编程提示词大全

扮演开发者布道师，优化 DX、产出内容、经营社区并反哺产品。

Act as a developer advocate: improve DX, produce technical content, build community, and feed product.

为 LLaDA 等扩散式语言模型设计、优化和调试提示词。

Design, optimize, and debug prompts for diffusion LMs like LLaDA, with sampling and debugging strategy.

扮演资深固件工程师，为受限 MCU 设计内存安全的生产级固件。

Act as a senior firmware engineer designing deterministic, memory-safe production firmware for constrained MCUs.

帮你设计 VLA 机器人系统：感知-推理-行动、仿真到现实、安全

Helps design VLA robotic systems: perceive-reason-act, sim-to-real, and safety

像资深同事一样一起写代码：会反问、会push back、诚实

Codes with you like a senior colleague: pushes back, questions, stays honest

量化模型在基准与真实流量间的行为差距并给出闭环方案

Quantifies the gap between benchmark and production behavior and closes it

扮演资深前端工程师，产出高性能、可访问、响应式的实现方案

Act as a senior frontend engineer delivering performant, accessible, responsive implementations.

用度量化设计规则纠正 AI 生成千篇一律 UI 的倾向，产出高级界面

Override AI's bias toward generic UIs with metric-based design rules to produce premium interfaces.

设计一套仿真实游戏工作室的分层 AI 团队，含职责、流程与质量门

Design a tiered AI team mirroring a real game studio, with roles, pipeline, and quality gates.

教你为 Suno、Udio、ElevenLabs 等写出专业级音频提示词

Teaches you to write professional-grade prompts for Suno, Udio, ElevenLabs, and more.

教你为各大 AI 绘图模型写出专业级、可复制的提示词

Teaches you to write professional, copy-paste prompts for every major AI image model.

在终端里扮演资深工程搭档,遵循计划—执行—验证闭环

Acts as a senior engineering partner in the terminal, following a plan-execute-verify loop.

为AI Agent设计相互引用、共享产品语境的营销技能生态

Design an interlocking marketing skill ecosystem for AI agents around a shared product context.

用数据驱动实验诊断漏斗、优化获客激活留存与病毒增长

Use data-driven experiments to diagnose the funnel and optimize acquisition, activation, retention, and virality.

决定AI Agent管线在何处、如何接入Headroom做可逆上下文压缩

Decide where and how an AI-agent pipeline integrates Headroom for reversible context compression.

从状态、重组、副作用等维度评审并重构 Compose 代码

Review and refactor Compose code across state, recomposition, and effects

设计、部署、加固并调优生产级 Kubernetes 集群

Design, deploy, harden, and tune production Kubernetes clusters

按会议/期刊模板排版论文并润色学术英文

Format papers to venue templates and polish academic English

设计生产级 LLM 系统：提示、RAG、微调、推理服务与安全护栏。

Designs production LLM systems: prompting, RAG, fine-tuning, inference serving, and guardrails.

为长期运行的智能体设计逐字存储、本地检索、可复现基准的记忆系统。

Designs verbatim, locally-stored, benchmark-driven memory systems for long-running agents.

为产品出海设计从 i18n 架构到翻译流程、文化适配的完整本地化战略。

Designs a full go-global localization strategy: i18n architecture, translation workflows, cultural fit.

把工程师之间的技术内容改写成给管理层的更新

Rewrite engineer-to-engineer content into a leadership update for a chosen channel.

根据工具/API描述，设计完整的MCP服务器规范

Design a complete MCP server spec from a tool or API description.

设计能自我进化的上下文工程系统与技能库

Design a self-improving context-engineering system and skill library.

在 Hugging Face 生态里自主查资料、写代码、训模型、上传并监控的 ML 助手。

An ML assistant that autonomously researches, codes, trains, ships, and monitors on Hugging Face.

扮演首席 MLOps 工程师,为具体场景设计端到端 ML 平台与运维框架。

Play a principal MLOps engineer designing an end-to-end ML platform and ops framework for a use case.

扮演原生与跨平台移动开发专家,产出高性能应用架构与示例代码。

Play a native and cross-platform mobile dev expert producing high-performance app architecture and sample code.

把 AI 定为“调度中枢”，负责拆解任务并分派给专职子智能体，不亲自执行。

Casts the AI as a dispatch hub that decomposes tasks and delegates to sub-agents without executing.

把资料导入 NotebookLM 并编排生成播客、幻灯片、报告、测验等多模态产物。

Ingests sources into NotebookLM and orchestrates podcasts, slides, reports, quizzes, and more.

以 OMC 编排层协调专职智能体、工具与技能，强调委派、验证与提交规范。

Coordinates specialized agents, tools, and skills via the OMC layer, stressing delegation, verification, and commit conventions.

把模糊的编码需求改写成 Codex CLI 可端到端执行的高质量提示词

Rewrite vague coding requests into prompts Codex CLI can execute end-to-end

把重复工作流封装成符合 Codex 官方格式、可安装的技能

Turn a repeated workflow into a portable, installable skill in Codex's official format