research-workspace-standard

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Research Workspace Standard

研究工作区标准

This skill establishes a unified environment for research data, enabling agents (Researcher, Writer, Reviewer) to collaborate using either a structured physical directory or a "Smart Discovery" virtual map of an existing codebase.

本Skill为研究数据构建了统一环境，支持智能体（Researcher、Writer、Reviewer）通过结构化实体目录或现有代码库的“智能发现”虚拟地图开展协作。

🎯 Triggers

🎯 触发指令

"khởi tạo thư mục nghiên cứu"
"setup research workspace"
"organize my PhD files"
"đọc cấu trúc codebase để nghiên cứu"
"map existing docs to research"

"khởi tạo thư mục nghiên cứu"
"设置研究工作区"
"整理我的博士文件"
"đọc cấu trúc codebase để nghiên cứu"
"将现有文档映射到研究工作"

🏛 Workspace Modes

🏛 工作区模式

Mode 1: Physical (Greenfield)

模式1：实体模式（全新项目）

...

Mode 2: Virtual (Brownfield & Monorepo)

模式2：虚拟模式（存量项目与单体仓库）

Use this for existing codebases or Monorepos. Create a WORKSPACE.md
at the root to map logical research roles.

适用于现有代码库或单体仓库（Monorepo）。在根目录创建**

WORKSPACE.md

**文件，用于映射逻辑研究角色。

📦 Monorepo & Submodule Support

📦 单体仓库（Monorepo）与子模块支持

If the project uses submodules or a monorepo structure (e.g., Nx, Turborepo), the

WORKSPACE.md

must explicitly map them:

markdown

undefined

如果项目采用子模块或单体仓库结构（如Nx、Turborepo），

WORKSPACE.md

必须明确映射这些结构：

markdown

undefined

Research Workspace Map

Core Engine:
```
packages/core/
```
,
```
.gitmodules:engine-v1
```
Sub-module Research:
```
submodules/research-utils/
```
Cross-Linkage:
```
../neighbor-project/research/notes/
```
(Relative path to related research)

undefined

Core Engine:
```
packages/core/
```
,
```
.gitmodules:engine-v1
```
Sub-module Research:
```
submodules/research-utils/
```
Cross-Linkage:
```
../neighbor-project/research/notes/
```
(Relative path to related research)

undefined

🔍 Smart Discovery Heuristics

🔍 智能发现规则

When in an existing codebase, the Agent must:

Locate Architecture: Search for C4 models, design docs, or
```
*.gitmodules
```
.
Identify Experiments: Scan for performance logs, benchmark scripts, or
```
tests/
```
.
Cross-Project Context: If a neighboring directory contains a
```
WORKSPACE.md
```
, read its State section to establish research continuity.

在现有代码库中，智能体必须执行以下操作：

定位架构： 搜索C4模型、设计文档或
```
*.gitmodules
```
文件。
识别实验： 扫描性能日志、基准测试脚本或
```
tests/
```
目录。
跨项目上下文： 如果相邻目录包含
```
WORKSPACE.md
```
，读取其State部分以确保研究连续性。

🔄 Artifact Contracts (I/O)

🔄 工件契约（输入/输出）

Regardless of the mode, skills communicate via these core artifacts (mapped in

WORKSPACE.md

sota-matrix.md
: Systematic comparison of current works (Output of
```
sota-survey
```
).
rq-statement.md
: Formal research questions (Output of
```
research-question
```
).
protocol.json
: Experimental setup (Output of
```
research-design
```
).
results.csv
: Raw results (Output of
```
experiment-tracking
```
).

无论采用哪种模式，Skill都通过

WORKSPACE.md

中映射的核心工件进行交互：

sota-matrix.md
：现有研究成果的系统性对比（
```
sota-survey
```
的输出）。
rq-statement.md
：正式研究问题（
```
research-question
```
的输出）。
protocol.json
：实验设置（
```
research-design
```
的输出）。
results.csv
：原始结果（
```
experiment-tracking
```
的输出）。

⚖️ Decision Hierarchy (Lex Specialis)

⚖️ 决策层级（特别法优先原则）

...

🔄 Session Evolution Protocol (Living Knowledge)

🔄 会话演进协议（动态知识）

WORKSPACE.md

is not static; it must evolve after every research session. The Agent is responsible for:

State Updates: Move the "Current Milestone" forward as tasks are completed.
Insight Logging: Add a
```
## 🧠 Recent Insights
```
section to
```
WORKSPACE.md
```
to capture new findings, rejected hypotheses, or architecture discoveries.
Dependency Mapping: If a new library, submodule, or external dataset is introduced, update the
```
🗺 Path Mapping
```
.
Closing Ritual: Before ending a session, the Agent MUST propose a diff for
```
WORKSPACE.md
```
summarizing:
- What was achieved.
- What new "Knowledge Artifacts" were created.
- What the next logical research step is.

WORKSPACE.md

并非静态文件，每次研究会话后都必须更新。智能体需负责以下事项：

状态更新： 随着任务完成，推进“当前里程碑”。
洞察记录： 在
```
WORKSPACE.md
```
中添加
```
## 🧠 近期洞察
```
部分，记录新发现、被否决的假设或架构探索结果。
依赖映射： 若引入新库、子模块或外部数据集，更新
```
🗺 路径映射
```
部分。
收尾流程： 会话结束前，智能体必须提交
```
WORKSPACE.md
```
的修改建议，总结：
- 已完成的工作。
- 新增的“知识工件”。
- 下一步合理的研究方向。

🚀 Initialization Workflow

🚀 初始化流程

Discovery: Scan root for
```
docs/
```
,
```
research/
```
,
```
tests/
```
, or
```
.gitmodules
```
.
Mapping: Create or update
```
WORKSPACE.md
```
with discovered paths and submodule roles.
Memory Linkage: Instruct sub-agents: "Read WORKSPACE.md for project paths and recent insights, then cross-reference with MEMORY.md for relevant lessons learned, adhering to the Hierarchy of Authority."
Evolution: Ensure
```
WORKSPACE.md
```
is updated with any new knowledge gained during the task execution.

发现： 扫描根目录下的
```
docs/
```
、
```
research/
```
、
```
tests/
```
或
```
.gitmodules
```
文件。
映射： 创建或更新
```
WORKSPACE.md
```
，填入发现的路径和子模块角色。
关联记忆： 告知子智能体：“阅读WORKSPACE.md获取项目路径和近期洞察，然后关联MEMORY.md中的相关经验教训，遵循权限层级规则。”
演进： 确保
```
WORKSPACE.md
```
更新了任务执行过程中获得的新知识。