diff --git a/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/Getting Started/rock-agent.md b/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/Getting Started/rock-agent.md
index bd2dd69b05..1898fbdcd0 100644
--- a/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/Getting Started/rock-agent.md	
+++ b/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/Getting Started/rock-agent.md	
@@ -4,70 +4,73 @@ sidebar_position: 4
 
 # Rock Agent 快速启动
 
-Rock Agent 是 ROCK 提供的 AI Agent 运行框架，支持在沙箱环境中运行各种类型的 Agent。
+ROCK 提供两种并列的 agent 使用能力,各自适用不同场景:
+
+- **Job**:通过 BashJob / HarborJob 在 sandbox 里跑一次 agent 评测/任务(典型基准:SWE-bench、Terminal Bench),是入门主要场景。
+- **install-agent**:直接在单个沙箱里安装并运行 agent,适合本地开发、单次调试。
+
+下面优先介绍 Job 用法,install-agent 用法见末尾或 [Install Agent in Sandbox (Experimental)](../References/Python%20SDK%20References/rock-agent.md)。
 
 ## 前置条件
 
-- 确保有可用的ROCK服务, 如果需要本地拉起服务端, 参考[快速启动](quickstart.md)
+- 确保有可用的 ROCK 服务,如果需要本地拉起服务端,参考[快速启动](quickstart.md)
 
-## 使用示例
+---
 
-ROCK 提供了两个Hello World Agent 示例，位于 `examples/agents/` 目录下：
+## 一、用 Job 运行 Agent
 
-```
-examples/agents/
-├── claude_code/      # ClaudeCode Agent 示例
-└── iflow_cli/        # IFlowCli Agent 示例
-```
+Job 有两种 backend:**Harbor Job** 用于运行 AI agent 基准评测任务(SWE-bench、Terminal Bench 等);**Bash Job** 用于在沙箱里跑自定义 shell 脚本。
 
-### 运行 IFlowCli 示例
+### 1.1 准备 yaml
 
-```bash
-cd examples/agents/iflow_cli
-python iflow_cli_demo.py
-```
+挑一类作为起点,直接复制对应模板:
 
-### 运行 ClaudeCode 示例
+- Harbor Job(Terminal Bench):[`examples/job/harbor/tb_job_config.yaml.template`](https://github.com/alibaba/ROCK/tree/master/examples/job/harbor/tb_job_config.yaml.template)
+- Bash Job(claw-eval):[`examples/job/bash/claw_eval/claw_eval_bashjob.yaml.template`](https://github.com/alibaba/ROCK/tree/master/examples/job/bash/claw_eval/claw_eval_bashjob.yaml.template)
 
-```bash
-cd examples/agents/claude_code
-python claude_code_demo.py
-```
+按模板填好对应字段即可。两类 Job 的完整字段说明见 [Use Job to Run Agent](../References/Python%20SDK%20References/job.md)。
 
-## IFlowCli 配置文件
+### 1.2 通过 Python SDK 启动
 
-配置文件位于 `examples/agents/iflow_cli/rock_agent_config.yaml`：
+```python
+import asyncio
+from rock.sdk.job import Job, JobConfig
 
-```yaml
-run_cmd: "iflow -p ${prompt} --yolo"
+async def main():
+    config = JobConfig.from_yaml("swe_job_config.yaml")
+    result = await Job(config).run()
 
-runtime_env_config:
-  type: node
-  npm_registry: "https://registry.npmmirror.com"
-  custom_install_cmd: "npm i -g @iflow-ai/iflow-cli@latest"
+    print(f"status={result.status}, score={result.score}")
+    for trial in result.trial_results:
+        print(f"  {trial.task_name}: score={trial.score} ({trial.status})")
 
-env:
-  IFLOW_API_KEY: "" # 填入你的 API Key
-  IFLOW_BASE_URL: "" # 填入你的 Base URL
-  IFLOW_MODEL_NAME: "" # 填入你的模型名称
+asyncio.run(main())
 ```
 
-## ClaudeCode 配置文件
+BashJob 的用法、完整字段说明、结果处理详见 [Use Job to Run Agent](../References/Python%20SDK%20References/job.md)。
 
-配置文件位于 `examples/agents/claude_code/rock_agent_config.yaml`：
+---
 
-```yaml
-run_cmd: "claude -p ${prompt}"
+## 二、install-agent:在沙箱里安装并运行 Agent
 
-runtime_env_config:
-  type: node
-  custom_install_cmd: "npm install -g @anthropic-ai/claude-code"
+适合本地开发或单次调试 agent 的场景,核心 API:
 
-env:
-  ANTHROPIC_BASE_URL: "" # 填入你的anthropic base url
-  ANTHROPIC_API_KEY: "" # 填入你的anthropic api key
+```python
+await sandbox.agent.install(config="rock_agent_config.yaml")
+result = await sandbox.agent.run(prompt="hello")
 ```
 
-## 相关文档
+`examples/install-agents/` 下提供了多个开箱即用的示例:
+
+- `examples/install-agents/iflow_cli/` — IFlowCli
+- `examples/install-agents/claude_code/` — Claude Code
+- `examples/install-agents/cursor_cli/`、`qwen_code/`、`swe_agent/`、`openclaw/` — 其他
+
+运行 Claude Code 示例:
+
+```bash
+cd examples/install-agents/claude_code
+python claude_code_demo.py
+```
 
-- [RockAgent 参考](../References/Python%20SDK%20References/rock-agent.md)
+完整 RockAgentConfig 字段说明、占位符语义、API 参考详见 [Install Agent in Sandbox (Experimental)](../References/Python%20SDK%20References/rock-agent.md)。
diff --git a/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/job.md b/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/job.md
new file mode 100644
index 0000000000..6dee1f678b
--- /dev/null
+++ b/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/job.md	
@@ -0,0 +1,150 @@
+# Use Job to Run Agent
+
+> 这是 ROCK 两种并列的 agent 使用能力中 **Job** 的参考文档,核心 API 是 `rock.sdk.job.Job` 与 `JobConfig`,用于在沙箱里跑一次 agent 评测/任务。有两种 backend:**Bash Job** 与 **Harbor Bench Job**。
+>
+> 另一种能力是在单个沙箱里安装并运行 agent,见 [Install Agent in Sandbox](./rock-agent.md)。两种能力使用各自独立的配置 schema,**不要互相套用**。
+
+`rock.sdk.job` 通过同一套 `Job` API 支持两种模式,通过配置类型区分:
+
+- **Bash Job**:在沙箱中运行自定义 Shell 脚本,适合数据处理、外部评测工具等
+- **Harbor Bench Job**:通过 Harbor 框架运行 AI agent 基准评测任务(SWE-bench、Terminal Bench 等)
+
+## 端到端示例
+
+最小可跑通的 Python 用法:
+
+```python
+import asyncio
+from rock.sdk.job import Job, JobConfig
+
+async def main():
+    config = JobConfig.from_yaml("swe_job_config.yaml")  # 含 agents: 与 datasets:
+    result = await Job(config).run()
+
+    print(f"status={result.status}, score={result.score}")
+    for trial in result.trial_results:
+        print(f"  {trial.task_name}: score={trial.score} ({trial.status})")
+
+asyncio.run(main())
+```
+
+完整 yaml 模板参考 `examples/job/harbor/swe_job_config.yaml.template`。
+
+---
+
+## Bash Job
+
+Bash Job 适用于在沙箱内执行任意 Shell 脚本的场景,例如运行评测工具、数据处理流程等。
+
+完整示例参考:[`examples/job/bash/claw_eval/`](https://github.com/alibaba/ROCK/tree/master/examples/job/bash/claw_eval)
+
+- `run_claw_eval.py` — 主入口,演示 `JobConfig.from_yaml()` + `Job(config).run()`
+- `claw_eval_bashjob.yaml.template` — YAML 配置模板,含 `script_path`、`environment`、`uploads`、`env` 等字段
+- `run_claw_eval.sh` — 沙箱内实际执行的脚本,演示 DinD 启动、日志写入和评分输出
+
+### BashJobConfig 配置字段
+
+| 字段 | 类型 | 默认值 | 说明 |
+|------|------|--------|------|
+| `script` | `str \| None` | `None` | 内联脚本内容,与 `script_path` 二选一 |
+| `script_path` | `str \| None` | `None` | 本地脚本文件路径,运行时读取并上传执行 |
+| `job_name` | `str` | 当前时间戳 | 任务名称,用于日志和产物路径区分 |
+| `environment` | `EnvironmentConfig` | — | 沙箱连接及资源配置,详见下表 |
+| `namespace` | `str \| None` | `None` | 命名空间 |
+| `experiment_id` | `str \| None` | `None` | 实验 ID |
+| `timeout` | `int` | `7200` | 整体超时秒数(2 小时) |
+
+**`environment` 常用字段:**
+
+| 字段 | 类型 | 说明 |
+|------|------|------|
+| `image` | `str` | 沙箱 Docker 镜像 |
+| `base_url` | `str` | ROCK 平台地址 |
+| `xrl_authorization` | `str` | 鉴权 Token |
+| `cluster` | `str` | 目标集群 |
+| `memory` | `str` | 内存大小(如 `"64g"`) |
+| `cpus` | `int` | CPU 核数 |
+| `auto_stop` | `bool` | 任务完成后是否自动停止沙箱 |
+| `uploads` | `list` | 本地文件/目录上传列表,格式:`[本地路径, 沙箱目标路径]` |
+| `env` | `dict[str, str]` | 注入沙箱会话的环境变量 |
+
+---
+
+## Harbor Bench Job
+
+Harbor Bench Job 适用于通过 Harbor 框架运行 AI agent 基准评测任务,如 SWE-bench、Terminal Bench 等。
+
+> **注意**:`rock.sdk.bench.Job` 已废弃,将在未来移除。请改用 `rock.sdk.job.Job` + `HarborJobConfig`。
+
+完整示例参考:[`examples/job/harbor/`](https://github.com/alibaba/ROCK/tree/master/examples/job/harbor)
+
+- `harbor_demo.py` — 主入口,演示 `JobConfig.from_yaml()` + `Job(config).run()` + 结果遍历
+- `swe_job_config.yaml.template` — SWE-bench 任务配置模板
+- `swe_job_config-verifier.yaml.template` — 附带 `verifier.mode: native` 的变体
+- `tb_job_config.yaml.template` — Terminal Bench 任务配置模板
+
+### HarborJobConfig 核心配置字段
+
+**基础字段:**
+
+| 字段 | 类型 | 默认值 | 说明 |
+|------|------|--------|------|
+| `experiment_id` | `str` | 必填 | 实验 ID,Harbor 中必须提供 |
+| `job_name` | `str \| None` | 自动生成 | 格式:`{dataset}_{task}_{uuid[:8]}` |
+| `namespace` | `str \| None` | `None` | 命名空间,从沙箱自动反填 |
+| `environment` | `RockEnvironmentConfig` | — | 沙箱连接及资源配置 |
+
+**执行控制字段:**
+
+| 字段 | 类型 | 默认值 | 说明 |
+|------|------|--------|------|
+| `n_attempts` | `int` | `1` | 每个 Trial 的尝试次数 |
+| `timeout` | `int` | `7200` | 整体超时秒数(自动从 agent_timeout 推算) |
+| `debug` | `bool` | `False` | 调试模式,保留更多中间产物 |
+
+**组件字段:**
+
+| 字段 | 类型 | 说明 |
+|------|------|------|
+| `agents` | `list[AgentConfig]` | Harbor 框架自身的 agent 配置(典型字段:`name`、`model_name`),完整字段见 `examples/job/harbor/swe_job_config.yaml.template` |
+| `datasets` | `list[DatasetConfig]` | 数据集配置列表 |
+| `verifier` | `VerifierConfig` | Verifier 评测配置 |
+| `orchestrator` | `OrchestratorConfig` | 并发调度配置 |
+
+---
+
+## 结果处理
+
+两种 Job 模式均返回 `JobResult`:
+
+```python
+result = await Job(config).run()
+
+print(f"status={result.status}, score={result.score}")
+for trial in result.trial_results:
+    print(f"  {trial.task_name}: score={trial.score} ({trial.status})")
+    if trial.exception_info:
+        print(f"    {trial.exception_info.exception_type}: {trial.exception_info.exception_message}")
+```
+
+### JobResult 字段
+
+| 字段 / 属性 | 类型 | 说明 |
+|------------|------|------|
+| `status` | `JobStatus` | 任务整体状态 |
+| `trial_results` | `list[TrialResult]` | 所有 Trial 结果列表 |
+| `score` | `float`(属性) | 所有 Trial `score` 的平均值 |
+| `n_completed` | `int`(属性) | 状态为 `completed` 的 Trial 数 |
+| `n_failed` | `int`(属性) | 状态为 `failed` 的 Trial 数 |
+
+### TrialResult 字段
+
+| 字段 / 属性 | 类型 | 说明 |
+|------------|------|------|
+| `task_name` | `str` | 任务名称 |
+| `exit_code` | `int` | 进程退出码 |
+| `raw_output` | `str` | 进程原始输出 |
+| `exception_info` | `ExceptionInfo \| None` | 若有异常则填充 |
+| `status` | `str`(属性) | `"completed"` 或 `"failed"` |
+| `duration_sec` | `float`(属性) | 执行耗时(秒) |
+| `score` | `float`(属性) | 评分(Bash Job 默认 `0.0`,Harbor 模式来自 verifier) |
diff --git a/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/rock-agent.md b/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/rock-agent.md
index c3f03b1efc..cd252750aa 100644
--- a/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/rock-agent.md	
+++ b/docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.7.x/References/Python SDK References/rock-agent.md	
@@ -1,8 +1,12 @@
-# Rock Agent（实验性）
+# Install Agent in Sandbox (Experimental)
 
-RockAgent 是 ROCK 框架中的核心 Agent 实现，直接继承自 `Agent` 抽象基类。它提供了完整的 Agent 生命周期管理，包括环境初始化、ModelService 集成、命令执行等功能。
+> 这是 ROCK 两种并列的 agent 使用能力中 **install-agent** 的参考文档,核心 API 是 `sandbox.agent.install()` 与 `sandbox.agent.run(prompt)`,用于在单个沙箱里安装并运行 agent。
+>
+> 另一种能力是用 Job 在沙箱里跑一次 agent 评测/任务,见 [Use Job to Run Agent](./job.md)。两种能力使用各自独立的配置 schema。
 
-使用 `sandbox.agent.install()` 以及 `sandbox.agent.run(prompt)` 就可以在 Rock 提供的 Sandbox 环境中安装和运行 Agent。
+RockAgent 是 ROCK 框架用来在沙箱中 install 自定义 agent 的能力,负责完整的 agent 生命周期管理:环境初始化、ModelService 集成、命令执行等。
+
+使用 `sandbox.agent.install()` 与 `sandbox.agent.run(prompt)` 就可以在 Rock 提供的 Sandbox 环境中安装和运行 Agent。
 
 ## 核心概念
 
@@ -185,10 +189,10 @@ model_service_config:                           # 具体参考 ModelService 有
 执行 Agent 任务。
 
 **执行流程**：
-1. 替换占位符, 准备Agent 运行命令
-4. 启动 agent 进程
-5. 如果启用 ModelService，启动 `watch_agent`
-6. 等待任务完成并返回结果
+1. 替换占位符,准备 Agent 运行命令
+2. 启动 agent 进程
+3. 如果启用 ModelService,启动 `watch_agent`
+4. 等待任务完成并返回结果
 
 ## 高级用法
 
@@ -281,10 +285,26 @@ model_service_config:
 
 ## 使用示例
 
-### 使用 YAML 配置文件（推荐）
+### 使用 YAML 配置文件(推荐)
 
 ```python
-# prepare a rock_agent_config.yaml
-await sandbox.agent.install(config="rock_agent_config.yaml")
-await sandbox.agent.run(prompt="hello")
+import asyncio
+from rock.sdk.sandbox import Sandbox, SandboxConfig
+
+async def main():
+    sandbox = Sandbox(SandboxConfig())
+    await sandbox.start()
+    try:
+        # rock_agent_config.yaml 与本文档「快速开始」中的示例一致
+        await sandbox.agent.install(config="rock_agent_config.yaml")
+        result = await sandbox.agent.run(prompt="hello")
+        print(result)
+    finally:
+        await sandbox.stop()
+
+asyncio.run(main())
 ```
+
+更多开箱即用的示例参见 `examples/install-agents/`(Claude Code、IFlowCli、Cursor CLI、Qwen Code、SWE-agent、OpenClaw 等)。
+
+如需通过 Job 跑 agent 评测/基准任务(另一条代码路径,有独立的配置 schema),见 [Use Job to Run Agent](./job.md)。
diff --git a/docs/versioned_docs/version-1.7.x/Getting Started/rock-agent.md b/docs/versioned_docs/version-1.7.x/Getting Started/rock-agent.md
index eb6b54a65b..19369c1aff 100644
--- a/docs/versioned_docs/version-1.7.x/Getting Started/rock-agent.md	
+++ b/docs/versioned_docs/version-1.7.x/Getting Started/rock-agent.md	
@@ -4,69 +4,73 @@ sidebar_position: 4
 
 # Rock Agent Quick Start
 
-Rock Agent is an AI Agent runtime framework provided by ROCK, supporting various types of Agents running in sandbox environments.
+ROCK provides two parallel ways to use agents, each suited to a different scenario:
+
+- **Job**: Run an agent evaluation/task in a sandbox via BashJob / HarborJob (typical benchmarks: SWE-bench, Terminal Bench) — the primary entry point.
+- **install-agent**: Install and run an agent directly inside a single sandbox — for local development and one-off debugging.
+
+Job is covered first. The install-agent section follows at the end, with full reference at [Install Agent in Sandbox (Experimental)](../References/Python%20SDK%20References/rock-agent.md).
 
 ## Prerequisites
-- Make sure you have a working ROCK service, if you need to locally start the service side, refer to [Quick Start](quickstart.md).
 
-## Examples
+- Make sure you have a working ROCK service. If you need to start the service locally, refer to [Quick Start](quickstart.md).
 
-ROCK provides two Hello World Agent examples in the `examples/agents/` directory:
+---
 
-```
-examples/agents/
-├── claude_code/      # ClaudeCode Agent example
-└── iflow_cli/        # IFlowCli Agent example
-```
+## 1. Use Job to Run Agent
 
-### Run IFlowCli Example
+Job has two backends: **Harbor Job** runs an AI agent benchmark task (SWE-bench, Terminal Bench, etc.); **Bash Job** runs a custom shell script inside a sandbox.
 
-```bash
-cd examples/agents/iflow_cli
-python iflow_cli_demo.py
-```
+### 1.1 Prepare a yaml
 
-### Run ClaudeCode Example
+Pick a starting point and copy the matching template:
 
-```bash
-cd examples/agents/claude_code
-python claude_code_demo.py
-```
+- Harbor Job (Terminal Bench): [`examples/job/harbor/tb_job_config.yaml.template`](https://github.com/alibaba/ROCK/tree/master/examples/job/harbor/tb_job_config.yaml.template)
+- Bash Job (claw-eval): [`examples/job/bash/claw_eval/claw_eval_bashjob.yaml.template`](https://github.com/alibaba/ROCK/tree/master/examples/job/bash/claw_eval/claw_eval_bashjob.yaml.template)
+
+Fill in the fields per the template. See [Use Job to Run Agent](../References/Python%20SDK%20References/job.md) for the full field reference of both backends.
 
-## IFlowCli Configuration File
+### 1.2 Launch via Python SDK
 
-The configuration file is located at `examples/agents/iflow_cli/rock_agent_config.yaml`:
+```python
+import asyncio
+from rock.sdk.job import Job, JobConfig
 
-```yaml
-run_cmd: "iflow -p ${prompt} --yolo"
+async def main():
+    config = JobConfig.from_yaml("swe_job_config.yaml")
+    result = await Job(config).run()
 
-runtime_env_config:
-  type: node
-  npm_registry: "https://registry.npmmirror.com"
-  custom_install_cmd: "npm i -g @iflow-ai/iflow-cli@latest"
+    print(f"status={result.status}, score={result.score}")
+    for trial in result.trial_results:
+        print(f"  {trial.task_name}: score={trial.score} ({trial.status})")
 
-env:
-  IFLOW_API_KEY: "" # Enter your API key
-  IFLOW_BASE_URL: "" # Enter your base URL
-  IFLOW_MODEL_NAME: "" # Enter your model name
+asyncio.run(main())
 ```
 
-## ClaudeCode Configuration File
+For BashJob usage, full field references, and result-handling details, see [Use Job to Run Agent](../References/Python%20SDK%20References/job.md).
 
-The configuration file is located at `examples/agents/claude_code/rock_agent_config.yaml`:
+---
 
-```yaml
-run_cmd: "claude -p ${prompt}"
+## 2. install-agent: Install and Run an Agent in a Sandbox
 
-runtime_env_config:
-  type: node
-  custom_install_cmd: "npm install -g @anthropic-ai/claude-code"
+For local development or debugging a single agent run, the core API is:
 
-env:
-  ANTHROPIC_BASE_URL: "" # Enter your anthropic base url
-  ANTHROPIC_API_KEY: "" # Enter your anthropic api key
+```python
+await sandbox.agent.install(config="rock_agent_config.yaml")
+result = await sandbox.agent.run(prompt="hello")
 ```
 
-## Related Documentation
+The `examples/install-agents/` directory ships ready-to-run examples:
+
+- `examples/install-agents/iflow_cli/` — IFlowCli
+- `examples/install-agents/claude_code/` — Claude Code
+- `examples/install-agents/cursor_cli/`, `qwen_code/`, `swe_agent/`, `openclaw/` — others
+
+Run the Claude Code example:
+
+```bash
+cd examples/install-agents/claude_code
+python claude_code_demo.py
+```
 
-- [RockAgent Reference](../References/Python%20SDK%20References/rock-agent.md)
+For full RockAgentConfig field details, placeholder semantics, and API reference, see [Install Agent in Sandbox (Experimental)](../References/Python%20SDK%20References/rock-agent.md).
diff --git a/docs/versioned_docs/version-1.7.x/References/Python SDK References/job.md b/docs/versioned_docs/version-1.7.x/References/Python SDK References/job.md
new file mode 100644
index 0000000000..a7dd7bcfd1
--- /dev/null
+++ b/docs/versioned_docs/version-1.7.x/References/Python SDK References/job.md	
@@ -0,0 +1,150 @@
+# Use Job to Run Agent
+
+> This is the reference for **Job**, one of ROCK's two parallel ways to use agents. Its core API is `rock.sdk.job.Job` with `JobConfig`, used to run an agent evaluation/task in a sandbox. Two backends are supported: **Bash Job** and **Harbor Bench Job**.
+>
+> The other way is to install and run an agent inside a single sandbox — see [Install Agent in Sandbox](./rock-agent.md). The two ways use distinct config schemas — **do not mix them**.
+
+`rock.sdk.job` exposes a single `Job` API that supports two modes, distinguished by the config type:
+
+- **Bash Job**: Runs an arbitrary shell script inside a sandbox — useful for data processing, external evaluation tools, etc.
+- **Harbor Bench Job**: Runs an AI agent benchmark task via the Harbor framework (SWE-bench, Terminal Bench, etc.).
+
+## End-to-End Example
+
+A minimal runnable Python snippet:
+
+```python
+import asyncio
+from rock.sdk.job import Job, JobConfig
+
+async def main():
+    config = JobConfig.from_yaml("swe_job_config.yaml")  # contains agents: and datasets:
+    result = await Job(config).run()
+
+    print(f"status={result.status}, score={result.score}")
+    for trial in result.trial_results:
+        print(f"  {trial.task_name}: score={trial.score} ({trial.status})")
+
+asyncio.run(main())
+```
+
+The full yaml template is in `examples/job/harbor/swe_job_config.yaml.template`.
+
+---
+
+## Bash Job
+
+Bash Job is for running arbitrary shell scripts inside a sandbox — running an external evaluation tool, processing data, etc.
+
+Full example: [`examples/job/bash/claw_eval/`](https://github.com/alibaba/ROCK/tree/master/examples/job/bash/claw_eval)
+
+- `run_claw_eval.py` — Entry point demonstrating `JobConfig.from_yaml()` + `Job(config).run()`
+- `claw_eval_bashjob.yaml.template` — YAML template with `script_path`, `environment`, `uploads`, `env`, etc.
+- `run_claw_eval.sh` — The script that actually runs in the sandbox (DinD startup, log writing, score output)
+
+### BashJobConfig Fields
+
+| Field | Type | Default | Description |
+|------|------|---------|-------------|
+| `script` | `str \| None` | `None` | Inline script content (mutually exclusive with `script_path`) |
+| `script_path` | `str \| None` | `None` | Local script path; the file is read and uploaded at runtime |
+| `job_name` | `str` | current timestamp | Name used for log and artifact paths |
+| `environment` | `EnvironmentConfig` | — | Sandbox connection and resource config (see below) |
+| `namespace` | `str \| None` | `None` | Namespace |
+| `experiment_id` | `str \| None` | `None` | Experiment ID |
+| `timeout` | `int` | `7200` | Overall timeout in seconds (2 hours) |
+
+**Common `environment` fields:**
+
+| Field | Type | Description |
+|------|------|-------------|
+| `image` | `str` | Sandbox Docker image |
+| `base_url` | `str` | ROCK platform URL |
+| `xrl_authorization` | `str` | Auth token |
+| `cluster` | `str` | Target cluster |
+| `memory` | `str` | Memory size (e.g. `"64g"`) |
+| `cpus` | `int` | Number of CPUs |
+| `auto_stop` | `bool` | Whether to stop the sandbox after the job |
+| `uploads` | `list` | Local-to-sandbox file/dir uploads, format: `[local_path, sandbox_path]` |
+| `env` | `dict[str, str]` | Environment variables injected into the sandbox session |
+
+---
+
+## Harbor Bench Job
+
+Harbor Bench Job runs AI agent benchmark tasks like SWE-bench and Terminal Bench via the Harbor framework.
+
+> **Note**: `rock.sdk.bench.Job` is deprecated and will be removed in a future release. Use `rock.sdk.job.Job` + `HarborJobConfig` instead.
+
+Full example: [`examples/job/harbor/`](https://github.com/alibaba/ROCK/tree/master/examples/job/harbor)
+
+- `harbor_demo.py` — Entry point demonstrating `JobConfig.from_yaml()` + `Job(config).run()` + result iteration
+- `swe_job_config.yaml.template` — SWE-bench task config template
+- `swe_job_config-verifier.yaml.template` — Variant with `verifier.mode: native`
+- `tb_job_config.yaml.template` — Terminal Bench task config template
+
+### HarborJobConfig Core Fields
+
+**Basic fields:**
+
+| Field | Type | Default | Description |
+|------|------|---------|-------------|
+| `experiment_id` | `str` | required | Experiment ID — required by Harbor |
+| `job_name` | `str \| None` | auto-generated | Format: `{dataset}_{task}_{uuid[:8]}` |
+| `namespace` | `str \| None` | `None` | Namespace, auto-filled from the sandbox |
+| `environment` | `RockEnvironmentConfig` | — | Sandbox connection and resource config |
+
+**Execution control:**
+
+| Field | Type | Default | Description |
+|------|------|---------|-------------|
+| `n_attempts` | `int` | `1` | Attempts per Trial |
+| `timeout` | `int` | `7200` | Overall timeout (auto-derived from agent_timeout) |
+| `debug` | `bool` | `False` | Debug mode — keeps more intermediate artifacts |
+
+**Components:**
+
+| Field | Type | Description |
+|------|------|-------------|
+| `agents` | `list[AgentConfig]` | Harbor's own agent config (typical fields: `name`, `model_name`) — see `examples/job/harbor/swe_job_config.yaml.template` for the canonical shape |
+| `datasets` | `list[DatasetConfig]` | Dataset configs |
+| `verifier` | `VerifierConfig` | Verifier evaluation config |
+| `orchestrator` | `OrchestratorConfig` | Concurrency / scheduling config |
+
+---
+
+## Result Handling
+
+Both Job modes return a `JobResult`:
+
+```python
+result = await Job(config).run()
+
+print(f"status={result.status}, score={result.score}")
+for trial in result.trial_results:
+    print(f"  {trial.task_name}: score={trial.score} ({trial.status})")
+    if trial.exception_info:
+        print(f"    {trial.exception_info.exception_type}: {trial.exception_info.exception_message}")
+```
+
+### JobResult Fields
+
+| Field / Property | Type | Description |
+|------------------|------|-------------|
+| `status` | `JobStatus` | Overall task status |
+| `trial_results` | `list[TrialResult]` | List of all Trial results |
+| `score` | `float` (property) | Average `score` across all Trials |
+| `n_completed` | `int` (property) | Number of Trials with status `completed` |
+| `n_failed` | `int` (property) | Number of Trials with status `failed` |
+
+### TrialResult Fields
+
+| Field / Property | Type | Description |
+|------------------|------|-------------|
+| `task_name` | `str` | Task name |
+| `exit_code` | `int` | Process exit code |
+| `raw_output` | `str` | Raw process output |
+| `exception_info` | `ExceptionInfo \| None` | Populated if an exception occurred |
+| `status` | `str` (property) | `"completed"` or `"failed"` |
+| `duration_sec` | `float` (property) | Execution time in seconds |
+| `score` | `float` (property) | Score (Bash Job defaults to `0.0`; Harbor mode comes from the verifier) |
diff --git a/docs/versioned_docs/version-1.7.x/References/Python SDK References/rock-agent.md b/docs/versioned_docs/version-1.7.x/References/Python SDK References/rock-agent.md
index f24ade49ac..9cc184d46c 100644
--- a/docs/versioned_docs/version-1.7.x/References/Python SDK References/rock-agent.md	
+++ b/docs/versioned_docs/version-1.7.x/References/Python SDK References/rock-agent.md	
@@ -1,6 +1,10 @@
-# Rock Agent (Experimental)
+# Install Agent in Sandbox (Experimental)
 
-RockAgent is the core Agent implementation in the ROCK framework, directly inheriting from the `Agent` abstract base class. It provides complete Agent lifecycle management, including environment initialization, ModelService integration, command execution, and more.
+> This is the reference for **install-agent**, one of ROCK's two parallel ways to use agents. Its core API is `sandbox.agent.install()` and `sandbox.agent.run(prompt)`, used to install and run an agent inside a single sandbox.
+>
+> The other way is to run an agent evaluation/task via Job — see [Use Job to Run Agent](./job.md). The two ways use distinct config schemas.
+
+RockAgent is the ROCK framework's mechanism for installing a custom agent inside a sandbox. It manages the full agent lifecycle — environment initialization, ModelService integration, command execution, and so on.
 
 Using `sandbox.agent.install()` and `sandbox.agent.run(prompt)`, you can install and run Agents in the Sandbox environment provided by Rock.
 
@@ -284,7 +288,23 @@ model_service_config:
 ### Using YAML Configuration File (Recommended)
 
 ```python
-# prepare a rock_agent_config.yaml
-await sandbox.agent.install(config="rock_agent_config.yaml")
-await sandbox.agent.run(prompt="hello")
+import asyncio
+from rock.sdk.sandbox import Sandbox, SandboxConfig
+
+async def main():
+    sandbox = Sandbox(SandboxConfig())
+    await sandbox.start()
+    try:
+        # rock_agent_config.yaml matches the examples in "Quick Start" above
+        await sandbox.agent.install(config="rock_agent_config.yaml")
+        result = await sandbox.agent.run(prompt="hello")
+        print(result)
+    finally:
+        await sandbox.stop()
+
+asyncio.run(main())
 ```
+
+More ready-to-run examples are in `examples/install-agents/` (Claude Code, IFlowCli, Cursor CLI, Qwen Code, SWE-agent, OpenClaw, etc.).
+
+To run an agent evaluation/benchmark task via Job (a different code path with its own config schema), see [Use Job to Run Agent](./job.md).
diff --git a/examples/evaluation/README.md b/examples/evaluation/README.md
new file mode 100644
index 0000000000..291fe57697
--- /dev/null
+++ b/examples/evaluation/README.md
@@ -0,0 +1,19 @@
+# evaluation
+
+End-to-end evaluation demos that combine sandbox lifecycle, agent install/run, and a test suite — useful for understanding how individual pieces fit together at the script level.
+
+## Layout
+
+| Subdir | Path | Description |
+|--------|------|-------------|
+| [`swe_bench/`](./swe_bench/) | install-agent | Single-task SWE-bench Verified demo: starts a sandbox, installs an agent via `sandbox.agent.install()`, runs the agent on the task, runs the test suite, parses the result |
+
+## When to use this vs `job/harbor/`
+
+| | `evaluation/swe_bench/` | [`job/harbor/`](../job/harbor/) |
+|--|------------------------|-------------------------------|
+| Path | install-agent | Job (Harbor) |
+| When | Debugging task setup or test parsing — full pipeline visible in script form | Production benchmark runs through the Harbor framework |
+| API | `Sandbox` + `sandbox.agent.install()` | `Job(JobConfig.from_yaml(...)).run()` |
+
+If you're running SWE-bench through the standard pipeline, prefer [`job/harbor/`](../job/harbor/).
diff --git a/examples/install-agents/README.md b/examples/install-agents/README.md
new file mode 100644
index 0000000000..951430d556
--- /dev/null
+++ b/examples/install-agents/README.md
@@ -0,0 +1,28 @@
+# install-agents
+
+Examples for the **install-agent** way of using ROCK: install and run an agent inside a single sandbox via `sandbox.agent.install()` + `sandbox.agent.run(prompt)`.
+
+To run an agent evaluation/benchmark task via Job, see [`../job/`](../job/) instead.
+
+## Layout
+
+| Subdir | Agent runtime |
+|--------|---------------|
+| [`claude_code/`](./claude_code/) | Anthropic Claude Code CLI (`@anthropic-ai/claude-code`) |
+| [`cursor_cli/`](./cursor_cli/) | Cursor CLI |
+| [`iflow_cli/`](./iflow_cli/) | iFlow CLI (`@iflow-ai/iflow-cli`) |
+| [`openclaw/`](./openclaw/) | OpenClaw — admin/proxy split-mode demo, has its own README |
+| [`qwen_code/`](./qwen_code/) | qwen-code (`@qwen-code/qwen-code`) |
+| [`swe_agent/`](./swe_agent/) | SWE-agent (`pip install -e` from GitHub) |
+
+Each subdir contains a `*_demo.py` entry point and a `rock_agent_config.yaml` driving the install/run.
+
+## Run
+
+```bash
+# pick any subdir
+cd iflow_cli
+python iflow_cli_demo.py
+```
+
+See the [Install Agent in Sandbox (Experimental)](../../docs/versioned_docs/version-1.7.x/References/Python%20SDK%20References/rock-agent.md) reference for the full RockAgentConfig schema.
diff --git a/examples/agents/claude_code/claude_code_demo.py b/examples/install-agents/claude_code/claude_code_demo.py
similarity index 100%
rename from examples/agents/claude_code/claude_code_demo.py
rename to examples/install-agents/claude_code/claude_code_demo.py
diff --git a/examples/agents/claude_code/rock_agent_config.yaml b/examples/install-agents/claude_code/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/claude_code/rock_agent_config.yaml
rename to examples/install-agents/claude_code/rock_agent_config.yaml
diff --git a/examples/agents/cursor_cli/cursor_cli_demo.py b/examples/install-agents/cursor_cli/cursor_cli_demo.py
similarity index 100%
rename from examples/agents/cursor_cli/cursor_cli_demo.py
rename to examples/install-agents/cursor_cli/cursor_cli_demo.py
diff --git a/examples/agents/cursor_cli/rock_agent_config.yaml b/examples/install-agents/cursor_cli/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/cursor_cli/rock_agent_config.yaml
rename to examples/install-agents/cursor_cli/rock_agent_config.yaml
diff --git a/examples/agents/iflow_cli/iflow_cli_demo.py b/examples/install-agents/iflow_cli/iflow_cli_demo.py
similarity index 100%
rename from examples/agents/iflow_cli/iflow_cli_demo.py
rename to examples/install-agents/iflow_cli/iflow_cli_demo.py
diff --git a/examples/agents/iflow_cli/integration_with_model_service/local/local_demo.py b/examples/install-agents/iflow_cli/integration_with_model_service/local/local_demo.py
similarity index 100%
rename from examples/agents/iflow_cli/integration_with_model_service/local/local_demo.py
rename to examples/install-agents/iflow_cli/integration_with_model_service/local/local_demo.py
diff --git a/examples/agents/iflow_cli/integration_with_model_service/local/rock_agent_config.yaml b/examples/install-agents/iflow_cli/integration_with_model_service/local/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/iflow_cli/integration_with_model_service/local/rock_agent_config.yaml
rename to examples/install-agents/iflow_cli/integration_with_model_service/local/rock_agent_config.yaml
diff --git a/examples/agents/iflow_cli/integration_with_model_service/proxy/proxy_demo.py b/examples/install-agents/iflow_cli/integration_with_model_service/proxy/proxy_demo.py
similarity index 100%
rename from examples/agents/iflow_cli/integration_with_model_service/proxy/proxy_demo.py
rename to examples/install-agents/iflow_cli/integration_with_model_service/proxy/proxy_demo.py
diff --git a/examples/agents/iflow_cli/integration_with_model_service/proxy/rock_agent_config.yaml b/examples/install-agents/iflow_cli/integration_with_model_service/proxy/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/iflow_cli/integration_with_model_service/proxy/rock_agent_config.yaml
rename to examples/install-agents/iflow_cli/integration_with_model_service/proxy/rock_agent_config.yaml
diff --git a/examples/agents/iflow_cli/rock_agent_config.yaml b/examples/install-agents/iflow_cli/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/iflow_cli/rock_agent_config.yaml
rename to examples/install-agents/iflow_cli/rock_agent_config.yaml
diff --git a/examples/agents/openclaw/REAMDE.md b/examples/install-agents/openclaw/README.md
similarity index 98%
rename from examples/agents/openclaw/REAMDE.md
rename to examples/install-agents/openclaw/README.md
index 707c87f636..89f75685cc 100644
--- a/examples/agents/openclaw/REAMDE.md
+++ b/examples/install-agents/openclaw/README.md
@@ -52,7 +52,7 @@ rock admin start --env local-proxy --role proxy --port 9001
 ## 4. Run the Demo
 
 ```bash
-cd examples/agents/openclaw
+cd examples/install-agents/openclaw
 python openclaw_demo.py
 ```
 
diff --git a/examples/agents/openclaw/openclaw.json b/examples/install-agents/openclaw/openclaw.json
similarity index 100%
rename from examples/agents/openclaw/openclaw.json
rename to examples/install-agents/openclaw/openclaw.json
diff --git a/examples/agents/openclaw/openclaw_demo.py b/examples/install-agents/openclaw/openclaw_demo.py
similarity index 100%
rename from examples/agents/openclaw/openclaw_demo.py
rename to examples/install-agents/openclaw/openclaw_demo.py
diff --git a/examples/agents/openclaw/rock_agent_config.yaml b/examples/install-agents/openclaw/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/openclaw/rock_agent_config.yaml
rename to examples/install-agents/openclaw/rock_agent_config.yaml
diff --git a/examples/agents/qwen_code/qwen_code_demo.py b/examples/install-agents/qwen_code/qwen_code_demo.py
similarity index 100%
rename from examples/agents/qwen_code/qwen_code_demo.py
rename to examples/install-agents/qwen_code/qwen_code_demo.py
diff --git a/examples/agents/qwen_code/rock_agent_config.yaml b/examples/install-agents/qwen_code/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/qwen_code/rock_agent_config.yaml
rename to examples/install-agents/qwen_code/rock_agent_config.yaml
diff --git a/examples/agents/swe_agent/rock_agent_config.yaml b/examples/install-agents/swe_agent/rock_agent_config.yaml
similarity index 100%
rename from examples/agents/swe_agent/rock_agent_config.yaml
rename to examples/install-agents/swe_agent/rock_agent_config.yaml
diff --git a/examples/agents/swe_agent/swe_agent_demo.py b/examples/install-agents/swe_agent/swe_agent_demo.py
similarity index 100%
rename from examples/agents/swe_agent/swe_agent_demo.py
rename to examples/install-agents/swe_agent/swe_agent_demo.py
diff --git a/examples/job/README.md b/examples/job/README.md
new file mode 100644
index 0000000000..ed591838fa
--- /dev/null
+++ b/examples/job/README.md
@@ -0,0 +1,16 @@
+# job
+
+Examples for the **Job** way of using ROCK: run an agent evaluation/task in a sandbox via `rock.sdk.job.Job` + `JobConfig`.
+
+For installing and running an agent inside a single sandbox, see [`../install-agents/`](../install-agents/) instead.
+
+## Layout
+
+| Subdir | Backend | Use it for |
+|--------|---------|-----------|
+| [`bash/`](./bash/) | `BashJobConfig` | Run an arbitrary shell script inside a sandbox (data processing, external evaluation tools) |
+| [`harbor/`](./harbor/) | `HarborJobConfig` | Run an AI agent benchmark task (SWE-bench, Terminal Bench, …) via the Harbor framework |
+
+Both backends share a single `Job(config).run()` entrypoint — pick the config type based on your scenario.
+
+See the [Use Job to Run Agent](../../docs/versioned_docs/version-1.7.x/References/Python%20SDK%20References/job.md) reference for the full schema.
diff --git a/examples/job/bash/README.md b/examples/job/bash/README.md
new file mode 100644
index 0000000000..aff46add5e
--- /dev/null
+++ b/examples/job/bash/README.md
@@ -0,0 +1,24 @@
+# job/bash
+
+`BashJob` examples: run an arbitrary shell script inside a sandbox.
+
+## Layout
+
+| File / dir | Form | Description |
+|------------|------|-------------|
+| [`simple_bash_job_demo.sh`](./simple_bash_job_demo.sh) | CLI | Minimal `rock job run --script-content ...` demo |
+| [`claw_eval/`](./claw_eval/) | Python SDK | `claw-eval` benchmark wrapped as a BashJob — uses `JobConfig.from_yaml()` + `Job(config).run()` |
+
+Both forms use the same underlying `BashJobConfig` schema; the CLI just wraps it.
+
+## Quick run
+
+```bash
+# CLI form
+bash simple_bash_job_demo.sh
+
+# SDK form
+cd claw_eval
+cp claw_eval_bashjob.yaml.template claw_eval_bashjob.yaml  # fill in real values
+python run_claw_eval.py
+```
diff --git a/examples/evaluation/claw_eval/claw_eval_bashjob.yaml.template b/examples/job/bash/claw_eval/claw_eval_bashjob.yaml.template
similarity index 100%
rename from examples/evaluation/claw_eval/claw_eval_bashjob.yaml.template
rename to examples/job/bash/claw_eval/claw_eval_bashjob.yaml.template
diff --git a/examples/evaluation/claw_eval/claw_eval_config.yaml.template b/examples/job/bash/claw_eval/claw_eval_config.yaml.template
similarity index 100%
rename from examples/evaluation/claw_eval/claw_eval_config.yaml.template
rename to examples/job/bash/claw_eval/claw_eval_config.yaml.template
diff --git a/examples/evaluation/claw_eval/run_claw_eval.py b/examples/job/bash/claw_eval/run_claw_eval.py
similarity index 95%
rename from examples/evaluation/claw_eval/run_claw_eval.py
rename to examples/job/bash/claw_eval/run_claw_eval.py
index f6beaefec7..21d8609c86 100644
--- a/examples/evaluation/claw_eval/run_claw_eval.py
+++ b/examples/job/bash/claw_eval/run_claw_eval.py
@@ -1,7 +1,7 @@
 """Run claw-eval via BashJob SDK.
 
 Usage:
-    cd examples/agents/claw_eval
+    cd examples/job/bash/claw_eval
     cp claw_eval_bashjob.yaml.template claw_eval_bashjob.yaml
     # fill in real values
     python run_claw_eval.py
diff --git a/examples/evaluation/claw_eval/run_claw_eval.sh b/examples/job/bash/claw_eval/run_claw_eval.sh
similarity index 100%
rename from examples/evaluation/claw_eval/run_claw_eval.sh
rename to examples/job/bash/claw_eval/run_claw_eval.sh
diff --git a/examples/bash/simple_bash_job_demo.sh b/examples/job/bash/simple_bash_job_demo.sh
similarity index 100%
rename from examples/bash/simple_bash_job_demo.sh
rename to examples/job/bash/simple_bash_job_demo.sh
diff --git a/examples/job/harbor/README.md b/examples/job/harbor/README.md
new file mode 100644
index 0000000000..061b7bdb54
--- /dev/null
+++ b/examples/job/harbor/README.md
@@ -0,0 +1,27 @@
+# job/harbor
+
+`HarborJob` examples: run an AI agent benchmark task via the Harbor framework.
+
+## Files
+
+| File | Purpose |
+|------|---------|
+| [`harbor_demo.py`](./harbor_demo.py) | Entry point — loads `JobConfig.from_yaml()`, runs `Job(config).run()`, iterates trial results |
+| [`swe_job_config.yaml.template`](./swe_job_config.yaml.template) | SWE-bench task config template |
+| [`swe_job_config-verifier.yaml.template`](./swe_job_config-verifier.yaml.template) | SWE-bench variant with `verifier.mode: native` |
+| [`tb_job_config.yaml.template`](./tb_job_config.yaml.template) | Terminal Bench task config template |
+
+## Quick run
+
+```bash
+# 1. copy a template and fill in real values
+cp swe_job_config.yaml.template swe_job_config.yaml
+
+# 2. set required env vars (OSS credentials, etc.) — see harbor_demo.py docstring
+source .env
+
+# 3. run
+python harbor_demo.py -c swe_job_config.yaml
+```
+
+The `agents:` block uses Harbor's own minimal schema (typical fields: `name`, `model_name`) — see the templates above for the canonical shape.
diff --git a/examples/harbor/harbor_demo.py b/examples/job/harbor/harbor_demo.py
similarity index 88%
rename from examples/harbor/harbor_demo.py
rename to examples/job/harbor/harbor_demo.py
index 09f48b3198..e6201acdb7 100644
--- a/examples/harbor/harbor_demo.py
+++ b/examples/job/harbor/harbor_demo.py
@@ -1,13 +1,11 @@
-"""Harbor benchmark demo using ROCK Job SDK (new path).
+"""Harbor benchmark demo using ROCK Job SDK.
 
 Uses ``rock.sdk.job.Job`` with ``HarborJobConfig`` — the recommended path
 with full feature parity (G1-G7 fixed) and scatter / multiple trial types.
 
-For the legacy path (``rock.sdk.bench.Job``), see ``harbor_demo_legacy.py``.
-
 Usage:
-    python examples/harbor/harbor_demo.py -c examples/harbor/swe.intern.yaml
-    python examples/harbor/harbor_demo.py -c examples/harbor/tb_job_config.yaml -t mailman
+    python examples/job/harbor/harbor_demo.py -c examples/job/harbor/swe.intern.yaml
+    python examples/job/harbor/harbor_demo.py -c examples/job/harbor/tb_job_config.yaml -t mailman
 
 Required environment variables (OSS_* are auto-forwarded into the sandbox):
     OSS_ACCESS_KEY_ID       Alibaba Cloud OSS access key ID
@@ -20,7 +18,7 @@
 Recommended setup:
     1. Copy .env.example to .env and fill in your credentials
     2. source .env
-    3. python examples/harbor/harbor_demo.py -c ...
+    3. python examples/job/harbor/harbor_demo.py -c ...
 """
 
 import argparse
diff --git a/examples/harbor/swe_job_config-verifier.yaml.template b/examples/job/harbor/swe_job_config-verifier.yaml.template
similarity index 100%
rename from examples/harbor/swe_job_config-verifier.yaml.template
rename to examples/job/harbor/swe_job_config-verifier.yaml.template
diff --git a/examples/harbor/swe_job_config.yaml.template b/examples/job/harbor/swe_job_config.yaml.template
similarity index 100%
rename from examples/harbor/swe_job_config.yaml.template
rename to examples/job/harbor/swe_job_config.yaml.template
diff --git a/examples/harbor/tb_job_config.yaml.template b/examples/job/harbor/tb_job_config.yaml.template
similarity index 100%
rename from examples/harbor/tb_job_config.yaml.template
rename to examples/job/harbor/tb_job_config.yaml.template