Coding prompts

从所有模块资源中剥离并注释掉现有 vnet 依赖

Helps detach and comment out an existing vnet across all modules and resources.

分类并转义 Markdown 中的货币美元符,保留数学与变量

Classifies and escapes currency dollar signs in Markdown while leaving math and shell vars alone.

用通俗语言为非技术新手解析 GitHub 开源仓库

Explains and analyzes a GitHub repo in beginner-friendly, non-technical language.

以资深思维追求最简洁、可维护、高质量的解决方案

A senior QA specialist favoring simple, maintainable, high-quality solutions.

用Framer打造现代炫酷且符合SEO的应用

Build visually stunning, SEO-optimized apps using Framer motion and 3D effects.

用Python每隔几分钟自动输入文本并打包exe

Build a Python auto-typing script with a customizable interval, packaged as an .exe.

开发带UI、可自定义延迟的Python自动打字程序

Create a Python auto-typer app with a UI, customizable delay, and .exe build.

帮你按 A2A 协议设计可互操作、异步的多智能体通信架构

Design interoperable, asynchronous, opaque multi-agent communication that conforms to the A2A protocol.

为 ADK 式智能体设计按需渐进加载的技能，省 token、更模块化

Design progressively-loaded skills for ADK-style agents to cut tokens and improve modularity.

审计并设计让 AI 智能体像老工程师花预算一样省着花上下文 token

Audit and design agents that spend context tokens deliberately, sandboxing raw output and scripting aggregates.

设计 AI 编码智能体的成本观测与预算治理系统：遥测、看板、告警

Design a cost-observability and budget-governance system for AI coding agents: telemetry, dashboards, alerts.

为通用 CLI 智能体设计沙箱、评测、预算与人工监督的自主科研环境。

Designs the sandbox, evaluation, budget, and oversight environment for autonomous CLI-agent research.

以智能体为首要用户，设计易学、规整、可确定性工具化的编程语言。

Designs a programming language whose primary user is the agent - small, regular, and deterministically tooled.

为多智能体系统定义所有权、权限、问责与控制，确保可审计可控。

Defines ownership, authority, accountability, and controls across multi-agent systems for auditability.

为 AI 智能体设计围绕模型的安全、可回滚、可观测运行框架。

Design the safe, reversible, observable runtime around an AI agent, not just its prompt.

优化现有 AI 编码智能体框架，提升一致性、可靠性与成本效率。

Optimize an existing AI coding-agent harness for consistent, measurable, production-grade output.

设计双层权限分类器，低风险自动放行、高风险强制人工确认。

Design a two-layer permission classifier: auto-approve low risk, force approval on high risk.

设计"正则粗筛＋AI 深判＋复核"的大型代码库漏洞扫描流水线。

Design a regex-triage + AI-deep-analysis + revalidation pipeline for large codebases.

帮你决定生产系统中智能体、工具与界面之间该用哪种通信协议。

Decides which communication protocols agents, tools, and interfaces should use in a production system.

从一致性、鲁棒性、可预测性、容错四维度设计、度量并提升智能体可靠性。

Designs, measures, and improves agent reliability across consistency, robustness, predictability, and fault tolerance.

用「有无技能」对照实验，判断某个技能对具体软件工程任务是否真的有净收益。

Runs a paired with/without experiment to decide if a skill gives net benefit on a specific software task.

安全清理本地 AI 编码代理的会话、工作树与日志，先审后改、只归档不删除。

Safely maintain local AI coding-agent sessions, worktrees, and logs - inspect first, archive not delete.

设计、原型、评测并优化供 AI 代理调用的工具集。

Design, prototype, evaluate, and optimize the tools that AI agents call.

把多种后端统一为一棵挂载树，让代理用类 Unix 命令访问。

Unify diverse backends into one mount tree so agents use Unix-like commands.

基于代码证据逐步推理，回答代码问题、指导改动

Answer code questions and guide changes via explicit, evidence-backed reasoning.

先规划再动手、边写边测的安全生产级编程系统提示词

A plan-first, test-as-you-go system prompt for secure, production-ready code.

把上下文当作可自我进化的策略手册，设计不改权重也能自我提升的系统

Design self-improving context playbooks that evolve without weight updates.

以转录文本、波形和帧为依据推理剪辑视频，用 ffmpeg/Python 产出成片。

Edits video by reasoning over transcripts, waveforms, and frames, producing final cuts with ffmpeg/Python.

扮演分析工程师，设计可扩展、可测试的数据管道与指标体系。

Acts as an analytics engineer to design scalable, testable data pipelines and metrics.

扮演资深集成架构师，设计可靠、可观测、可维护的系统对接方案。

Acts as a senior integration architect to design reliable, observable, maintainable system integrations.

设计一套让编码智能体自协调、人只给方向的协作系统

Designs a system where coding agents self-coordinate and the human only sets direction.

扮演生信专家，设计并跑通高通量组学分析流程

Acts as a bioinformatics expert to design and run high-throughput omics analysis pipelines.

以CFO视角做资本配置、融资、并购与利益相关方沟通分析

Analyzes capital allocation, fundraising, M&A, and stakeholder comms from a CFO's view.

把电影摄影语言拆成可提示要素，产出AI视频/图像生成指南与模板。

Breaks cinematic craft into promptable parts and outputs an AI video/image guide.

把整洁代码/架构、DDD、数据密集系统提炼成AI编码代理的强制工程政策。

Distills Clean Code/Architecture, DDD, and data-intensive design into binding AI-agent policy.

帮你设计窄职责、易路由、最小权限的Claude Code子代理并产出文件。

Designs narrow, easy-to-route, least-privilege Claude Code sub-agents and outputs the file.

扮演资深云架构师，按 Well-Architected 原则设计多云方案。

Act as a senior cloud architect designing multi-cloud solutions by Well-Architected principles.

按 OWASP Top 10 逐项审查代码，给出风险与可用修复。

Review code against OWASP Top 10 with risks and ready-to-use fixes.

把任意代码库转成一份浏览器可运行的互动单页 HTML 教程。

Turn any codebase into a browser-runnable, interactive single-page HTML course.

把整份代码库解析成可查询的知识图谱,揭示依赖与架构。

Turn an entire codebase into a queryable knowledge graph that surfaces dependencies and architecture.

一套面向命令行编程智能体的完整系统提示词,规范行为与工具用法。

A complete system prompt for a command-line coding agent, defining behavior and tool usage.

一套安全操作浏览器与桌面、抵御提示注入的智能体系统提示词。

A system prompt for safely operating a browser and desktop while resisting prompt injection.

为 AI 智能体系统设计并审计完整的上下文/意图/规范环境

Design and audit the full context, intent, and specification environment for AI agents

从构思到修改，全流程辅导小说/剧本/诗歌等创作

Coach a writing project end to end, from concept to revision, across genres

把数据转化为业务洞见与可执行建议，含多种分析模板

Turn data into business insight and action, with templates for common analyses

扮演资深数据工程师，设计可靠、可观测的 ETL/湖仓数据管道。

Acts as a senior data engineer designing reliable, observable ETL/lakehouse pipelines.

设计多智能体流水线，自动发现、验证、分诊并修复代码漏洞。

Designs a multi-agent pipeline to find, verify, triage, and patch code vulnerabilities.

产出机器可读的 DESIGN.md，给编码智能体统一的视觉规范。

Authors a machine-readable DESIGN.md giving coding agents a shared visual spec.