Add local docs web app and Codex skills

.claude/skills/mujoco-cpp-build/SKILL.md

@@ -0,0 +1,16 @@
+---
+name: mujoco-cpp-build
+description: Use when building C++ examples, compiling simulation executables, configuring CMake, or resolving C++ building errors in the MuJoCo learning repository. It requires the agent to actively prompt the user for build requirements (source vs release path, target directories).
+---
+
+# MuJoCo C++ Build
+
+This is the Claude Code project-local adapter for the canonical skill at:
+
+`../../../skills/mujoco-cpp-build/SKILL.md`
+
+When this skill is loaded, read and follow the canonical skill instructions:
+
+- `../../../skills/mujoco-cpp-build/SKILL.md`
+
+Use repository-relative tutorial paths only. Do not store user-specific paths in this adapter.

.claude/skills/mujoco-engineering/SKILL.md

@@ -0,0 +1,17 @@
+---
+name: mujoco-engineering
+description: Use when an agent needs to implement, reproduce, debug, or extend MuJoCo projects using this repository's MJCF, Python, C++, ray casting, soft contact, sensors, rendering, viewer, force, or build examples.
+---
+
+# MuJoCo Engineering
+
+This is the Claude Code project-local adapter for the canonical skill at:
+
+`../../../skills/mujoco-engineering/SKILL.md`
+
+When this skill is loaded, read and follow the canonical skill instructions and its reference map:
+
+- `../../../skills/mujoco-engineering/SKILL.md`
+- `../../../skills/mujoco-engineering/references/tutorial-map.md`
+
+Use repository-relative tutorial paths only. Do not store user-specific paths in this adapter.

.claude/skills/mujoco-teaching/SKILL.md

@@ -0,0 +1,17 @@
+---
+name: mujoco-teaching
+description: Use when a user wants to learn MuJoCo concepts, MJCF modeling, Python/C++ MuJoCo APIs, sensors, rendering, contact, soft contact, ray casting, or asks conceptual/tutorial questions about this repository's MuJoCo lessons.
+---
+
+# MuJoCo Teaching
+
+This is the Claude Code project-local adapter for the canonical skill at:
+
+`../../../skills/mujoco-teaching/SKILL.md`
+
+When this skill is loaded, read and follow the canonical skill instructions and its reference map:
+
+- `../../../skills/mujoco-teaching/SKILL.md`
+- `../../../skills/mujoco-teaching/references/tutorial-map.md`
+
+Use repository-relative tutorial paths only. Do not store user-specific paths in this adapter.

.gitignore

@@ -22,8 +22,21 @@ extend/equality/.cache/
 
 extend/plugin/.cache/
 
+__pycache__/
+*.py[cod]
+.venv/
+*.egg-info/
+build/
+dist/
+.pytest_cache/
+.ruff_cache/
+.mypy_cache/
+.uv/
+uv.lock
+skills/*/.local/
+
 extend/soft_contact/C++/build/
 extend/soft_contact/C++/.cache/
 
 extend/touch/C++/build/
-extend/touch/C++/.cache/
+extend/touch/C++/.cache/

.opencode/skills/mujoco-cpp-build/SKILL.md

@@ -0,0 +1,18 @@
+---
+name: mujoco-cpp-build
+description: Use when building C++ examples, compiling simulation executables, configuring CMake, or resolving C++ building errors in the MuJoCo learning repository. It requires the agent to actively prompt the user for build requirements (source vs release path, target directories).
+metadata:
+  canonical-source: skills/mujoco-cpp-build
+---
+
+# MuJoCo C++ Build
+
+This is the OpenCode project-local adapter for the canonical skill at:
+
+`../../../skills/mujoco-cpp-build/SKILL.md`
+
+When this skill is loaded, read and follow the canonical skill instructions:
+
+- `../../../skills/mujoco-cpp-build/SKILL.md`
+
+Use repository-relative tutorial paths only. Do not store user-specific paths in this adapter.

.opencode/skills/mujoco-engineering/SKILL.md

@@ -0,0 +1,19 @@
+---
+name: mujoco-engineering
+description: Use when an agent needs to implement, reproduce, debug, or extend MuJoCo projects using this repository's MJCF, Python, C++, ray casting, soft contact, sensors, rendering, viewer, force, or build examples.
+metadata:
+  canonical-source: skills/mujoco-engineering
+---
+
+# MuJoCo Engineering
+
+This is the OpenCode project-local adapter for the canonical skill at:
+
+`../../../skills/mujoco-engineering/SKILL.md`
+
+When this skill is loaded, read and follow the canonical skill instructions and its reference map:
+
+- `../../../skills/mujoco-engineering/SKILL.md`
+- `../../../skills/mujoco-engineering/references/tutorial-map.md`
+
+Use repository-relative tutorial paths only. Do not store user-specific paths in this adapter.

.opencode/skills/mujoco-teaching/SKILL.md

@@ -0,0 +1,19 @@
+---
+name: mujoco-teaching
+description: Use when a user wants to learn MuJoCo concepts, MJCF modeling, Python/C++ MuJoCo APIs, sensors, rendering, contact, soft contact, ray casting, or asks conceptual/tutorial questions about this repository's MuJoCo lessons.
+metadata:
+  canonical-source: skills/mujoco-teaching
+---
+
+# MuJoCo Teaching
+
+This is the OpenCode project-local adapter for the canonical skill at:
+
+`../../../skills/mujoco-teaching/SKILL.md`
+
+When this skill is loaded, read and follow the canonical skill instructions and its reference map:
+
+- `../../../skills/mujoco-teaching/SKILL.md`
+- `../../../skills/mujoco-teaching/references/tutorial-map.md`
+
+Use repository-relative tutorial paths only. Do not store user-specific paths in this adapter.

.vscode/c_cpp_properties.json

@@ -0,0 +1,24 @@
+{
+  "configurations": [
+    {
+      "browse": {
+        "databaseFilename": "${workspaceFolder}/.vscode/browse.vc.db",
+        "limitSymbolsToIncludedHeaders": false
+      },
+      "includePath": [
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_hg/include/**",
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_go/include/**",
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_api/include/**",
+        "/home/albusgive2/rlfw/rlfw_interface/install/rlfw_msgs/include/**",
+        "/opt/ros/humble/include/**",
+        "/usr/include/**"
+      ],
+      "name": "ros2",
+      "intelliSenseMode": "gcc-x64",
+      "compilerPath": "/usr/bin/gcc",
+      "cStandard": "gnu11",
+      "cppStandard": "c++17"
+    }
+  ],
+  "version": 4
+}

.vscode/settings.json

@@ -2,5 +2,21 @@
     "cmake.sourceDirectory": "/home/albusgive2/mujoco_learning/CPP/Chapter1-make",
     "python-envs.defaultEnvManager": "ms-python.python:conda",
     "python-envs.defaultPackageManager": "ms-python.python:conda",
-    "python-envs.pythonProjects": []
+    "ROS2.distro": "humble",
+    "python.autoComplete.extraPaths": [
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_hg/local/lib/python3.10/dist-packages",
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_go/local/lib/python3.10/dist-packages",
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_api/local/lib/python3.10/dist-packages",
+        "/home/albusgive2/rlfw/rlfw_interface/install/rlfw_msgs/local/lib/python3.10/dist-packages",
+        "/opt/ros/humble/lib/python3.10/site-packages",
+        "/opt/ros/humble/local/lib/python3.10/dist-packages"
+    ],
+    "python.analysis.extraPaths": [
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_hg/local/lib/python3.10/dist-packages",
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_go/local/lib/python3.10/dist-packages",
+        "/home/albusgive2/unitree_ros2/cyclonedds_ws/install/unitree_api/local/lib/python3.10/dist-packages",
+        "/home/albusgive2/rlfw/rlfw_interface/install/rlfw_msgs/local/lib/python3.10/dist-packages",
+        "/opt/ros/humble/lib/python3.10/site-packages",
+        "/opt/ros/humble/local/lib/python3.10/dist-packages"
+    ]
 }

CPP/Chapter3-get_obj/tutorial.md

@@ -1,23 +1,107 @@
-# get obj
-## 获取名字
-数量
-![](../../MJCF/asset/entity.png)
-**先看mjModel结构体开头部分，这里的命名方式都是nXXX，这代表各个元素的数量**
-![](../../MJCF/asset/names.png)
-## 获取id
-`MJAPI int mj_name2id(const mjModel* m, int type, const char* name);`
-**通过name获取实体的id          
-m :mjModel
-type:   mjmodel.h文件中的mjtObj中定义，这个是要获取id的实体类型一下是部分type类型枚举，在mjtObj中找到       
-name: name
-**     
-![](../../MJCF/asset/enum_mjtobj.png)
-## 获取位置
-![](../../MJCF/asset/xpos.png)
-**可以通过 xpos和xxx_xpos获取各个对象的位置**
-## 获取姿态
-**通过xquat可以获取body的姿态**
+# 获取仿真世界中的实体信息 (Get Object Info)
 
+在 MuJoCo 中，获取仿真世界中各种实体（Body, Joint, Geom 等）的数量、ID、位置及姿态是非常高频的操作。以下是实现逻辑与核心数据结构。
 
+---
 
+## 1. 获取实体数量与名称 (Get Counts & Names)
+所有的数量与命名信息都存储在 `mjModel` 结构体中。
 
+### 1.1 实体数量定义
+在 `mjModel` 的开头，定义了所有仿真元素的数量（以 `n` 开头）：
+```cpp
+// 摘自 mujoco/mjmodel.h: 实体数量声明
+int       nbody;                // 身体 (Body) 数量
+int       njnt;                 // 关节 (Joint) 数量
+int       ngeom;                // 几何体 (Geom) 数量
+int       nsite;                // 标记点 (Site) 数量
+int       ncam;                 // 相机 (Camera) 数量
+int       nlight;               // 光源 (Light) 数量
+int       nmesh;                // 网格 (Mesh) 数量
+int       nsensor;              // 传感器 (Sensor) 数量
+int       nactuator;            // 驱动器 (Actuator) 数量
+```
+* 源码链接：[mujoco/mjmodel.h (sizes段)](https://github.com/google-deepmind/mujoco/blob/main/include/mujoco/mjmodel.h)
+
+### 1.2 实体名字缓冲区
+在 `mjModel` 中，名字均以一维字符缓冲区存储，并通过对应的地址数组记录每个实体的名字指针偏置：
+```cpp
+// 摘自 mujoco/mjmodel.h: 命名缓冲区
+char*     names;                // 名字缓存区，字符大数组 (nnames x 1)
+int*      name_bodyadr;         // 各 Body 名字在 names 中的起始索引 (nbody x 1)
+int*      name_jntadr;          // 各 Joint 名字在 names 中的起始索引 (njnt x 1)
+int*      name_geomadr;         // 各 Geom 名字在 names 中的起始索引 (ngeom x 1)
+int*      name_sensoradr;       // 各 Sensor 名字在 names 中的起始索引 (nsensor x 1)
+```
+* 源码链接：[mujoco/mjmodel.h (names段)](https://github.com/google-deepmind/mujoco/blob/main/include/mujoco/mjmodel.h)
+
+---
+
+## 2. 获取实体 ID (Get ID by Name)
+因为 MuJoCo 底层物理计算全部基于数组索引（即 ID），在 API 中我们通常需要通过名字（string）获取其 ID。
+使用 API 函数：
+```cpp
+MJAPI int mj_name2id(const mjModel* m, int type, const char* name);
+```
+
+### 2.1 参数说明
+* `m`: `mjModel` 指针。
+* `type`: 实体的类型，由枚举类型 `mjtObj` 定义。
+* `name`: 实体的名称。
+
+### 2.2 实体类型枚举 `mjtObj`
+`mjtObj` 定义在 `mjmodel.h` 中，代表各种可能的实体类型：
+```cpp
+typedef enum mjtObj_ {
+  mjOBJ_UNKNOWN        = 0,     // 未知对象类型
+  mjOBJ_BODY,                   // 身体 (Body)
+  mjOBJ_XBODY,                  // Body的替代表示，用于读取常规参考系 (而非惯性系)
+  mjOBJ_JOINT,                  // 关节 (Joint)
+  mjOBJ_DOF,                    // 自由度 (Degrees of freedom)
+  mjOBJ_GEOM,                   // 几何体 (Geom)
+  mjOBJ_SITE,                   // 标记点 (Site)
+  mjOBJ_CAMERA,                 // 相机 (Camera)
+  mjOBJ_LIGHT,                  // 光源 (Light)
+  mjOBJ_FLEX,                   // 柔性可变形体 (Flex)
+  mjOBJ_MESH,                   // 三角网格 (Mesh)
+  mjOBJ_SKIN,                   // 皮肤网格 (Skin)
+  mjOBJ_HFIELD,                 // 高度图 (Heightfield)
+  mjOBJ_TEXTURE,                // 纹理 (Texture)
+  mjOBJ_MATERIAL,               // 材质 (Material)
+  mjOBJ_PAIR,                   // 特殊碰撞 geom 对 (Pair)
+  mjOBJ_SENSOR                  // 传感器 (Sensor)
+} mjtObj;
+```
+* 源码链接：[mujoco/mjmodel.h (mjtObj定义)](https://github.com/google-deepmind/mujoco/blob/main/include/mujoco/mjmodel.h)
+
+演示代码：
+```cpp
+int body_id = mj_name2id(model, mjOBJ_BODY, "robot_link1");
+```
+
+---
+
+## 3. 获取位置与姿态 (Get Pos & Ori)
+物体的实时笛卡尔空间状态（三维位置、姿态矩阵等）全部存储在 **`mjData`** 中，由求解器在 `mj_step` 时进行实时更新更新：
+
+```cpp
+// 摘自 mujoco/mjdata.h: 笛卡尔坐标状态
+mjtNum*   xpos;                 // 各个 Body 的三维笛卡尔中心坐标 (nbody x 3)
+mjtNum*   xquat;                // 各个 Body 的笛卡尔四元数旋转姿态 (nbody x 4)
+mjtNum*   xmat;                 // 各个 Body 的笛卡尔旋转矩阵姿态 (nbody x 9)
+mjtNum*   xipos;                // 各个 Body 质心 (com) 的笛卡尔坐标 (nbody x 3)
+mjtNum*   ximat;                // 各个 Body 质心的笛卡尔旋转矩阵 (nbody x 9)
+```
+* 源码链接：[mujoco/mjdata.h (Cartesian state)](https://github.com/google-deepmind/mujoco/blob/main/include/mujoco/mjdata.h)
+
+演示代码：
+```cpp
+// 获取特定 body 实时三维位置坐标
+int body_id = mj_name2id(model, mjOBJ_BODY, "support");
+if (body_id != -1) {
+    double x = data->xpos[body_id * 3 + 0];
+    double y = data->xpos[body_id * 3 + 1];
+    double z = data->xpos[body_id * 3 + 2];
+    std::cout << "Body position: " << x << ", " << y << ", " << z << std::endl;
+}
+```

CPP/Chapter4-sensor_data/tutorial.md

@@ -1,9 +1,15 @@
 # 传感器数据获取
 
-![](../../MJCF/asset/sensor_adr.png)
-**sensordata的索引需要依靠mjData的sensor_adr获取，这个可以使用sensor的id**
-**这个我们要注意传感器具有的数据量，有的传感器是一个值，而有的传感器是三个值。我们可以使用mjModel中的sensor_dim获得传感器输出的参数量**
-<font color=Green>*演示：*</font>
+**`sensordata` 的索引需要依靠 `mjModel` 中的 `sensor_adr` 获取，也可以使用 sensor 的 id。**
+
+**我们要注意传感器具有的数据量，有的传感器是一个值，而有的传感器是三个值。我们可以使用 `mjModel` 中的 `sensor_dim` 获得传感器输出的参数量。**
+
+* `mjModel.sensor_adr` 的定义类似于：
+```cpp
+int*    sensor_adr;             // address in sensor array        (nsensor x 1)
+```
+
+演示：
 
 ```C++
 std::vector<float> get_sensor_data(const mjModel *model, const mjData *data,
@@ -21,8 +27,24 @@ std::vector<float> get_sensor_data(const mjModel *model, const mjData *data,
   return sensor_data;
 }
 ```
-对于传感器其他的属性在 mjModel中可以直接获得。如下：
-![](../../MJCF/asset/modelsensors.png)
+
+对于传感器的其他属性，在 `mjModel` 中可以直接获得：
+```cpp
+// mjModel 中的传感器相关成员变量说明
+int*      sensor_type;       // 传感器类型 (mjtSensor)          (nsensor x 1)
+int*      sensor_datatype;   // 数值数据类型 (mjtDataType)       (nsensor x 1)
+int*      sensor_needstage;  // 所需 Jun 算阶段 (mjtStage)       (nsensor x 1)
+int*      sensor_objtype;    // 被测物体的类型 (mjtObj)          (nsensor x 1)
+int*      sensor_objid;      // 被测物体的 ID                   (nsensor x 1)
+int*      sensor_reftype;    // 参考系物体的类型 (mjtObj)         (nsensor x 1)
+int*      sensor_refid;      // 参考系物体的 ID; -1: 全局参考系   (nsensor x 1)
+int*      sensor_dim;        // 标量输出的数量                  (nsensor x 1)
+int*      sensor_adr;        // 传感器数据在 sensordata 中的地址 (nsensor x 1)
+mjtNum*   sensor_cutoff;     // 实数/正数的截止值; 0: 忽略        (nsensor x 1)
+mjtNum*   sensor_noise;      // 噪声音频标准差                   (nsensor x 1)
+mjtNum*   sensor_user;       // 用户数据                        (nsensor x nuser_sensor)
+int*      sensor_plugin;     // 插件实例 ID; -1: 非插件          (nsensor x 1)
+```
 
 ### 读取相机画面
 &emsp;&emsp;相机来源一般是在模型文件中创建相机，或者创建一个相机手动控制，就像 base中