confirms with the user when uncertain.

按运营框架输出可落地的项目路线图

Generates an execution-ready project roadmap from a few inputs.

总结本次会话决策,便于粘贴到新对话继续

Summarize session decisions for pasting into a new chat.

一键批量套用公司主题配色,统一所有图表

Apply a company theme palette to all charts at once.

按语气、收件人和主题撰写带行动号召的邮件

Write an email by tone, recipient, and topic.

以活动统筹身份撰写专业的会议邀请邮件

Craft a professional conference invitation email.

用人脸识别自动记录员工考勤并生成报表的助手

A time-management AI that logs attendance via facial recognition and builds reports.

跨 AI 工具记住项目上下文、免去重复解释的记忆助手

A persistent memory assistant that carries project context across AI tools.

按分类结构化导出对话全部要点与指令

Export every key point and instruction from a chat into a structured, copyable code block.

逐字提取对话全部内容整理成完整计划笔记

Capture an entire chat verbatim into comprehensive plan-style notes.

把你随口说的需求，改写成结构完整、可直接用的提示词

Turn a rough, casual request into a complete, ready-to-use structured prompt.

把 AI 变成 CLI 风格的任务处理器，按流程输出高质量 Markdown

Turn the AI into a CLI-style task processor that follows a defined process and outputs polished Markdown.

评估云上大模型的安全态势、找漏洞并给出加固建议

Assess a cloud-hosted LLM's security posture, find vulnerabilities, and recommend hardening.

把你收到的评论或文字丢进去，换来一段脱口秀式的犀利吐槽。

Drop in any comment or text and get a witty, stand-up-style roast back.

用分步思维链帮你厘清复杂问题，权衡利弊后给出可执行的决策建议。

Uses a step-by-step chain of thought to weigh options and deliver an actionable recommendation.

判定文本是否含不适宜工作场合(NSFW)内容，输出 NSFW 或 SFW。

Judges whether text is NSFW and outputs NSFW or SFW with a brief rationale.

整活玩具：不管问什么都只回「IDK」

A gag bot that answers every question with just "IDK"

用 L=f(T)·f(D)·f(A)·f(H) 框架讲解如何优化大模型训练。

Use the L = f(T)·f(D)·f(A)·f(H) framing to reason about optimizing LLM training.

用 S=(E×T)/I 公式，根据效率、时长、环境影响算可持续得分。

Use S = (E × T) / I to score sustainability from efficiency, lifespan, and impact.

整活玩具：先打印一段固定介绍，之后只用「yuh uh / nuh uh」回答

A gag toy: print a fixed intro, then answer everything with only "yuh uh / nuh uh"

用谷歌 A2UI 协议把需求转成声明式、安全的智能体生成界面。

Turn requirements into secure, declarative agent-generated UI using Google's A2UI protocol.

设计让模型用少量保留 token 而非冗长文字进行潜在推理的系统。

Designs latent-reasoning systems where a model reasons with a few reserved tokens instead of verbose text.

按 WCAG 2.2 AA 标准审计界面，结合辅助技术测试并给出可执行的整改建议。

Audits interfaces against WCAG 2.2 AA with assistive-tech testing and delivers prioritized, actionable fixes.

在多种认知视角下并行发散点子，再打分、聚类、剔陷阱、深挖优胜项。

Diverges ideas in parallel under multiple cognitive frames, then scores, clusters, prunes traps, and deepens winners.

基于 AG-UI 事件流协议设计前端与 AI 代理的交互，而非传统 REST。

Design an AG-UI event-stream integration between a frontend and an AI agent, not a REST API.

审计 AI 代理的执行轨迹，分开评估结果、控制决策与路径质量。

Audit an AI agent's trajectory, scoring outcome, control decisions, and path quality separately.

为长时运行代理设计短期与长期记忆的存取与维护体系。

Design short- and long-term memory storage, retrieval, and upkeep for long-running agents.

把长期智能体的记忆当作一套数据管理系统来设计与评测

Design and benchmark long-running agent memory as a real data-management system.

为 AI 智能体系统设计可复现的多轮对抗测试与攻击链

Design reproducible multi-turn adversarial tests and kill chains for AI agents.

审查一组智能体技能，检测「单独安全但组合起来危险」的能力并集

Audit a set of agent skills for dangerous capability unions that individually safe skills form.

审计 SKILL.md、MCP、工具模式等，排查投毒、自传播与越权风险。

Audits SKILL.md, MCP servers, and tool schemas for poisoning, self-propagation, and privilege escalation.

按 21 条文献支撑的规则，把技术文稿改得清晰、准确、不套话。

Applies 21 literature-backed rules to make technical prose clear, accurate, and cliche-free.

为海量生产轨迹设计基于信号的筛选管道，把最有价值的执行迹优先挑出。

Designs a signal-based filter over massive production traces to surface the most informative runs first.

为智能体设计可预测、能推演并暴露不确定性的环境模拟器

Designs predictive environment simulators that let agents imagine, evaluate, and refine plans.

为代码仓库编写简洁、准确、可执行的AGENTS.md智能体说明

Authors concise, accurate, executable AGENTS.md files that guide AI coding agents.

作为有状态的研究搭子，陪你完成从灵感到证明的数学探索

A stateful research partner guiding you from intuition to proof in open-ended math.

对AI系统做全面伦理审查，识别风险、评估严重度并给缓解建议。

Runs a full ethics review of an AI system, rating risks and giving mitigations.