diff --git a/docs/developer-manual/develop-guideline.md b/docs/developer-manual/develop-guideline.md
index 5cdd6f85..e684e7fe 100644
--- a/docs/developer-manual/develop-guideline.md
+++ b/docs/developer-manual/develop-guideline.md
@@ -1,4 +1,175 @@
+---
+title: Development Guideline
+---
 
-# 开发指南
+This page covers developing Kubedoop itself — the operators and the framework they share. For
+writing documentation see [Document Writing Guidelines](./document-guideline.md), and for the fork
+and pull request mechanics see [First Contribution](./first-commiter.md).
 
-TODO
+## What you would be working on
+
+Kubedoop is not one repository. Each piece lives under
+[github.com/zncdatadev](https://github.com/zncdatadev):
+
+| Repository | Contains |
+|------------|----------|
+| `operator-go` | The Go framework every operator builds on: shared CRD types, the reconciler, resource builders |
+| `<product>-operator` | One operator per product — `zookeeper-operator`, `hdfs-operator`, `trino-operator`, and so on |
+| `commons-operator`, `listener-operator`, `secret-operator` | The built-in operators every product cluster depends on |
+| `containers` | Product container images |
+| `kubedoop-helm-charts` | Published Helm charts |
+| `docs` | This site |
+
+Each operator is an independent Go module that depends on `operator-go`. A change to shared
+behaviour usually belongs in the framework; a change to one product's rendering belongs in that
+product's operator.
+
+## Prerequisites
+
+| Tool | Why |
+|------|-----|
+| Go | Building and testing. The required version is in each repository's `go.mod` |
+| Docker or Podman | Building operator images |
+| kubectl | Talking to a cluster |
+| [kind](https://kind.sigs.k8s.io/) | The local cluster end-to-end tests run against |
+| Helm | Installing operator dependencies during e2e |
+
+Everything else — `controller-gen`, `kustomize`, `setup-envtest`, `golangci-lint`, `chainsaw` — is
+downloaded into the repository's `bin/` directory by the Makefile on first use. Do not install those
+globally; a mismatched version is a source of failures that do not reproduce in CI.
+
+Kubernetes 1.29 is the supported floor.
+
+## Repository layout
+
+Operators follow the standard [Kubebuilder](https://book.kubebuilder.io/) layout, with a few
+Kubedoop additions:
+
+```text
+api/v1alpha1/       # CRD Go types — the API surface
+internal/           # Reconcilers, builders, product-specific logic
+cmd/                # Manager entrypoint
+config/             # Kustomize manifests, including generated CRDs
+deploy/helm/        # The operator's Helm chart
+test/e2e/           # Chainsaw end-to-end suites
+examples/           # Example custom resources
+Makefile            # Every task below
+PROJECT             # Kubebuilder project metadata
+```
+
+`make help` lists the targets in any operator repository.
+
+## The development loop
+
+```bash
+make manifests generate   # regenerate CRDs and deepcopy code after API changes
+make fmt vet              # format and vet
+make lint                 # golangci-lint
+make test                 # unit tests
+make build                # build the manager binary
+make run                  # run the controller against your current kubecontext
+```
+
+`make test` depends on `manifests generate fmt vet`, so it regenerates before running. If that
+produces a diff, commit it — generated files are checked in.
+
+## Changing the API
+
+API types live in `api/v1alpha1/`. After editing them:
+
+```bash
+make manifests generate
+```
+
+This regenerates the CRD YAML under `config/` and the `zz_generated.deepcopy.go` files. Both are
+committed, so a change to a type with no accompanying regeneration will show up as a diff in CI.
+
+The Helm chart carries its own copy of the CRDs and the operator's RBAC rules, and they do not
+update themselves:
+
+```bash
+make helm-crd-sync    # copy generated CRDs into deploy/helm/
+make helm-rbac-sync   # copy the generated ClusterRole rules into deploy/helm/
+```
+
+Forgetting these is the usual cause of a chart that installs an operator against a stale CRD.
+
+### Adding fields
+
+Prefer the shared structures in `operator-go`'s `pkg/apis/commons/v1alpha1` — resources, logging,
+affinity, PodDisruptionBudget — over defining a product-local equivalent. A field that ends up
+copied into three operators is a sign it belongs in the framework instead.
+
+Be careful with `+kubebuilder:default` on fields inside `config`. That block is folded from role to
+role group, and structural defaulting fills a leaf as soon as its enclosing object exists — so a
+default there makes "unset" indistinguishable from "explicitly this value", and the role's setting
+can never win. Several fields in the framework carry comments explaining exactly this; defaults for
+folded fields belong at consumption time.
+
+## Testing
+
+### Unit tests
+
+```bash
+make test
+```
+
+These run against [envtest](https://book.kubebuilder.io/reference/envtest.html) — a real API server
+and etcd, no scheduler or controllers. Enough to assert what the operator *renders*: that a given
+CR produces the StatefulSet, ConfigMap and Service you expect.
+
+### End-to-end tests
+
+End-to-end tests use [Chainsaw](https://kyverno.github.io/chainsaw/) against a kind cluster. Cases
+live in `test/e2e/<suite>/<case>/`, each a directory of YAML steps and assertions.
+
+```bash
+make setup-chainsaw-cluster   # create the kind cluster, install operator dependencies
+make setup-chainsaw-e2e       # build the image, load it, deploy the operator
+make chainsaw-e2e             # run the suites
+make cleanup-chainsaw-e2e     # undeploy
+```
+
+`setup-chainsaw-cluster` installs the operators the product depends on — `commons-operator`,
+`listener-operator` and `secret-operator` — from the published charts, because almost nothing works
+without them: secrets, listeners and pod enrichment are all provided by those three.
+
+Write an end-to-end case for anything that only shows up on a live cluster: rendered configuration
+a unit test would have to duplicate, rolling behaviour, discovery between two products. Keep
+assertions on the resources rather than on log output.
+
+## Working with operator-go
+
+Operators depend on a released `operator-go`. When a change needs both the framework and an operator
+at once, point the module at your local checkout while developing:
+
+```text
+replace github.com/zncdatadev/operator-go => ../operator-go
+```
+
+Remove the `replace` before opening the pull request — it cannot be merged. Land the framework
+change first, then bump the operator to the released version.
+
+## Before opening a pull request
+
+Run the same gates CI does, from the operator's repository:
+
+```bash
+make manifests generate   # and commit any resulting diff
+make fmt vet lint
+make test
+make helm-crd-sync helm-rbac-sync   # if the API changed
+```
+
+Then the end-to-end suite for anything touching rendering or reconcile behaviour.
+
+Follow the commit and branch conventions in [First Contribution](./first-commiter.md), and describe
+in the pull request what you actually ran. A checklist box ticked without running the command is
+worse than an honest note about what was skipped.
+
+## Related
+
+- [Collaboration Guide](./collaboration.md)
+- [First Contribution](./first-commiter.md)
+- [Document Writing Guidelines](./document-guideline.md)
+- [Project Status](./project-status.md)
diff --git a/i18n/zh/docusaurus-plugin-content-docs/current/developer-manual/develop-guideline.md b/i18n/zh/docusaurus-plugin-content-docs/current/developer-manual/develop-guideline.md
index 5cdd6f85..8d2b4efa 100644
--- a/i18n/zh/docusaurus-plugin-content-docs/current/developer-manual/develop-guideline.md
+++ b/i18n/zh/docusaurus-plugin-content-docs/current/developer-manual/develop-guideline.md
@@ -1,4 +1,169 @@
+---
+title: 开发指南
+---
 
-# 开发指南
+本页讲的是开发 Kubedoop 本身——各个 Operator 以及它们共用的框架。撰写文档请看
+[文档编写规范](./document-guideline.md)，fork 与 PR 的具体流程请看
+[第一次贡献](./first-commiter.md)。
 
-TODO
+## 你会改动哪些代码
+
+Kubedoop 不是单个仓库，各部分都在
+[github.com/zncdatadev](https://github.com/zncdatadev) 下：
+
+| 仓库 | 内容 |
+|------|------|
+| `operator-go` | 所有 Operator 共用的 Go 框架：共享 CRD 类型、调谐器、资源构建器 |
+| `<product>-operator` | 每个产品一个 Operator——`zookeeper-operator`、`hdfs-operator`、`trino-operator` 等 |
+| `commons-operator`、`listener-operator`、`secret-operator` | 所有产品集群都依赖的内置 Operator |
+| `containers` | 产品容器镜像 |
+| `kubedoop-helm-charts` | 已发布的 Helm chart |
+| `docs` | 本站点 |
+
+每个 Operator 都是独立的 Go module，依赖 `operator-go`。改动共性行为通常应该落在框架里，
+只影响某个产品渲染结果的改动才放进该产品的 Operator。
+
+## 前置条件
+
+| 工具 | 用途 |
+|------|------|
+| Go | 构建与测试。所需版本见各仓库的 `go.mod` |
+| Docker 或 Podman | 构建 Operator 镜像 |
+| kubectl | 与集群交互 |
+| [kind](https://kind.sigs.k8s.io/) | 端到端测试所用的本地集群 |
+| Helm | e2e 过程中安装 Operator 依赖 |
+
+其余工具——`controller-gen`、`kustomize`、`setup-envtest`、`golangci-lint`、`chainsaw`——
+由 Makefile 在首次使用时下载到仓库的 `bin/` 目录。不要全局安装它们，
+版本不一致会导致本地失败而 CI 无法复现（反之亦然）。
+
+Kubernetes 支持的最低版本是 1.29。
+
+## 仓库结构
+
+Operator 遵循标准 [Kubebuilder](https://book.kubebuilder.io/) 布局，另有几处 Kubedoop 特有的目录：
+
+```text
+api/v1alpha1/       # CRD Go 类型——对外 API 面
+internal/           # 调谐器、构建器、产品特有逻辑
+cmd/                # manager 入口
+config/             # Kustomize 清单，含生成的 CRD
+deploy/helm/        # 该 Operator 的 Helm chart
+test/e2e/           # Chainsaw 端到端用例
+examples/           # 示例自定义资源
+Makefile            # 下文所有任务
+PROJECT             # Kubebuilder 项目元数据
+```
+
+在任意 Operator 仓库执行 `make help` 可列出全部目标。
+
+## 开发循环
+
+```bash
+make manifests generate   # 改完 API 后重新生成 CRD 与 deepcopy 代码
+make fmt vet              # 格式化与静态检查
+make lint                 # golangci-lint
+make test                 # 单元测试
+make build                # 构建 manager 二进制
+make run                  # 用当前 kubecontext 在本机运行控制器
+```
+
+`make test` 依赖 `manifests generate fmt vet`，因此会先重新生成。如果生成产生了 diff，请提交它
+——生成物是纳入版本控制的。
+
+## 修改 API
+
+API 类型位于 `api/v1alpha1/`。修改后执行：
+
+```bash
+make manifests generate
+```
+
+这会重新生成 `config/` 下的 CRD YAML 和各处 `zz_generated.deepcopy.go`。两者都要提交，
+所以改了类型却没重新生成，会在 CI 里表现为 diff。
+
+Helm chart 自己保存了一份 CRD 和 Operator 的 RBAC 规则，它们**不会自动更新**：
+
+```bash
+make helm-crd-sync    # 把生成的 CRD 同步进 deploy/helm/
+make helm-rbac-sync   # 把生成的 ClusterRole 规则同步进 deploy/helm/
+```
+
+漏掉这两步，通常就是"chart 安装出来的 Operator 配着一份过期 CRD"的原因。
+
+### 新增字段
+
+优先复用 `operator-go` 的 `pkg/apis/commons/v1alpha1` 中已有的结构——资源、日志、亲和性、
+PodDisruptionBudget——而不是在产品侧另造一个等价物。一个字段如果被抄进了三个 Operator，
+就说明它本该属于框架。
+
+给 `config` 内部的字段加 `+kubebuilder:default` 要格外小心。该块会从角色折叠到角色组，
+而结构化默认值会在其**外层对象**一存在时就填充叶子字段——于是"未设置"和"显式设成该值"
+变得无法区分，角色级的取值就永远赢不了。框架里有若干字段的注释专门解释了这一点；
+折叠字段的默认值应当在消费时才施加。
+
+## 测试
+
+### 单元测试
+
+```bash
+make test
+```
+
+它们运行在 [envtest](https://book.kubebuilder.io/reference/envtest.html) 之上——
+一个真实的 API server 加 etcd，但没有调度器和内置控制器。这足以断言 Operator **渲染出了什么**：
+给定一个 CR，是否产出了预期的 StatefulSet、ConfigMap 和 Service。
+
+### 端到端测试
+
+端到端测试使用 [Chainsaw](https://kyverno.github.io/chainsaw/)，跑在 kind 集群上。
+用例位于 `test/e2e/<suite>/<case>/`，每个用例是一个装着 YAML 步骤与断言的目录。
+
+```bash
+make setup-chainsaw-cluster   # 创建 kind 集群，安装 Operator 依赖
+make setup-chainsaw-e2e       # 构建镜像、加载、部署 Operator
+make chainsaw-e2e             # 运行用例
+make cleanup-chainsaw-e2e     # 卸载
+```
+
+`setup-chainsaw-cluster` 会从已发布的 chart 安装产品所依赖的 Operator——`commons-operator`、
+`listener-operator` 和 `secret-operator`——因为缺了它们几乎什么都跑不起来：
+secret、listener 和 Pod 信息补全都由这三个提供。
+
+凡是只有在真实集群上才暴露的行为，都应该写端到端用例：单元测试需要重复实现一遍才能验证的渲染结果、
+滚动更新行为、两个产品之间的服务发现。断言请针对资源本身，而不是日志输出。
+
+## 与 operator-go 协同开发
+
+Operator 依赖的是已发布的 `operator-go`。当一处改动同时需要框架和 Operator 时，
+开发期间可以把 module 指向本地检出：
+
+```text
+replace github.com/zncdatadev/operator-go => ../operator-go
+```
+
+提 PR 之前务必移除这个 `replace`，带着它无法合并。正确顺序是先让框架的改动合入，
+再把 Operator 升到已发布的版本。
+
+## 提 PR 之前
+
+在 Operator 仓库里跑一遍 CI 会跑的那些门禁：
+
+```bash
+make manifests generate   # 并提交由此产生的 diff
+make fmt vet lint
+make test
+make helm-crd-sync helm-rbac-sync   # 如果改了 API
+```
+
+涉及渲染或调谐行为的改动，再跑一遍端到端用例。
+
+提交与分支命名请遵循[第一次贡献](./first-commiter.md)中的约定，并在 PR 里如实写明你实际跑了什么。
+勾了却没真跑的检查项，比诚实说明"这一项跳过了"更糟。
+
+## 相关内容
+
+- [协作指南](./collaboration.md)
+- [第一次贡献](./first-commiter.md)
+- [文档编写规范](./document-guideline.md)
+- [项目状态](./project-status.md)