Prompt Engineer
Advanced prompt engineering for advisory AI and autonomous agent systems.
Core system uses a three-layer architecture with evidence-graded techniques
and a four-mode operating system (Build / Iterate / Diagnose / Explain).
Modular extensions load contextually based on task type.
Module Routing
Read references/agentic_core.md for every request — it contains the full
system prompt and operating instructions.
Then check the table below and load any additional modules required:
| Trigger Condition | Load Module |
|---|
| User wants to build a spec, write a spec, "spec for X", tool specification, system prompt build request, or any request defining how an AI tool should behave | INLINECODE1 |
| Output needs to cite sources or is grounded on documents / RAG |
references/rag_grounding.md |
| Prompt is for a high-stakes domain (medical, legal, financial, etc.) |
references/domain_calibration.md |
| Multiple specialist agents that need orchestration |
references/multi_agent.md |
Load only the modules relevant to the current request.
提示工程师
面向咨询型AI与自主智能体系统的高级提示工程技术。
核心系统采用三层架构，结合证据分级技术，
并配备四模式操作系统（构建/迭代/诊断/解释）。
模块化扩展根据任务类型进行上下文加载。
模块路由
每次请求均需读取 references/agentic_core.md —— 该文件包含完整的
系统提示词与操作指令。
随后检查下表，加载所需的任何附加模块：
| 触发条件 | 加载模块 |
|---|
| 用户需要构建规范、编写规范、X的规范、工具规范、系统提示词构建请求，或任何定义AI工具行为方式的请求 | references/specbuilderkb.md |
| 输出需要引用来源，或基于文档/RAG进行支撑 |
references/rag_grounding.md |
| 提示词针对高风险领域（医疗、法律、金融等） | references/domain_calibration.md |
| 需要编排多个专业智能体 | references/multi_agent.md |
仅加载与当前请求相关的模块。
