improve agent part in the doc

Author: refraction-ray

docs/source/agentic.rst

@@ -80,10 +80,11 @@ This instructs the agent to clone the repository, read coding guidelines/skills,
     You are an AI coding assistant. Please follow these steps to set up our project workspace and solve my requests:
     
     1. Run `git clone https://github.com/tensorcircuit/tensorcircuit-ng.git .` to download the repository in this empty folder.
-    2. Read `AGENTS.md` at the root of the repository to understand the coding standards.
-    3. Review the available agentic workflow scripts and skill files under `.agents/skills/` to see if a specialized workflow is suitable for the current task.
-    4. Reference existing implementations in `examples/` and test cases in `tests/` to write differentiable, and JIT-friendly code.
-    5. Respond to my later requests, write code, and execute experiments/tests to analyse the results.
+    2. Install the git-cloned TensorCircuit-NG library locally in the appropriate Python environment. If unsure which Python environment is appropriate, ask the user.
+    3. Read `AGENTS.md` at the root of the repository to understand the coding standards.
+    4. Review the available agentic workflow scripts and skill files under `.agents/skills/` to see if a specialized workflow is suitable for the current task.
+    5. Reference existing implementations in `examples/` and test cases in `tests/` to write differentiable, and JIT-friendly code.
+    6. Respond to my later requests, write code, and execute experiments/tests to analyse the results.
     
     Let me know when the initialization is complete and ask for my requests.
 

docs/source/locale/zh_CN/LC_MESSAGES/agentic.po

@@ -9,7 +9,7 @@ msgid ""
 msgstr ""
 "Project-Id-Version: tensorcircuit-ng \n"
 "Report-Msgid-Bugs-To: \n"
-"POT-Creation-Date: 2026-06-10 09:49+0800\n"
+"POT-Creation-Date: 2026-06-15 16:24+0800\n"
 "PO-Revision-Date: 2026-03-31 13:30+0800\n"
 "Last-Translator: Gemini CLI\n"
 "Language: zh_CN\n"
@@ -177,10 +177,23 @@ msgid ""
 msgstr "通过将极致性能与自主的、意图驱动的 AI 工作流相结合，TensorCircuit-NG 助力研究人员从手动编码转向自动化的科学发现。"
 
 #: ../../source/agentic.rst:73
+msgid "Initial Prompt for Empty Workspace Setup"
+msgstr "空白工作区初始设置提示词"
+
+#: ../../source/agentic.rst:75
+msgid ""
+"When starting in a completely empty folder/workspace with an AI agent, "
+"you can copy and paste the following initial prompt to bootstrap the "
+"agent. This instructs the agent to clone the repository, read coding "
+"guidelines/skills, and execute tasks accordingly:"
+msgstr ""
+"当在完全空白的文件夹/工作区中使用 AI 智能体时，您可以复制并粘贴以下初始提示词以引导该智能体。这将指示智能体克隆仓库、阅读编码规范与技能文件，并相应地执行任务："
+
+#: ../../source/agentic.rst:93
 msgid "AI-Native Documentation"
 msgstr "AI 原生文档"
 
-#: ../../source/agentic.rst:75
+#: ../../source/agentic.rst:95
 msgid ""
 "One can also refer to AI-native docs for tensorcircuit-ng: `Devin "
 "Deepwiki <https://deepwiki.com/tensorcircuit/tensorcircuit-ng>`_, `Google"