diff --git a/HOWTO.md b/HOWTO.md
new file mode 100644
index 0000000..7c5a02a
--- /dev/null
+++ b/HOWTO.md
@@ -0,0 +1,725 @@
+# docfill HOWTO
+
+本文档说明如何使用 `docfill` 完成本地文档自动填充。文档分为两部分：
+
+- “最小实践”用于快速完成一次可验证运行。
+- “详细说明”用于理解任务配置、运行变量、筛选、模板、输出、错误和日志。
+
+## 一、最小实践
+
+### 1. 进入项目目录
+
+```bash
+cd /Users/zhaoyilun/Documents/xls_doc
+```
+
+### 2. 构建命令行工具
+
+```bash
+bash scripts/build.sh
+```
+
+输出示例：
+
+```text
+/Users/zhaoyilun/Documents/xls_doc/dist/docfill-darwin-arm64
+```
+
+构建后的可执行文件是：
+
+```text
+dist/docfill-darwin-arm64
+```
+
+### 3. 准备测试素材
+
+仓库中已经包含 `manual_tests/` 测试素材。如果需要重新生成，可执行：
+
+```bash
+/Users/zhaoyilun/.cache/codex-runtimes/codex-primary-runtime/dependencies/python/bin/python3 scripts/make-manual-test-materials.py
+```
+
+输出示例：
+
+```text
+/Users/zhaoyilun/Documents/xls_doc/manual_tests
+```
+
+测试素材包含四类任务：
+
+- `manual_tests/01_excel_to_word`: Excel 源数据生成 Word 日报。
+- `manual_tests/02_sqlite_to_excel`: SQLite 源数据生成 Excel 汇总表。
+- `manual_tests/03_excel_to_excel`: Excel 源数据生成 Excel 质检表。
+- `manual_tests/04_filter_work_order_to_word`: 按日期和工单编号筛选一条工单，再生成 Word。
+
+### 4. 运行筛选工单案例
+
+命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002 --verbose
+```
+
+输出示例：
+
+```text
+log: load config path=manual_tests/04_filter_work_order_to_word/task.json
+log: read data type=excel path=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx
+log: read data type=excel path=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx rows=3
+log: filter rows=3 matched=1 expect=one
+log: render row=1 output=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+generated 1 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+```
+
+查看输出目录：
+
+```bash
+open manual_tests/04_filter_work_order_to_word/out
+```
+
+输入 Excel 内容示例：
+
+```text
+日期          工单编号   客户名称  系统来源   处理结果  下一步
+2026-06-18   GD-001    A 公司   PMS3.0   已派单   等待现场反馈
+2026-06-18   GD-002    B 公司   D5000    已完成   归档并同步日报
+2026-06-19   GD-001    C 公司   风控平台  待复核   补充截图
+```
+
+本次命令传入：
+
+```text
+date = 2026-06-18
+ticket_no = GD-002
+```
+
+所以工具会筛选：
+
+```text
+日期 = 2026-06-18，并且 工单编号 = GD-002
+```
+
+最终只生成 `GD-002` 对应的 Word 文件。
+
+## 二、详细说明
+
+### 1. task.json 与 --var 的关系
+
+`task.json` 是某一种业务报表的固定任务说明书。它通常由开发人员或维护人员准备好，普通运行时不需要频繁修改。
+
+命令行 `--var` 是本次运行传入的具体条件，例如日期、工单编号、人员、批次。
+
+可以这样理解：
+
+```text
+不同业务 = 不同 task.json
+本次运行的具体条件 = --var
+真实业务数据 = Excel 或 SQLite
+输出模板 = Word 或 Excel
+```
+
+例如不同业务可以分别维护：
+
+```text
+工单日报.json
+部门汇总.json
+质检报告.json
+```
+
+运行时只传本次变化的信息：
+
+```bash
+dist/docfill-darwin-arm64 run -c 工单日报.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+变量优先级：
+
+```text
+数据表里的字段 > 命令行 --var > task.json 里的 vars
+```
+
+示例：
+
+```json
+"vars": {
+  "date": "2026-06-18"
+}
+```
+
+如果运行时执行：
+
+```bash
+dist/docfill-darwin-arm64 run -c task.json --var date=2026-06-19
+```
+
+实际使用的是：
+
+```text
+date = 2026-06-19
+```
+
+也就是命令行 `--var` 会覆盖 `task.json` 中的同名默认变量。
+
+### 2. 命令格式
+
+`docfill` 当前只提供一个核心命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c <task.json> [--var key=value] [--verbose]
+```
+
+参数说明：
+
+| 参数 | 作用 | 示例 |
+| --- | --- | --- |
+| `run` | 执行一次填充任务 | `dist/docfill-darwin-arm64 run -c task.json` |
+| `-c` | 指定任务配置文件 | `-c manual_tests/04_filter_work_order_to_word/task.json` |
+| `--config` | `-c` 的完整写法 | `--config task.json` |
+| `--var` | 传入运行时变量，可以重复使用 | `--var date=2026-06-18 --var ticket_no=GD-002` |
+| `--verbose` | 输出诊断日志到 stderr | `--verbose` |
+
+命令行参数错误示例：
+
+```bash
+dist/docfill-darwin-arm64 run --var date
+```
+
+输出示例：
+
+```text
+docfill: [CLI_USAGE] invalid value "date" for flag -var: invalid --var "date", want key=value
+hint: Check command flags and use key=value for --var.
+```
+
+### 3. 任务配置结构
+
+典型的筛选任务配置如下：
+
+```json
+{
+  "vars": {
+    "date": "2026-06-18"
+  },
+  "data": {
+    "type": "excel",
+    "path": "./source_work_orders.xlsx",
+    "sheet": "工单",
+    "header_row": 1
+  },
+  "filter": {
+    "equals": {
+      "日期": "{{date}}",
+      "工单编号": "{{ticket_no}}"
+    },
+    "expect": "one"
+  },
+  "template": {
+    "path": "./work_order_template.docx"
+  },
+  "output": {
+    "path": "./out/work_order_{{日期}}_{{工单编号}}.docx",
+    "overwrite": true
+  }
+}
+```
+
+配置字段说明：
+
+| 字段 | 必填 | 说明 |
+| --- | --- | --- |
+| `vars` | 否 | 默认变量。适合放日期、批次、班次等默认值。 |
+| `data.type` | 是 | 数据源类型。支持 `excel`、`xlsx`、`sqlite`。 |
+| `data.path` | 是 | 数据源路径。支持 `.xlsx` 或 SQLite `.db`。 |
+| `data.sheet` | Excel 可选 | Excel 工作表名称。未填写时读取第一个工作表。 |
+| `data.header_row` | Excel 可选 | Excel 表头行号，从 `1` 开始。未填写时默认为 `1`。 |
+| `data.query` | SQLite 必填 | SQLite 查询语句。查询结果列名作为字段名。 |
+| `data.params` | SQLite 可选 | SQL `?` 参数列表，支持 `{{变量名}}`。 |
+| `filter.equals` | 否 | 等值筛选条件。左侧是数据字段，右侧是匹配值。 |
+| `filter.expect` | 否 | 筛选结果数量要求。支持 `one`、`many`，默认 `one`。 |
+| `template.path` | 是 | Word 或 Excel 模板路径。支持 `.docx`、`.xlsx`。 |
+| `output.path` | 是 | 输出文件路径模板。可以使用 `{{字段名}}`。 |
+| `output.overwrite` | 否 | 是否覆盖已存在文件。默认是 `false`。 |
+
+相对路径规则：
+
+```text
+task.json 所在目录就是相对路径的基准目录。
+```
+
+### 4. 业务筛选 filter
+
+`filter` 的作用是从数据表中筛选出本次要使用的数据行。它的目标不是提供数据库查询能力，而是提供符合表格使用习惯的简单筛选。
+
+配置示例：
+
+```json
+"filter": {
+  "equals": {
+    "日期": "{{date}}",
+    "工单编号": "{{ticket_no}}"
+  },
+  "expect": "one"
+}
+```
+
+含义：
+
+```text
+找到 日期 等于 date，并且 工单编号 等于 ticket_no 的数据行。
+```
+
+命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c task.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+实际筛选：
+
+```text
+日期 = 2026-06-18
+工单编号 = GD-002
+```
+
+筛选规则：
+
+- `filter.equals` 左侧必须是 Excel 表头或 SQLite 查询结果列名。
+- `filter.equals` 右侧可以写固定值，也可以写 `{{变量名}}`。
+- 多个条件之间是“全部满足”关系。
+- 比较前会去掉单元格值和筛选值两侧的空格。
+- 只支持等值匹配。
+- 不支持模糊匹配、范围查询、OR、排序表达式或复杂条件。
+
+`filter.expect` 支持两个值：
+
+| 值 | 含义 | 适用场景 |
+| --- | --- | --- |
+| `one` | 必须只匹配一行 | 工单、单份日报、单个报表。 |
+| `many` | 允许匹配多行 | 按日期批量生成多个人或多个部门的报表。 |
+
+未填写 `filter.expect` 时，默认按 `one` 处理。
+
+如果没有配置 `filter`，工具会沿用旧行为：读取到的每一行都会生成一个输出文件。
+
+复杂情况的处理建议：
+
+- 如果多个字段才能确定一条数据，就在 `filter.equals` 中写多个条件。
+- 如果表格本身不容易筛选，建议在 Excel 中新增辅助列，例如 `填报批次`、`业务唯一标识`、`筛选日期`。
+- 不建议把复杂业务规则写进 `docfill`，复杂规则应在数据准备阶段完成。
+
+### 5. Excel 数据源
+
+Excel 数据源配置示例：
+
+```json
+{
+  "data": {
+    "type": "excel",
+    "path": "./source_work_orders.xlsx",
+    "sheet": "工单",
+    "header_row": 1
+  }
+}
+```
+
+运行示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+Excel 读取规则：
+
+- `header_row` 指定表头行，表头单元格内容就是字段名。
+- 表头下方的每一行是候选数据。
+- 配置了 `filter` 时，先筛选，再生成输出。
+- 未配置 `filter` 时，每一行都会生成一个输出文件。
+- 空行会被跳过。
+- 空表头列会被忽略。
+
+表头在第 1 行的输入示例：
+
+```text
+日期          工单编号   客户名称  系统来源   处理结果  下一步
+2026-06-18   GD-001    A 公司   PMS3.0   已派单   等待现场反馈
+2026-06-18   GD-002    B 公司   D5000    已完成   归档并同步日报
+2026-06-19   GD-001    C 公司   风控平台  待复核   补充截图
+```
+
+表头在第 2 行的配置示例：
+
+```json
+{
+  "data": {
+    "type": "excel",
+    "path": "./source_audit_data.xlsx",
+    "sheet": "Audit",
+    "header_row": 2
+  }
+}
+```
+
+对应输入示例：
+
+```text
+日报质检导出
+
+employee  date        item        score  comment
+王五       2026-06-18  日报完整性    92     字段齐全，说明清楚
+赵六       2026-06-18  系统截图一致性  87     缺少 1 张截图，需要补充
+```
+
+### 6. SQLite 数据源
+
+SQLite 数据源配置示例：
+
+```json
+{
+  "vars": {
+    "date": "2026-06-18"
+  },
+  "data": {
+    "type": "sqlite",
+    "path": "./source_summary.db",
+    "query": "select department, report_date, total_count, abnormal_count, owner from department_summary order by department",
+    "params": []
+  },
+  "filter": {
+    "equals": {
+      "report_date": "{{date}}"
+    },
+    "expect": "many"
+  }
+}
+```
+
+运行示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/02_sqlite_to_excel/task.json --var date=2026-06-18
+```
+
+SQLite 读取规则：
+
+- `data.query` 的查询结果列名就是模板字段名。
+- `data.params` 按顺序对应 SQL 中的 `?`。
+- `data.params` 可以引用 `vars` 或命令行 `--var`。
+- 配置了 `filter` 时，SQL 查询结果还会再经过业务筛选。
+- 查询结果的每一行或筛选后的每一行生成一个输出文件。
+
+输入数据示例：
+
+```text
+department  report_date  total_count  abnormal_count  owner
+营销部        2026-06-17   115          2               张三
+客服部        2026-06-18   76           0               李四
+营销部        2026-06-18   128          3               张三
+```
+
+SQLite 可以继续使用 SQL 参数：
+
+```json
+"data": {
+  "type": "sqlite",
+  "path": "./source_summary.db",
+  "query": "select department, report_date, total_count, abnormal_count, owner from department_summary where report_date = ? order by department",
+  "params": ["{{date}}"]
+}
+```
+
+这适合由开发人员维护的 SQLite 数据源。面向表格使用习惯时，优先使用 `filter.equals` 表达业务筛选。
+
+### 7. Word 模板
+
+Word 模板使用 `.docx` 文件，模板中写入 `{{字段名}}`：
+
+```text
+日期：{{日期}}
+工单编号：{{工单编号}}
+客户名称：{{客户名称}}
+系统来源：{{系统来源}}
+处理结果：{{处理结果}}
+下一步：{{下一步}}
+```
+
+运行示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+生成结果示例：
+
+```text
+日期：2026-06-18
+工单编号：GD-002
+客户名称：B 公司
+系统来源：D5000
+处理结果：已完成
+下一步：归档并同步日报
+```
+
+Word 支持同一段落内被 Word 拆分的文本占位符。例如模板内部可能把 `{{name}}` 拆成相邻文本节点 `{{na` 和 `me}}`，工具会按同一段落进行最小兼容处理。
+
+这种兼容会把同一段落内的拆分文本合并后再写回，渲染结果会写入该段落第一个文本节点的位置，后续拆分节点的独立格式不再体现。建议模板中每个占位符使用同一种字体、颜色和加粗状态，不要在 `{{`、字段名、`}}` 之间混用格式。
+
+当前不支持复杂模板语法，例如表格循环、条件判断、图片插入、跨段落拼接占位符。
+
+### 8. Excel 模板
+
+Excel 模板使用 `.xlsx` 文件，单元格中写入 `{{字段名}}`：
+
+```text
+B2: {{department}}
+B3: {{report_date}}
+B4: {{owner}}
+B5: {{total_count}}
+B6: {{abnormal_count}}
+```
+
+运行示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/02_sqlite_to_excel/task.json --var date=2026-06-18
+```
+
+生成结果示例：
+
+```text
+B2: 营销部
+B3: 2026-06-18
+B4: 张三
+B5: 128
+B6: 3
+```
+
+Excel 模板会遍历所有工作表，替换单元格文本中的占位符。当前不支持公式重算、透视表刷新、复杂样式模板语法。
+
+### 9. 输出路径
+
+输出路径可以使用字段占位符：
+
+```json
+{
+  "output": {
+    "path": "./out/work_order_{{日期}}_{{工单编号}}.docx",
+    "overwrite": true
+  }
+}
+```
+
+运行示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+输出文件示例：
+
+```text
+work_order_2026-06-18_GD-002.docx
+```
+
+文件名清理规则：
+
+```text
+/ \ : * ? " < > |
+```
+
+以上字符会在输出路径字段值中替换为 `_`。例如：
+
+```text
+张三/测试 -> 张三_测试
+```
+
+覆盖规则：
+
+- `output.overwrite` 未设置或为 `false` 时，输出文件已存在会失败。
+- `output.overwrite` 为 `true` 时，允许覆盖同名输出文件。
+
+### 10. 日志、标准输出和退出码
+
+成功时，stdout 输出生成文件清单：
+
+```text
+generated 1 file(s)
+- /path/to/out/work_order_2026-06-18_GD-002.docx
+```
+
+启用 `--verbose` 后，stderr 输出诊断日志：
+
+```text
+log: load config path=task.json
+log: read data type=excel path=/path/data.xlsx
+log: read data type=excel path=/path/data.xlsx rows=3
+log: filter rows=3 matched=1 expect=one
+log: render row=1 output=/path/out/work_order_2026-06-18_GD-002.docx
+```
+
+退出码：
+
+| 退出码 | 含义 | 适用场景 |
+| --- | --- | --- |
+| `0` | 成功 | 可以读取 stdout 中的生成文件清单。 |
+| `1` | 执行失败 | 配置、数据、模板、输出等处理失败。 |
+| `2` | 命令行错误 | 参数缺失、命令错误、`--var` 格式错误。 |
+
+数字员工或调度脚本建议：
+
+- 判断退出码确认任务是否成功。
+- 成功时读取 stdout 获取生成文件路径。
+- 失败时记录 stderr，用于定位错误码、提示和上下文。
+- 仅在诊断或审计时启用 `--verbose`。
+
+### 11. 错误格式与错误码
+
+错误输出格式：
+
+```text
+docfill: [ERROR_CODE] message
+hint: repair suggestion
+context: key=value
+```
+
+配置文件不存在示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/not-exist.json
+```
+
+输出示例：
+
+```text
+docfill: [CONFIG_READ] read config: open manual_tests/not-exist.json: no such file or directory
+hint: Check that the config file exists and is readable.
+context: config=manual_tests/not-exist.json
+```
+
+筛选无匹配示例：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-999 --verbose
+```
+
+输出示例：
+
+```text
+log: load config path=manual_tests/04_filter_work_order_to_word/task.json
+log: read data type=excel path=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx
+log: read data type=excel path=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx rows=3
+docfill: [DATA_FILTER_NO_MATCH] filter matched no rows
+hint: Check filter.equals, --var values, or source table contents.
+context: expect=one filter=工单编号=GD-999,日期=2026-06-18
+```
+
+错误码说明：
+
+| 错误码 | 含义 | 常见处理方式 |
+| --- | --- | --- |
+| `CLI_USAGE` | 命令行参数错误 | 检查命令、`-c`、`--var key=value`。 |
+| `CONFIG_READ` | 配置文件读取失败 | 检查路径、文件是否存在、读权限。 |
+| `CONFIG_PARSE` | 配置 JSON 解析失败 | 检查 JSON 语法。 |
+| `CONFIG_VALIDATE` | 配置字段缺失或不支持 | 检查 `data.type`、`data.path`、`template.path`、`output.path`、`filter.expect`。 |
+| `DATA_EXCEL` | Excel 数据读取失败 | 检查 `.xlsx` 文件、工作表名、表头行。 |
+| `DATA_SQLITE` | SQLite 数据读取或查询失败 | 检查 `.db` 文件、SQL、参数。 |
+| `DATA_EMPTY` | 数据源没有返回可生成的数据行 | 检查源数据是否为空。 |
+| `DATA_FILTER_FIELD` | 筛选字段不在数据中 | 检查 Excel 表头或 SQLite 查询列名。 |
+| `DATA_FILTER_VALUE` | 筛选条件引用的变量没有提供 | 补充 `--var` 或 `vars`。 |
+| `DATA_FILTER_NO_MATCH` | 筛选后没有符合条件的数据行 | 检查筛选条件，或整理源表。 |
+| `DATA_FILTER_MULTI_MATCH` | 期望一行但匹配多行 | 增加筛选条件，或整理重复数据。 |
+| `TEMPLATE_OPEN` | 模板读取失败 | 检查模板路径和文件格式。 |
+| `TEMPLATE_FIELD_MISSING` | 模板字段缺失 | 在数据源、`vars` 或 `--var` 中补充字段。 |
+| `TEMPLATE_UNRESOLVED` | 渲染后仍残留占位符 | 检查模板占位符拼写和字段值。 |
+| `OUTPUT_EXISTS` | 输出文件已存在 | 修改输出路径或设置 `output.overwrite: true`。 |
+| `OUTPUT_WRITE` | 输出写入失败 | 检查输出目录权限、文件占用状态。 |
+| `INTERNAL` | 未分类内部错误 | 保留 stderr 和输入文件，开启 `--verbose` 复现。 |
+
+### 12. 自动化调用示例
+
+以下示例展示数字员工或调度脚本如何调用 `docfill`。该脚本只做最小封装，不承担模板解析或业务逻辑。
+
+```bash
+#!/usr/bin/env bash
+set -euo pipefail
+
+DOCFILL="/Users/zhaoyilun/Documents/xls_doc/dist/docfill-darwin-arm64"
+CONFIG="/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/task.json"
+DATE="${1:?date is required, example: 2026-06-18}"
+TICKET_NO="${2:?ticket_no is required, example: GD-002}"
+
+"$DOCFILL" run -c "$CONFIG" --var date="$DATE" --var ticket_no="$TICKET_NO" --verbose
+```
+
+执行：
+
+```bash
+bash run-docfill.sh 2026-06-18 GD-002
+```
+
+预期结果：
+
+```text
+generated 1 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+```
+
+集成建议：
+
+- 把业务系统接口采集结果先落成 Excel 或 SQLite。
+- 为每类报表维护一个 `task.json`。
+- 数字员工只负责准备运行变量并执行命令。
+- 复杂筛选优先通过整理 Excel 表解决，例如新增辅助列。
+- 不要把账号密码、审批状态、业务系统访问逻辑写入 `docfill` 配置。
+
+### 13. 开发与验证
+
+运行全部 Go 测试：
+
+```bash
+go test -count=1 ./...
+```
+
+运行冒烟测试：
+
+```bash
+bash scripts/smoke.sh
+```
+
+构建发布用本机二进制：
+
+```bash
+bash scripts/build.sh
+```
+
+重新生成手工测试素材：
+
+```bash
+/Users/zhaoyilun/.cache/codex-runtimes/codex-primary-runtime/dependencies/python/bin/python3 scripts/make-manual-test-materials.py
+```
+
+冒烟测试覆盖内容：
+
+- Go 单元测试。
+- 临时构建 `docfill`。
+- 生成临时 Excel、Word、SQLite 示例数据。
+- 执行无筛选 Excel 到 Word 示例。
+- 执行筛选 Excel 到 Word 示例。
+- 执行 SQLite 到 Excel 示例。
+- 检查输出文件是否生成。
+
+### 14. 当前边界
+
+为了保持工具小型、稳定、便于自动化调用，当前明确不包含以下能力：
+
+- 不提供 Web 界面。
+- 不提供模板设计器。
+- 不直接登录业务系统。
+- 不管理账号、密码或权限。
+- 不实现审批流。
+- 不实现 Word 表格循环或条件语法。
+- 不实现 Excel 公式重算和透视表刷新。
+- 不实现复杂筛选语法。
+- 不实现复杂任务编排。
+
+如需扩展，应优先通过外部脚本准备数据，再交给 `docfill` 执行单次文档填充。
diff --git a/README.md b/README.md
index 3458507..f5572e7 100644
--- a/README.md
+++ b/README.md
@@ -1,79 +1,151 @@
 # docfill
 
-`docfill` 是一个最小化本地填报工具：从 Excel 或 SQLite 读取数据，把 `{{字段名}}` 填入 Word/Excel 模板，生成输出文件。
+`docfill` 是一个本地文档填充命令行工具。它从 Excel 或 SQLite 读取结构化数据，按业务条件筛选出需要使用的数据行，再把数据填入 Word 或 Excel 模板中的 `{{字段名}}` 占位符。
 
-## Build
+本工具的定位是“小型、稳定、便于数字员工调用”。它只负责本地数据到本地模板的填充，不内置业务系统登录、审批流、Web 界面、模板设计器或权限系统。
+
+## 核心理解
+
+`task.json` 是某一种业务报表的固定任务说明书，命令行 `--var` 是本次运行传入的具体条件。
+
+```text
+不同业务 = 不同 task.json
+本次运行的日期、工单编号、人员、批次 = --var
+真实数据 = Excel 或 SQLite
+模板文件 = Word 或 Excel
+```
+
+优先级：
+
+```text
+数据表里的字段 > 命令行 --var > task.json 里的 vars
+```
+
+例如 `task.json` 中写了：
+
+```json
+"vars": {
+  "date": "2026-06-18"
+}
+```
+
+运行时执行：
+
+```bash
+dist/docfill-darwin-arm64 run -c task.json --var date=2026-06-19 --var ticket_no=GD-001
+```
+
+实际使用的是：
+
+```text
+date = 2026-06-19
+ticket_no = GD-001
+```
+
+也就是 `--var date=2026-06-19` 会覆盖 `task.json` 里的默认 `date=2026-06-18`。
+
+## 最小实践
+
+进入仓库目录：
+
+```bash
+cd /Users/zhaoyilun/Documents/xls_doc
+```
+
+构建本机二进制：
 
 ```bash
-go build -o docfill ./cmd/docfill
 bash scripts/build.sh
 ```
 
-`scripts/build.sh` writes the local macOS arm64 binary to `dist/docfill-darwin-arm64`.
-
-## Run
-
-```bash
-./docfill run --config task.json
-./docfill run -c task.json --var date=2026-06-18 --var batch=早班
-```
-
-成功时 stdout 输出生成文件清单：
+输出示例：
 
 ```text
-generated 2 file(s)
-- /path/to/out/report_张三.xlsx
-- /path/to/out/report_李四.xlsx
+/Users/zhaoyilun/Documents/xls_doc/dist/docfill-darwin-arm64
 ```
 
-退出码：
-
-- `0`: 成功。
-- `1`: 配置、数据、模板或输出失败。
-- `2`: 命令行参数错误。
-
-## Run Case Tests
+运行“按日期 + 工单编号筛选一条工单并生成 Word”的测试任务：
 
 ```bash
-go test -run TestCase -v ./internal/docfill
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002 --verbose
 ```
 
-当前案例覆盖：
+输出示例：
 
-- Excel 日报数据填 Excel 模板，批量生成个人日报。
-- SQLite 日报数据填 Word 模板，批量生成个人日报。
-- SQLite 汇总数据填 Excel 模板，批量生成部门汇总表。
+```text
+log: load config path=manual_tests/04_filter_work_order_to_word/task.json
+log: read data type=excel path=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx
+log: read data type=excel path=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx rows=3
+log: filter rows=3 matched=1 expect=one
+log: render row=1 output=/Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+generated 1 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+```
 
-## Smoke Test
+查看生成结果：
+
+```bash
+open manual_tests/04_filter_work_order_to_word/out
+```
+
+更多可运行案例见 [HOWTO.md](HOWTO.md) 和 [manual_tests/README.md](manual_tests/README.md)。
+
+## 常用命令
+
+构建：
+
+```bash
+bash scripts/build.sh
+```
+
+运行一个任务：
+
+```bash
+dist/docfill-darwin-arm64 run -c task.json
+```
+
+运行任务并传入本次筛选条件：
+
+```bash
+dist/docfill-darwin-arm64 run -c 工单日报.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+输出诊断日志：
+
+```bash
+dist/docfill-darwin-arm64 run -c task.json --verbose
+```
+
+重新生成本地测试素材：
+
+```bash
+/Users/zhaoyilun/.cache/codex-runtimes/codex-primary-runtime/dependencies/python/bin/python3 scripts/make-manual-test-materials.py
+```
+
+运行自动化冒烟测试：
 
 ```bash
 bash scripts/smoke.sh
 ```
 
-Smoke test 会：
+运行 Go 测试：
 
-- 跑全量 Go 测试。
-- 构建临时 `docfill` 二进制。
-- 复制 `examples/` 到临时目录。
-- 生成临时 Excel、Word、SQLite 示例数据。
-- 执行两个示例任务并检查输出文件。
+```bash
+go test -count=1 ./...
+```
 
-## Digital Employee Runbook
+## 筛选配置示例
 
-给数字员工或调度脚本使用时，看：
+Excel 工单表：
 
-- `docs/digital-employee-runbook.md`
+```text
+日期          工单编号   客户名称  系统来源   处理结果  下一步
+2026-06-18   GD-001    A 公司   PMS3.0   已派单   等待现场反馈
+2026-06-18   GD-002    B 公司   D5000    已完成   归档并同步日报
+2026-06-19   GD-001    C 公司   风控平台  待复核   补充截图
+```
 
-## Runnable Examples
-
-示例配置在 `examples/`：
-
-- `examples/excel-to-word/task.json`
-- `examples/sqlite-to-excel/task.json`
-
-示例里的二进制 Office/SQLite 文件由 `scripts/smoke-fixtures` 临时生成，不提交到仓库。
-
-## Excel Data Example
+`task.json`：
 
 ```json
 {
@@ -82,63 +154,88 @@ Smoke test 会：
   },
   "data": {
     "type": "excel",
-    "path": "./data.xlsx",
-    "sheet": "Sheet1",
+    "path": "./source_work_orders.xlsx",
+    "sheet": "工单",
     "header_row": 1
   },
+  "filter": {
+    "equals": {
+      "日期": "{{date}}",
+      "工单编号": "{{ticket_no}}"
+    },
+    "expect": "one"
+  },
   "template": {
-    "path": "./template.docx"
+    "path": "./work_order_template.docx"
   },
   "output": {
-    "path": "./out/report_{{date}}_{{name}}_{{_row}}.docx",
-    "overwrite": false
+    "path": "./out/work_order_{{日期}}_{{工单编号}}.docx",
+    "overwrite": true
   }
 }
 ```
 
-Excel 第一行是字段名，后续每一行生成一个文件。
+运行：
 
-## SQLite Data Example
-
-```json
-{
-  "vars": {
-    "date": "2026-06-18"
-  },
-  "data": {
-    "type": "sqlite",
-    "path": "./data.db",
-    "query": "select name, date, done from daily_report where date = ?",
-    "params": ["{{date}}"]
-  },
-  "template": {
-    "path": "./template.xlsx"
-  },
-  "output": {
-    "path": "./out/report_{{name}}_{{_row}}.xlsx"
-  }
-}
+```bash
+dist/docfill-darwin-arm64 run -c task.json --var date=2026-06-18 --var ticket_no=GD-002
 ```
 
-SQL 查询结果的列名就是模板占位符字段名。
+筛选含义：
 
-## P0 Rules
+```text
+找到 日期 = 2026-06-18，并且 工单编号 = GD-002 的那一行。
+```
 
-- 配置里的相对路径按 `task.json` 所在目录解析。
-- `vars` 可以提供默认变量，命令行 `--var key=value` 会覆盖同名默认变量。
-- 模板填充时，每行数据字段优先于 `vars`。
-- `_row` 是内置变量，表示当前数据行序号，从 `1` 开始。
-- SQLite `params` 对应 SQL 里的 `?` 参数。
-- Excel `header_row` 从 `1` 开始，默认是 `1`。
-- 默认不覆盖已有输出文件；只有 `output.overwrite: true` 才允许覆盖。
-- 输出文件路径中的字段值会清理 `/ \ : * ? " < > |`。
-- 模板里缺字段或渲染后仍有 `{{字段}}`，会直接失败。
-- Word 模板支持同一段落内被拆开的文本占位符，例如 `{{na` 和 `me}}` 位于相邻文本节点。
+`filter.expect` 支持：
 
-## MVP Limits
+- `one`: 必须只匹配一行，适合工单、单份日报、单个报表。
+- `many`: 允许匹配多行，适合按日期生成多个人或多个部门的报表。
 
-- 只支持 JSON 配置。
-- 只支持 `.xlsx` 数据源、SQLite 数据源。
-- 只支持 `.docx` 和 `.xlsx` 模板。
-- Word 占位符拆分只做最小兼容：支持同一段落内的文本节点拆分，不做复杂表格循环或样式级模板语法。
-- 不做 Web、权限、审批流、模板设计器或业务系统直连。
+未配置 `filter` 时，工具沿用旧行为：读取到的每一行都会生成一个输出文件。
+
+## 核心规则
+
+- 配置文件使用 JSON。
+- 数据源支持 `.xlsx` 和 SQLite `.db`。
+- 模板与输出支持 `.docx` 和 `.xlsx`。
+- 模板占位符格式为 `{{字段名}}`。
+- Excel 数据的表头行提供字段名，SQLite 查询结果的列名提供字段名。
+- 常用场景是通过 `filter.equals` 先筛选数据，再填充模板。
+- 多个筛选条件之间是“全部满足”关系。
+- 筛选只支持等值匹配，不支持模糊匹配、范围查询、OR 或复杂表达式。
+- 复杂情况建议业务人员整理 Excel，例如新增“填报批次”“业务唯一标识”等辅助列。
+- `vars` 提供默认变量，命令行 `--var key=value` 可以覆盖同名变量。
+- 数据行字段优先于 `vars` 和 `--var`。
+- `_row` 是内置变量，表示当前输出对应的结果行序号，从 `1` 开始。
+- 默认不覆盖已存在文件，只有 `output.overwrite: true` 才允许覆盖。
+- 成功结果写入 stdout，诊断日志和错误写入 stderr，便于数字员工或调度脚本判断执行结果。
+
+## 退出码
+
+- `0`: 执行成功。
+- `1`: 配置、数据、模板或输出处理失败。
+- `2`: 命令行参数错误。
+
+错误输出格式：
+
+```text
+docfill: [ERROR_CODE] message
+hint: repair suggestion
+context: key=value
+```
+
+常见筛选错误：
+
+- `DATA_FILTER_FIELD`: 筛选字段不在数据表中。
+- `DATA_FILTER_VALUE`: 筛选条件引用的变量没有提供。
+- `DATA_FILTER_NO_MATCH`: 没有找到符合条件的数据行。
+- `DATA_FILTER_MULTI_MATCH`: 期望匹配一行，但实际匹配多行。
+
+完整错误码、日志说明和集成建议见 [HOWTO.md](HOWTO.md)。
+
+## 文档索引
+
+- [HOWTO.md](HOWTO.md): 完整使用说明，包含最小实践、详细配置、输入输出示例、错误处理和数字员工调用建议。
+- [manual_tests/README.md](manual_tests/README.md): 本地真实测试素材说明。
+- [docs/digital-employee-runbook.md](docs/digital-employee-runbook.md): 面向数字员工或调度脚本的最小运行约定。
diff --git a/cmd/docfill/main.go b/cmd/docfill/main.go
index bf46060..7fff71b 100644
--- a/cmd/docfill/main.go
+++ b/cmd/docfill/main.go
@@ -1,10 +1,12 @@
 package main
 
 import (
+	"errors"
 	"flag"
 	"fmt"
 	"io"
 	"os"
+	"sort"
 	"strings"
 
 	"xlsdoc/internal/docfill"
@@ -18,11 +20,11 @@ func main() {
 
 func run(args []string, stdout, stderr io.Writer) int {
 	if len(args) == 0 {
-		fmt.Fprintln(stderr, "docfill: missing command: run")
+		writeCodedError(stderr, docfill.CodeCLIUsage, "missing command: run", "Use: docfill run -c task.json", nil)
 		return 2
 	}
 	if args[0] != "run" {
-		fmt.Fprintf(stderr, "docfill: unknown command: %s\n", args[0])
+		writeCodedError(stderr, docfill.CodeCLIUsage, "unknown command: "+args[0], "Use: docfill run -c task.json", nil)
 		return 2
 	}
 
@@ -43,38 +45,78 @@ func runFill(args []string, stderr io.Writer) (docfill.RunResult, int) {
 	fs.SetOutput(stderr)
 
 	var configPath string
+	var verbose bool
 	var vars varFlags = make(map[string]string)
 	fs.SetOutput(io.Discard)
 	fs.StringVar(&configPath, "config", "", "path to task json")
 	fs.StringVar(&configPath, "c", "", "path to task json")
+	fs.BoolVar(&verbose, "verbose", false, "print diagnostic logs to stderr")
 	fs.Var(vars, "var", "template variable as key=value; repeatable")
 
 	if err := fs.Parse(args); err != nil {
-		fmt.Fprintf(stderr, "docfill: %v\n", err)
+		writeCodedError(stderr, docfill.CodeCLIUsage, err.Error(), "Check command flags and use key=value for --var.", nil)
 		return docfill.RunResult{}, 2
 	}
 	if fs.NArg() > 0 {
-		fmt.Fprintf(stderr, "docfill: unexpected argument: %s\n", fs.Arg(0))
+		writeCodedError(stderr, docfill.CodeCLIUsage, "unexpected argument: "+fs.Arg(0), "Remove extra positional arguments.", nil)
 		return docfill.RunResult{}, 2
 	}
 	if vars == nil {
 		vars = make(map[string]string)
 	}
 	if configPath == "" {
-		fmt.Fprintln(stderr, "docfill: missing --config")
+		writeCodedError(stderr, docfill.CodeCLIUsage, "missing --config", "Use -c task.json or --config task.json.", nil)
 		fs.SetOutput(stderr)
 		fs.Usage()
 		return docfill.RunResult{}, 2
 	}
 
-	result, err := docfill.RunWithOptions(configPath, docfill.RunOptions{Vars: vars})
+	var logf func(format string, args ...any)
+	if verbose {
+		logf = func(format string, args ...any) {
+			fmt.Fprintf(stderr, "log: "+format+"\n", args...)
+		}
+	}
+
+	result, err := docfill.RunWithOptions(configPath, docfill.RunOptions{Vars: vars, Logf: logf})
 	if err != nil {
-		fmt.Fprintf(stderr, "docfill: %v\n", err)
+		writeExecutionError(stderr, err)
 		return docfill.RunResult{}, 1
 	}
 	return result, 0
 }
 
+func writeExecutionError(stderr io.Writer, err error) {
+	var appErr *docfill.AppError
+	if errors.As(err, &appErr) {
+		writeCodedError(stderr, appErr.Code, err.Error(), appErr.Hint, appErr.Context)
+		return
+	}
+	writeCodedError(stderr, docfill.CodeInternal, err.Error(), "Run with --verbose and inspect inputs if this repeats.", nil)
+}
+
+func writeCodedError(stderr io.Writer, code docfill.ErrorCode, message, hint string, context map[string]string) {
+	fmt.Fprintf(stderr, "docfill: [%s] %s\n", code, message)
+	if hint != "" {
+		fmt.Fprintf(stderr, "hint: %s\n", hint)
+	}
+	if len(context) == 0 {
+		return
+	}
+
+	keys := make([]string, 0, len(context))
+	for key := range context {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+
+	parts := make([]string, 0, len(keys))
+	for _, key := range keys {
+		parts = append(parts, key+"="+context[key])
+	}
+	fmt.Fprintf(stderr, "context: %s\n", strings.Join(parts, " "))
+}
+
 func (v varFlags) String() string {
 	if len(v) == 0 {
 		return ""
diff --git a/cmd/docfill/main_test.go b/cmd/docfill/main_test.go
index bc7ca9d..24de209 100644
--- a/cmd/docfill/main_test.go
+++ b/cmd/docfill/main_test.go
@@ -63,7 +63,7 @@ func TestRunCommandRejectsBadVarSyntax(t *testing.T) {
 	if exitCode != 2 {
 		t.Fatalf("exit code = %d, want 2", exitCode)
 	}
-	if !strings.Contains(stderr.String(), "docfill:") || !strings.Contains(stderr.String(), "invalid --var") {
+	if !strings.Contains(stderr.String(), "[CLI_USAGE]") || !strings.Contains(stderr.String(), "invalid --var") {
 		t.Fatalf("unexpected stderr: %q", stderr.String())
 	}
 }
@@ -84,6 +84,96 @@ func TestRunCommandRejectsUnexpectedArgs(t *testing.T) {
 	}
 }
 
+func TestRunCommandFormatsCodedExecutionErrors(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "data.xlsx")
+	templatePath := filepath.Join(dir, "template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeCLIWorkbook(t, dataPath, "Daily", [][]string{
+		{"name"},
+		{"张三"},
+	})
+	writeCLIWorkbook(t, templatePath, "日报", [][]string{
+		{"{{name}}", "{{missing}}"},
+	})
+	writeCLIJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "Daily",
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "日报_{{name}}.xlsx"),
+		},
+	})
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	exitCode := run([]string{"run", "-c", configPath}, &stdout, &stderr)
+
+	if exitCode != 1 {
+		t.Fatalf("exit code = %d, want 1", exitCode)
+	}
+	if stdout.Len() != 0 {
+		t.Fatalf("expected empty stdout, got %q", stdout.String())
+	}
+	got := stderr.String()
+	for _, want := range []string{"docfill: [TEMPLATE_FIELD_MISSING]", "hint:", "context:", "field=missing", "template="} {
+		if !strings.Contains(got, want) {
+			t.Fatalf("stderr missing %q: %q", want, got)
+		}
+	}
+}
+
+func TestRunCommandVerbosePrintsStepLogs(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "data.xlsx")
+	templatePath := filepath.Join(dir, "template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeCLIWorkbook(t, dataPath, "Daily", [][]string{
+		{"name", "done"},
+		{"张三", "完成日报"},
+	})
+	writeCLIWorkbook(t, templatePath, "日报", [][]string{
+		{"{{name}}", "{{done}}"},
+	})
+	writeCLIJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "Daily",
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "日报_{{name}}.xlsx"),
+		},
+	})
+
+	var stdout bytes.Buffer
+	var stderr bytes.Buffer
+	exitCode := run([]string{"run", "-c", configPath, "--verbose"}, &stdout, &stderr)
+
+	if exitCode != 0 {
+		t.Fatalf("exit code = %d, stderr=%q", exitCode, stderr.String())
+	}
+	if !strings.Contains(stdout.String(), "generated 1 file(s)") {
+		t.Fatalf("unexpected stdout: %q", stdout.String())
+	}
+	got := stderr.String()
+	for _, want := range []string{"log: load config", "log: read data", "log: render row=1"} {
+		if !strings.Contains(got, want) {
+			t.Fatalf("stderr missing %q: %q", want, got)
+		}
+	}
+}
+
 func writeCLIWorkbook(t *testing.T, path, sheet string, rows [][]string) {
 	t.Helper()
 
diff --git a/docs/digital-employee-runbook.md b/docs/digital-employee-runbook.md
index 70fa061..b15e1ab 100644
--- a/docs/digital-employee-runbook.md
+++ b/docs/digital-employee-runbook.md
@@ -8,11 +8,14 @@ Each run needs:
 
 - A `task.json` config file.
 - A data source: `.xlsx` or SQLite `.db`.
+- Optional business filters in `filter.equals`.
 - A template: `.docx` or `.xlsx`.
 - An output path pattern in `task.json`.
 
 Paths inside `task.json` are resolved from the config file directory.
 
+Use one `task.json` per business task. Use `--var` for values that change per run, such as date, ticket number, person, or batch.
+
 ## Command
 
 ```bash
@@ -22,10 +25,42 @@ docfill run -c /path/to/task.json --var date=2026-06-18
 Use repeated `--var key=value` arguments for runtime values:
 
 ```bash
-docfill run -c task.json --var date=2026-06-18 --var batch=早班
+docfill run -c task.json --var date=2026-06-18 --var ticket_no=GD-002
 ```
 
-CLI variables override `vars` in `task.json`. Row data overrides both.
+Priority:
+
+```text
+row data > CLI --var > task.json vars
+```
+
+CLI variables override `vars` in `task.json`. The selected row data overrides both when rendering templates and output paths.
+
+If `task.json` has a filter:
+
+```json
+"filter": {
+  "equals": {
+    "日期": "{{date}}",
+    "工单编号": "{{ticket_no}}"
+  },
+  "expect": "one"
+}
+```
+
+then the command:
+
+```bash
+docfill run -c task.json --var date=2026-06-18 --var ticket_no=GD-002
+```
+
+selects the row where `日期 = 2026-06-18` and `工单编号 = GD-002`.
+
+Use `--verbose` only when diagnosing a failed or suspicious run:
+
+```bash
+docfill run -c task.json --var date=2026-06-18 --var ticket_no=GD-002 --verbose
+```
 
 ## Success
 
@@ -48,9 +83,91 @@ Exit code `2` means the command line was invalid.
 Errors are written to stderr with the `docfill:` prefix:
 
 ```text
-docfill: render row 1: missing template field "done"
+docfill: [TEMPLATE_FIELD_MISSING] render row 1: missing template field "done"
+hint: Add column "done" to the data source or pass --var done=...
+context: field=done template=/path/template.docx
 ```
 
+Verbose logs are also written to stderr:
+
+```text
+log: load config path=/path/task.json
+log: read data type=excel path=/path/data.xlsx
+log: read data type=excel path=/path/data.xlsx rows=3
+log: filter rows=3 matched=1 expect=one
+log: render row=1 output=/path/out/work_order_2026-06-18_GD-002.docx
+```
+
+## Error Codes
+
+`CLI_USAGE`
+
+The command line is invalid.
+
+`CONFIG_READ`
+
+The config file cannot be read.
+
+`CONFIG_PARSE`
+
+The config file is not valid JSON.
+
+`CONFIG_VALIDATE`
+
+The config is missing required fields or uses unsupported values.
+
+`DATA_EXCEL`
+
+Excel data cannot be opened or read.
+
+`DATA_SQLITE`
+
+SQLite data cannot be opened, queried, or parameterized.
+
+`DATA_EMPTY`
+
+The data source returned no rows.
+
+`DATA_FILTER_FIELD`
+
+The filter references a field that is not in the data.
+
+`DATA_FILTER_VALUE`
+
+The filter references a variable that was not provided.
+
+`DATA_FILTER_NO_MATCH`
+
+The filter matched no rows.
+
+`DATA_FILTER_MULTI_MATCH`
+
+The filter expected one row but matched multiple rows.
+
+`TEMPLATE_OPEN`
+
+The template file cannot be opened or inspected.
+
+`TEMPLATE_FIELD_MISSING`
+
+The template references a field that is not available.
+
+`TEMPLATE_UNRESOLVED`
+
+The rendered file would still contain a `{{field}}` placeholder.
+
+`OUTPUT_EXISTS`
+
+The output file exists and overwrite is not enabled.
+
+`OUTPUT_WRITE`
+
+The output directory or file cannot be written.
+
+`INTERNAL`
+
+An unexpected unclassified error occurred.
+
 ## Common Errors
 
 `missing template field`
@@ -65,6 +182,18 @@ The output file exists and `output.overwrite` is not `true`.
 
 A SQLite query parameter such as `{{date}}` was not provided.
 
+`filter matched no rows`
+
+Check `filter.equals`, `--var` values, or the source table contents.
+
+`filter matched N rows but expected one`
+
+Add another filter condition or clean duplicate rows in the source table.
+
+`filter field is not in data`
+
+Check Excel headers or SQLite query column names.
+
 `unresolved placeholder`
 
 The rendered document still contains `{{field}}`. Check the source data value and the template.
@@ -87,4 +216,4 @@ dist/docfill-darwin-arm64
 bash scripts/smoke.sh
 ```
 
-Smoke test builds a temporary binary, generates temporary example files, runs both example configs, and checks generated outputs.
+Smoke test builds a temporary binary, generates temporary example files, runs example configs including the filter case, and checks generated outputs.
diff --git a/docs/evaluations/2026-06-19-full-evaluation.md b/docs/evaluations/2026-06-19-full-evaluation.md
new file mode 100644
index 0000000..1de2b43
--- /dev/null
+++ b/docs/evaluations/2026-06-19-full-evaluation.md
@@ -0,0 +1,186 @@
+# docfill 全方位评测报告
+
+评测日期：2026-06-19
+
+评测对象：`docfill` 当前工作区版本，包含错误码、日志、业务筛选 `filter.equals`、手工测试素材和用户文档更新。
+
+## 结论
+
+本轮评测结论：通过当前 MVP 质量门槛。
+
+适用范围：
+
+- 本地 macOS arm64 构建。
+- Excel / SQLite 数据源读取。
+- Word / Excel 模板填充。
+- `filter.equals` 多条件等值筛选。
+- `filter.expect` 的 `one` / `many` 行数约束。
+- CLI 成功输出、错误码、提示和上下文。
+- 用户文档、示例配置和手工测试素材一致性。
+
+未宣称覆盖：
+
+- Windows / Linux 二进制构建。
+- 大体量数据性能测试。
+- Excel 公式重算、透视表刷新。
+- 复杂 Word 模板语法。
+- 业务系统直连、权限、审批或调度系统。
+
+## 工作区状态
+
+当前工作区包含未提交变更，评测没有执行提交、回滚或清理。
+
+基线提交：
+
+```text
+ca444bd feat: add docfill mvp cli
+```
+
+评测时存在的主要变更类型：
+
+- CLI 错误码和日志。
+- `filter.equals` / `filter.expect` 筛选能力。
+- 新增 `HOWTO.md`。
+- 新增 `manual_tests/` 真实测试素材。
+- 新增 `examples/excel-filter-to-word/` smoke 示例。
+- 更新 README、runbook 和示例文档。
+
+## 自动化评测
+
+| 项目 | 命令 | 结果 |
+| --- | --- | --- |
+| Git diff 格式检查 | `git diff --check` | 通过，无输出 |
+| Go 单元测试 | `go test -count=1 ./...` | 通过 |
+| Go race 测试 | `go test -race -count=1 ./...` | 通过 |
+| Go 静态检查 | `go vet ./...` | 通过，无输出 |
+| 本机构建 | `bash scripts/build.sh` | 通过，输出 `dist/docfill-darwin-arm64` |
+| 冒烟测试 | `bash scripts/smoke.sh` | 通过，包含无筛选、筛选、SQLite 示例 |
+
+`bash scripts/smoke.sh` 覆盖：
+
+- Excel 到 Word：生成 2 个日报 Word 文件。
+- Excel 筛选到 Word：按 `date + ticket_no` 生成 1 个工单 Word 文件。
+- SQLite 到 Excel：生成 2 个部门汇总 Excel 文件。
+
+## 真实素材成功路径
+
+| 案例 | 命令 | 结果 |
+| --- | --- | --- |
+| Excel 筛选到 Word | `dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002 --verbose` | 生成 1 个 Word 文件 |
+| Excel 到 Word | `dist/docfill-darwin-arm64 run -c manual_tests/01_excel_to_word/task.json --var batch=早班 --verbose` | 生成 2 个 Word 文件 |
+| SQLite 到 Excel | `dist/docfill-darwin-arm64 run -c manual_tests/02_sqlite_to_excel/task.json --var date=2026-06-18 --verbose` | 生成 2 个 Excel 文件 |
+| Excel 到 Excel | `dist/docfill-darwin-arm64 run -c manual_tests/03_excel_to_excel/task.json --var date=2026-06-18 --verbose` | 生成 2 个 Excel 文件 |
+
+关键输出：
+
+```text
+manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+manual_tests/01_excel_to_word/out/daily_2026-06-18_张三_测试_1.docx
+manual_tests/01_excel_to_word/out/daily_2026-06-18_李四_2.docx
+manual_tests/02_sqlite_to_excel/out/summary_客服部_1.xlsx
+manual_tests/02_sqlite_to_excel/out/summary_营销部_2.xlsx
+manual_tests/03_excel_to_excel/out/audit_2026-06-18_王五_1.xlsx
+manual_tests/03_excel_to_excel/out/audit_2026-06-18_赵六_2.xlsx
+```
+
+## 真实素材内容核验
+
+使用 bundled Python 读取生成的 Word / Excel 文件并断言关键字段。
+
+核验结果：
+
+- 筛选工单 Word 包含 `GD-002`、`B 公司`、`D5000`、`已完成`。
+- 筛选工单 Word 不包含非目标行 `GD-001`、`A 公司`。
+- 日报 Word 包含 `张三/测试`、`早班`、`营销部`。
+- SQLite 汇总 Excel 中 `summary_营销部_2.xlsx` 的 B2:B6 为 `营销部 / 2026-06-18 / 张三 / 128 / 3`。
+- Excel 质检输出中 `audit_2026-06-18_赵六_2.xlsx` 的 B2:B6 为 `2026-06-18 / 赵六 / 系统截图一致性 / 87 / 缺少 1 张截图，需要补充`。
+- SQLite 源表包含 3 行测试数据。
+
+## DOCX 渲染核验
+
+使用 `documents` skill 的 `render_docx.py` 渲染两个 Word 输出：
+
+```text
+manual_tests/_evaluation_rendered/filter/page-1.png
+manual_tests/_evaluation_rendered/daily/page-1.png
+```
+
+视觉核验结果：
+
+- 筛选工单页可读，字段位置正常，关键值为 `2026-06-18 / GD-002 / B 公司 / D5000 / 已完成 / 归档并同步日报`。
+- 日报页可读，字段位置正常，关键值为 `张三/测试 / 营销部 / 早班`。
+- 未观察到明显空白页、文本重叠、缺字或表格错位。
+
+## CLI 失败路径
+
+| 场景 | 命令或配置 | 预期错误码 | 结果 |
+| --- | --- | --- | --- |
+| 筛选无匹配 | `ticket_no=GD-999` | `DATA_FILTER_NO_MATCH` | 通过 |
+| 筛选变量缺失 | 不传 `ticket_no` | `DATA_FILTER_VALUE` | 通过 |
+| 筛选匹配多行 | 临时配置只按 `日期` 筛选且 `expect=one` | `DATA_FILTER_MULTI_MATCH` | 通过 |
+| 非法 `filter.expect` | 临时配置 `expect=single` | `CONFIG_VALIDATE` | 通过 |
+| 配置不存在 | `-c manual_tests/not-exist.json` | `CONFIG_READ` | 通过 |
+| CLI 参数错误 | `--var date` | `CLI_USAGE`，退出码 2 | 通过 |
+
+典型错误输出：
+
+```text
+docfill: [DATA_FILTER_NO_MATCH] filter matched no rows
+hint: Check filter.equals, --var values, or source table contents.
+context: expect=one filter=工单编号=GD-999,日期=2026-06-18
+```
+
+```text
+docfill: [DATA_FILTER_MULTI_MATCH] filter matched 2 rows but expected one
+hint: Add another filter condition or clean duplicate rows in the source table.
+context: filter=日期=2026-06-18 matches=2
+```
+
+## 结构与文档一致性
+
+| 检查 | 结果 |
+| --- | --- |
+| 所有 `task.json` 可被 JSON 解析 | 通过 |
+| `manual_tests/**/out/*.{docx,xlsx}` 内部 XML 无残留 `{{` / `}}` 占位符 | 通过 |
+| README / HOWTO / runbook 均说明 `filter.equals`、`filter.expect`、`--var` 优先级 | 通过 |
+| 用户文档未出现不合适措辞 `小白`、`大白话` | 通过 |
+| 用户文档无 `TODO` / `TBD` / `FIXME` | 通过 |
+
+## 能力覆盖评估
+
+已验证：
+
+- Excel 数据源读取。
+- SQLite 数据源读取。
+- Word 模板填充。
+- Excel 模板填充。
+- Word 同段落拆分占位符替换。
+- 输出文件名非法字符清理。
+- `vars` 默认变量。
+- CLI `--var` 覆盖 `vars`。
+- 数据行字段优先于 `--var` / `vars`。
+- 多条件业务筛选。
+- `expect=one` 和 `expect=many`。
+- 筛选错误、配置错误、CLI 错误。
+- stdout 输出生成文件清单。
+- stderr 输出日志、错误码、hint、context。
+
+未覆盖或仅部分覆盖：
+
+- 大文件性能。
+- 多 sheet 大范围模板扫描性能。
+- Windows / Linux 路径与构建。
+- Excel 公式、图表、透视表类模板。
+- Word 跨段落拆分占位符。
+- 真实业务系统导出的复杂脏数据。
+
+## 风险与建议
+
+P0 无阻塞问题。
+
+建议进入下一轮前只做小范围补强：
+
+1. 增加一条大约 500 到 1000 行 Excel 的筛选性能测试，确认小工具在常见业务表规模下响应稳定。
+2. 增加文档中的“字段命名建议”，例如日期列、工单编号列、辅助列的命名方式。
+3. 保持筛选语法只支持等值 AND，不增加模糊匹配、范围查询或表达式。
+4. 发布前在目标机器上补一次 `bash scripts/build.sh && bash scripts/smoke.sh`。
diff --git a/docs/superpowers/plans/2026-06-18-docfill-p0.7-errors-logs.md b/docs/superpowers/plans/2026-06-18-docfill-p0.7-errors-logs.md
new file mode 100644
index 0000000..21780b6
--- /dev/null
+++ b/docs/superpowers/plans/2026-06-18-docfill-p0.7-errors-logs.md
@@ -0,0 +1,52 @@
+# Docfill P0.7 Errors And Logs Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Make failures easier for digital employees and humans to classify without turning `docfill` into a logging platform.
+
+**Architecture:** Add a small `AppError` type in the internal package with code, hint, and context. Keep CLI formatting in `cmd/docfill`; add only one runtime logging switch, `--verbose`, which writes step logs to stderr.
+
+**Tech Stack:** Go standard errors, existing CLI flag parsing, existing tests.
+
+---
+
+### Task 1: Error Contract
+
+**Files:**
+- Create: `internal/docfill/errors.go`
+- Modify: `internal/docfill/p0_test.go`
+- Modify: `cmd/docfill/main_test.go`
+
+- [ ] Add tests for coded missing-field and output-exists failures.
+- [ ] Add tests for CLI stderr formatting with `[CODE]`, `hint:`, and `context:`.
+
+### Task 2: Error Classification
+
+**Files:**
+- Modify: `internal/docfill/docfill.go`
+- Create: `internal/docfill/errors.go`
+
+- [ ] Classify config read, config parse, config validate, data Excel, data SQLite, template, output, and unresolved-placeholder failures where they already occur.
+- [ ] Keep existing return signatures.
+
+### Task 3: Verbose Logs
+
+**Files:**
+- Modify: `internal/docfill/docfill.go`
+- Modify: `cmd/docfill/main.go`
+- Modify: `cmd/docfill/main_test.go`
+
+- [ ] Add `RunOptions.Logf`.
+- [ ] Add `--verbose` to `docfill run`.
+- [ ] Log only config load, data read summary, and per-row output path.
+
+### Task 4: Docs And Verification
+
+**Files:**
+- Modify: `README.md`
+- Modify: `docs/digital-employee-runbook.md`
+
+- [ ] Document error shape and `--verbose`.
+- [ ] Run `go test -count=1 ./...`.
+- [ ] Run `bash scripts/smoke.sh`.
+- [ ] Run `bash scripts/build.sh`.
diff --git a/examples/README.md b/examples/README.md
index bb68185..4fa66ac 100644
--- a/examples/README.md
+++ b/examples/README.md
@@ -1,16 +1,32 @@
-# docfill examples
+# docfill 示例配置
 
-These examples are intentionally small and script-friendly.
+本目录保存可自动化运行的最小示例配置。示例中的 Office 文件和 SQLite 文件由 `scripts/smoke.sh` 在临时目录中生成，因此仓库只提交配置文件，不提交临时生成的数据文件。
 
-Run all examples through the smoke script:
+## 运行全部示例
 
 ```bash
 bash scripts/smoke.sh
 ```
 
-The smoke script copies this directory to a temp folder, generates the Office and SQLite fixture files, builds `docfill`, and runs the example tasks.
+输出示例：
 
-## Examples
+```text
+generated 2 file(s)
+- /tmp/.../examples/excel-to-word/out/daily_2026-06-18_张三_1.docx
+- /tmp/.../examples/excel-to-word/out/daily_2026-06-18_李四_2.docx
+generated 2 file(s)
+- /tmp/.../examples/sqlite-to-excel/out/summary_客服部_1.xlsx
+- /tmp/.../examples/sqlite-to-excel/out/summary_营销部_2.xlsx
+smoke ok
+```
 
-- `excel-to-word`: Excel daily rows to Word daily reports.
-- `sqlite-to-excel`: SQLite summary rows to Excel department reports.
+## 示例列表
+
+- `excel-to-word`: 从 Excel 日报数据生成 Word 日报。
+- `excel-filter-to-word`: 从 Excel 工单数据中按日期和工单编号筛选一条记录，再生成 Word 工单。
+- `sqlite-to-excel`: 从 SQLite 汇总数据生成 Excel 部门汇总表。
+
+## 与 manual_tests 的区别
+
+- `examples/` 用于自动化冒烟测试，文件尽量少，数据文件运行时临时生成。
+- `manual_tests/` 用于人工验证，包含可以直接打开的 Word、Excel 和 SQLite 测试素材。
diff --git a/examples/excel-filter-to-word/task.json b/examples/excel-filter-to-word/task.json
new file mode 100644
index 0000000..66d1f1a
--- /dev/null
+++ b/examples/excel-filter-to-word/task.json
@@ -0,0 +1,25 @@
+{
+  "vars": {
+    "date": "2026-06-18"
+  },
+  "data": {
+    "type": "excel",
+    "path": "./data.xlsx",
+    "sheet": "Tickets",
+    "header_row": 1
+  },
+  "filter": {
+    "equals": {
+      "date": "{{date}}",
+      "ticket_no": "{{ticket_no}}"
+    },
+    "expect": "one"
+  },
+  "template": {
+    "path": "./template.docx"
+  },
+  "output": {
+    "path": "./out/ticket_{{date}}_{{ticket_no}}.docx",
+    "overwrite": true
+  }
+}
diff --git a/internal/docfill/docfill.go b/internal/docfill/docfill.go
index 0098f16..60a6e24 100644
--- a/internal/docfill/docfill.go
+++ b/internal/docfill/docfill.go
@@ -4,12 +4,12 @@ import (
 	"archive/zip"
 	"database/sql"
 	"encoding/json"
-	"errors"
 	"fmt"
 	"io"
 	"os"
 	"path/filepath"
 	"regexp"
+	"sort"
 	"strconv"
 	"strings"
 
@@ -21,6 +21,7 @@ import (
 type config struct {
 	Vars     map[string]string `json:"vars"`
 	Data     dataConfig        `json:"data"`
+	Filter   filterConfig      `json:"filter"`
 	Template templateConfig    `json:"template"`
 	Output   outputConfig      `json:"output"`
 }
@@ -34,6 +35,11 @@ type dataConfig struct {
 	HeaderRow int      `json:"header_row"`
 }
 
+type filterConfig struct {
+	Equals map[string]string `json:"equals"`
+	Expect string            `json:"expect"`
+}
+
 type templateConfig struct {
 	Path string `json:"path"`
 }
@@ -45,6 +51,7 @@ type outputConfig struct {
 
 type RunOptions struct {
 	Vars map[string]string
+	Logf func(format string, args ...any)
 }
 
 type RunResult struct {
@@ -68,6 +75,7 @@ func Run(configPath string) error {
 
 // RunWithOptions executes one configured fill task and returns generated files.
 func RunWithOptions(configPath string, opts RunOptions) (RunResult, error) {
+	logf(opts, "load config path=%s", configPath)
 	cfg, err := loadConfig(configPath)
 	if err != nil {
 		return RunResult{}, err
@@ -77,12 +85,24 @@ func RunWithOptions(configPath string, opts RunOptions) (RunResult, error) {
 	}
 
 	vars := mergeVars(cfg.Vars, opts.Vars)
+	logf(opts, "read data type=%s path=%s", cfg.Data.Type, cfg.Data.Path)
 	rows, err := readRowsWithVars(cfg.Data, vars)
 	if err != nil {
 		return RunResult{}, err
 	}
 	if len(rows) == 0 {
-		return RunResult{}, errors.New("data returned no rows")
+		return RunResult{}, appError(CodeDataEmpty, "data returned no rows").
+			withHint("Check the data source, sheet/header row, SQL where clause, or --var values.").
+			withContext("data", cfg.Data.Path)
+	}
+	logf(opts, "read data type=%s path=%s rows=%d", cfg.Data.Type, cfg.Data.Path, len(rows))
+	sourceRowCount := len(rows)
+	rows, err = filterRows(cfg.Filter, rows, vars)
+	if err != nil {
+		return RunResult{}, err
+	}
+	if hasFilter(cfg.Filter) {
+		logf(opts, "filter rows=%d matched=%d expect=%s", sourceRowCount, len(rows), filterExpect(cfg.Filter))
 	}
 
 	result := RunResult{Files: make([]string, 0, len(rows))}
@@ -94,11 +114,16 @@ func RunWithOptions(configPath string, opts RunOptions) (RunResult, error) {
 
 		outputPath := replacePathPlaceholders(cfg.Output.Path, context)
 		if hasPlaceholder(outputPath) {
-			return RunResult{}, fmt.Errorf("render row %d: unresolved placeholder in output path %q", i+1, outputPath)
+			return RunResult{}, fmt.Errorf("render row %d: %w", i+1, appError(CodeTemplateUnresolved, "unresolved placeholder in output path").
+				withHint("Check output.path placeholders and available data fields or --var values.").
+				withContext("output", outputPath))
 		}
 		if !cfg.Output.Overwrite && fileExists(outputPath) {
-			return RunResult{}, fmt.Errorf("render row %d: output already exists: %s", i+1, outputPath)
+			return RunResult{}, fmt.Errorf("render row %d: %w", i+1, appError(CodeOutputExists, "output already exists").
+				withHint("Change output.path or set output.overwrite to true.").
+				withContext("output", outputPath))
 		}
+		logf(opts, "render row=%d output=%s", i+1, outputPath)
 		if err := renderTemplate(cfg.Template.Path, outputPath, context); err != nil {
 			return RunResult{}, fmt.Errorf("render row %d: %w", i+1, err)
 		}
@@ -109,19 +134,32 @@ func RunWithOptions(configPath string, opts RunOptions) (RunResult, error) {
 	return result, nil
 }
 
+func logf(opts RunOptions, format string, args ...any) {
+	if opts.Logf != nil {
+		opts.Logf(format, args...)
+	}
+}
+
 func loadConfig(path string) (config, error) {
 	if path == "" {
-		return config{}, errors.New("config path is required")
+		return config{}, appError(CodeConfigValidate, "config path is required").
+			withHint("Pass -c task.json or --config task.json.")
 	}
 
 	data, err := os.ReadFile(path)
 	if err != nil {
-		return config{}, fmt.Errorf("read config: %w", err)
+		return config{}, appError(CodeConfigRead, "read config").
+			withHint("Check that the config file exists and is readable.").
+			withContext("config", path).
+			withCause(err)
 	}
 
 	var cfg config
 	if err := json.Unmarshal(data, &cfg); err != nil {
-		return config{}, fmt.Errorf("parse config json: %w", err)
+		return config{}, appError(CodeConfigParse, "parse config json").
+			withHint("Fix task.json syntax and make sure it is valid JSON.").
+			withContext("config", path).
+			withCause(err)
 	}
 	resolveConfigPaths(&cfg, filepath.Dir(absPath(path)))
 	return cfg, nil
@@ -129,19 +167,27 @@ func loadConfig(path string) (config, error) {
 
 func validateConfig(cfg config) error {
 	if cfg.Data.Type == "" {
-		return errors.New("data.type is required")
+		return appError(CodeConfigValidate, "data.type is required").
+			withHint("Set data.type to excel or sqlite.")
 	}
 	if cfg.Data.Path == "" {
-		return errors.New("data.path is required")
+		return appError(CodeConfigValidate, "data.path is required").
+			withHint("Set data.path to an .xlsx or .db file.")
 	}
 	if cfg.Template.Path == "" {
-		return errors.New("template.path is required")
+		return appError(CodeConfigValidate, "template.path is required").
+			withHint("Set template.path to a .docx or .xlsx file.")
 	}
 	if cfg.Output.Path == "" {
-		return errors.New("output.path is required")
+		return appError(CodeConfigValidate, "output.path is required").
+			withHint("Set output.path to the generated file pattern.")
 	}
 	if cfg.Data.HeaderRow < 0 {
-		return errors.New("data.header_row must not be negative")
+		return appError(CodeConfigValidate, "data.header_row must not be negative").
+			withHint("Use a 1-based header_row value, or omit it for 1.")
+	}
+	if err := validateFilterConfig(cfg.Filter); err != nil {
+		return err
 	}
 	return nil
 }
@@ -157,14 +203,19 @@ func readRowsWithVars(cfg dataConfig, vars map[string]string) ([]map[string]stri
 	case "sqlite":
 		return readSQLiteRows(cfg, vars)
 	default:
-		return nil, fmt.Errorf("unsupported data.type %q", cfg.Type)
+		return nil, appError(CodeConfigValidate, fmt.Sprintf("unsupported data.type %q", cfg.Type)).
+			withHint("Set data.type to excel or sqlite.").
+			withContext("data.type", cfg.Type)
 	}
 }
 
 func readExcelRows(cfg dataConfig) ([]map[string]string, error) {
 	f, err := excelize.OpenFile(cfg.Path)
 	if err != nil {
-		return nil, fmt.Errorf("open excel data: %w", err)
+		return nil, appError(CodeDataExcel, "open excel data").
+			withHint("Check that data.path points to a readable .xlsx file.").
+			withContext("data", cfg.Path).
+			withCause(err)
 	}
 	defer f.Close()
 
@@ -173,15 +224,24 @@ func readExcelRows(cfg dataConfig) ([]map[string]string, error) {
 		sheet = f.GetSheetName(0)
 	}
 	if sheet == "" {
-		return nil, errors.New("excel data has no sheets")
+		return nil, appError(CodeDataExcel, "excel data has no sheets").
+			withHint("Check the Excel data file.").
+			withContext("data", cfg.Path)
 	}
 
 	rawRows, err := f.GetRows(sheet)
 	if err != nil {
-		return nil, fmt.Errorf("read excel sheet %q: %w", sheet, err)
+		return nil, appError(CodeDataExcel, fmt.Sprintf("read excel sheet %q", sheet)).
+			withHint("Check data.sheet and the workbook contents.").
+			withContext("data", cfg.Path).
+			withContext("sheet", sheet).
+			withCause(err)
 	}
 	if len(rawRows) == 0 {
-		return nil, errors.New("excel data has no header row")
+		return nil, appError(CodeDataExcel, "excel data has no header row").
+			withHint("Add a header row or set data.header_row to the correct row.").
+			withContext("data", cfg.Path).
+			withContext("sheet", sheet)
 	}
 
 	headerRow := cfg.HeaderRow
@@ -189,7 +249,10 @@ func readExcelRows(cfg dataConfig) ([]map[string]string, error) {
 		headerRow = 1
 	}
 	if headerRow > len(rawRows) {
-		return nil, fmt.Errorf("excel header_row %d is outside sheet %q", headerRow, sheet)
+		return nil, appError(CodeDataExcel, fmt.Sprintf("excel header_row %d is outside sheet %q", headerRow, sheet)).
+			withHint("Use a header_row value within the sheet row count.").
+			withContext("data", cfg.Path).
+			withContext("sheet", sheet)
 	}
 
 	headers := make([]string, len(rawRows[headerRow-1]))
@@ -221,12 +284,16 @@ func readExcelRows(cfg dataConfig) ([]map[string]string, error) {
 
 func readSQLiteRows(cfg dataConfig, vars map[string]string) ([]map[string]string, error) {
 	if cfg.Query == "" {
-		return nil, errors.New("data.query is required for sqlite")
+		return nil, appError(CodeConfigValidate, "data.query is required for sqlite").
+			withHint("Set data.query to the SQL select statement.")
 	}
 
 	db, err := sql.Open("sqlite", cfg.Path)
 	if err != nil {
-		return nil, fmt.Errorf("open sqlite data: %w", err)
+		return nil, appError(CodeDataSQLite, "open sqlite data").
+			withHint("Check that data.path points to a readable SQLite database.").
+			withContext("data", cfg.Path).
+			withCause(err)
 	}
 	defer db.Close()
 
@@ -237,13 +304,18 @@ func readSQLiteRows(cfg dataConfig, vars map[string]string) ([]map[string]string
 
 	result, err := db.Query(cfg.Query, args...)
 	if err != nil {
-		return nil, fmt.Errorf("query sqlite data: %w", err)
+		return nil, appError(CodeDataSQLite, "query sqlite data").
+			withHint("Check data.query, data.params, and --var values.").
+			withContext("data", cfg.Path).
+			withCause(err)
 	}
 	defer result.Close()
 
 	columns, err := result.Columns()
 	if err != nil {
-		return nil, fmt.Errorf("read sqlite columns: %w", err)
+		return nil, appError(CodeDataSQLite, "read sqlite columns").
+			withContext("data", cfg.Path).
+			withCause(err)
 	}
 
 	var rows []map[string]string
@@ -254,7 +326,9 @@ func readSQLiteRows(cfg dataConfig, vars map[string]string) ([]map[string]string
 			scanTargets[i] = &values[i]
 		}
 		if err := result.Scan(scanTargets...); err != nil {
-			return nil, fmt.Errorf("scan sqlite row: %w", err)
+			return nil, appError(CodeDataSQLite, "scan sqlite row").
+				withContext("data", cfg.Path).
+				withCause(err)
 		}
 
 		row := make(map[string]string)
@@ -264,18 +338,154 @@ func readSQLiteRows(cfg dataConfig, vars map[string]string) ([]map[string]string
 		rows = append(rows, row)
 	}
 	if err := result.Err(); err != nil {
-		return nil, fmt.Errorf("iterate sqlite rows: %w", err)
+		return nil, appError(CodeDataSQLite, "iterate sqlite rows").
+			withContext("data", cfg.Path).
+			withCause(err)
 	}
 
 	return rows, nil
 }
 
+func validateFilterConfig(cfg filterConfig) error {
+	expect := strings.ToLower(strings.TrimSpace(cfg.Expect))
+	if expect != "" && expect != "one" && expect != "many" {
+		return appError(CodeConfigValidate, fmt.Sprintf("unsupported filter.expect %q", cfg.Expect)).
+			withHint("Set filter.expect to one or many.").
+			withContext("filter.expect", cfg.Expect)
+	}
+	if expect != "" && len(cfg.Equals) == 0 {
+		return appError(CodeConfigValidate, "filter.equals is required when filter.expect is set").
+			withHint("Add at least one filter.equals condition or remove filter.expect.")
+	}
+	for field := range cfg.Equals {
+		if strings.TrimSpace(field) == "" {
+			return appError(CodeConfigValidate, "filter.equals field name must not be empty").
+				withHint("Use a data column name as each filter.equals key.")
+		}
+	}
+	return nil
+}
+
+func filterRows(cfg filterConfig, rows []map[string]string, vars map[string]string) ([]map[string]string, error) {
+	if !hasFilter(cfg) {
+		return rows, nil
+	}
+
+	conditions, err := filterConditions(cfg, vars)
+	if err != nil {
+		return nil, err
+	}
+	if len(rows) > 0 {
+		if err := validateFilterFields(rows[0], conditions); err != nil {
+			return nil, err
+		}
+	}
+
+	matches := make([]map[string]string, 0, len(rows))
+	for _, row := range rows {
+		if rowMatchesFilter(row, conditions) {
+			matches = append(matches, row)
+		}
+	}
+
+	expect := filterExpect(cfg)
+	if len(matches) == 0 {
+		return nil, appError(CodeDataFilterNoMatch, "filter matched no rows").
+			withHint("Check filter.equals, --var values, or source table contents.").
+			withContext("filter", filterSummary(conditions)).
+			withContext("expect", expect)
+	}
+	if expect == "one" && len(matches) != 1 {
+		return nil, appError(CodeDataFilterMultiMatch, fmt.Sprintf("filter matched %d rows but expected one", len(matches))).
+			withHint("Add another filter condition or clean duplicate rows in the source table.").
+			withContext("filter", filterSummary(conditions)).
+			withContext("matches", strconv.Itoa(len(matches)))
+	}
+
+	return matches, nil
+}
+
+func hasFilter(cfg filterConfig) bool {
+	return len(cfg.Equals) > 0
+}
+
+func filterExpect(cfg filterConfig) string {
+	expect := strings.ToLower(strings.TrimSpace(cfg.Expect))
+	if expect == "" {
+		return "one"
+	}
+	return expect
+}
+
+func filterConditions(cfg filterConfig, vars map[string]string) (map[string]string, error) {
+	conditions := make(map[string]string, len(cfg.Equals))
+	for field, pattern := range cfg.Equals {
+		field = strings.TrimSpace(field)
+		context := map[string]string{"filter": field}
+		if err := ensurePlaceholdersAvailableWithCode(
+			fmt.Sprintf("filter.equals[%s]", field),
+			pattern,
+			vars,
+			CodeDataFilterValue,
+			"Provide the missing filter value with vars or --var.",
+			context,
+		); err != nil {
+			return nil, err
+		}
+		value := replacePlaceholders(pattern, vars)
+		if hasPlaceholder(value) {
+			return nil, appError(CodeDataFilterValue, fmt.Sprintf("unresolved placeholder in filter.equals[%s]", field)).
+				withHint("Provide the missing filter value with vars or --var.").
+				withContext("filter", field)
+		}
+		conditions[field] = strings.TrimSpace(value)
+	}
+	return conditions, nil
+}
+
+func validateFilterFields(row map[string]string, conditions map[string]string) error {
+	for field := range conditions {
+		if _, ok := row[field]; !ok {
+			return appError(CodeDataFilterField, fmt.Sprintf("filter field %q is not in data", field)).
+				withHint("Check the data header/query columns or adjust filter.equals.").
+				withContext("field", field)
+		}
+	}
+	return nil
+}
+
+func rowMatchesFilter(row map[string]string, conditions map[string]string) bool {
+	for field, want := range conditions {
+		if strings.TrimSpace(row[field]) != want {
+			return false
+		}
+	}
+	return true
+}
+
+func filterSummary(conditions map[string]string) string {
+	keys := make([]string, 0, len(conditions))
+	for key := range conditions {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+
+	parts := make([]string, 0, len(keys))
+	for _, key := range keys {
+		parts = append(parts, key+"="+conditions[key])
+	}
+	return strings.Join(parts, ",")
+}
+
 func renderTemplate(templatePath, outputPath string, row map[string]string) error {
 	if err := validateTemplateFields(templatePath, row); err != nil {
 		return err
 	}
 	if err := os.MkdirAll(filepath.Dir(outputPath), 0o755); err != nil {
-		return fmt.Errorf("create output directory: %w", err)
+		return appError(CodeOutputWrite, "create output directory").
+			withHint("Check output.path directory permissions.").
+			withContext("output", outputPath).
+			withCause(err)
 	}
 
 	switch strings.ToLower(filepath.Ext(templatePath)) {
@@ -284,21 +494,29 @@ func renderTemplate(templatePath, outputPath string, row map[string]string) erro
 	case ".docx":
 		return renderDOCX(templatePath, outputPath, row)
 	default:
-		return fmt.Errorf("unsupported template extension %q", filepath.Ext(templatePath))
+		return appError(CodeConfigValidate, fmt.Sprintf("unsupported template extension %q", filepath.Ext(templatePath))).
+			withHint("Use a .docx or .xlsx template.").
+			withContext("template", templatePath)
 	}
 }
 
 func renderXLSX(templatePath, outputPath string, row map[string]string) error {
 	f, err := excelize.OpenFile(templatePath)
 	if err != nil {
-		return fmt.Errorf("open xlsx template: %w", err)
+		return appError(CodeTemplateOpen, "open xlsx template").
+			withHint("Check that template.path points to a readable .xlsx file.").
+			withContext("template", templatePath).
+			withCause(err)
 	}
 	defer f.Close()
 
 	for _, sheet := range f.GetSheetList() {
 		rows, err := f.GetRows(sheet)
 		if err != nil {
-			return fmt.Errorf("read template sheet %q: %w", sheet, err)
+			return appError(CodeTemplateOpen, fmt.Sprintf("read template sheet %q", sheet)).
+				withContext("template", templatePath).
+				withContext("sheet", sheet).
+				withCause(err)
 		}
 		for r, values := range rows {
 			for c, value := range values {
@@ -311,17 +529,25 @@ func renderXLSX(templatePath, outputPath string, row map[string]string) error {
 				}
 				rendered := replacePlaceholders(value, row)
 				if hasPlaceholder(rendered) {
-					return fmt.Errorf("unresolved placeholder in %s!%s", sheet, cell)
+					return appError(CodeTemplateUnresolved, fmt.Sprintf("unresolved placeholder in %s!%s", sheet, cell)).
+						withHint("Check source data values and template placeholders.").
+						withContext("template", templatePath).
+						withContext("cell", sheet+"!"+cell)
 				}
 				if err := f.SetCellValue(sheet, cell, rendered); err != nil {
-					return fmt.Errorf("set cell %s!%s: %w", sheet, cell, err)
+					return appError(CodeOutputWrite, fmt.Sprintf("set cell %s!%s", sheet, cell)).
+						withContext("output", outputPath).
+						withCause(err)
 				}
 			}
 		}
 	}
 
 	if err := f.SaveAs(outputPath); err != nil {
-		return fmt.Errorf("save xlsx output: %w", err)
+		return appError(CodeOutputWrite, "save xlsx output").
+			withHint("Check output.path and file permissions.").
+			withContext("output", outputPath).
+			withCause(err)
 	}
 	return nil
 }
@@ -329,13 +555,19 @@ func renderXLSX(templatePath, outputPath string, row map[string]string) error {
 func renderDOCX(templatePath, outputPath string, row map[string]string) error {
 	reader, err := zip.OpenReader(templatePath)
 	if err != nil {
-		return fmt.Errorf("open docx template: %w", err)
+		return appError(CodeTemplateOpen, "open docx template").
+			withHint("Check that template.path points to a readable .docx file.").
+			withContext("template", templatePath).
+			withCause(err)
 	}
 	defer reader.Close()
 
 	output, err := os.Create(outputPath)
 	if err != nil {
-		return fmt.Errorf("create docx output: %w", err)
+		return appError(CodeOutputWrite, "create docx output").
+			withHint("Check output.path and file permissions.").
+			withContext("output", outputPath).
+			withCause(err)
 	}
 
 	writer := zip.NewWriter(output)
@@ -348,10 +580,14 @@ func renderDOCX(templatePath, outputPath string, row map[string]string) error {
 	}
 	if err := writer.Close(); err != nil {
 		_ = output.Close()
-		return fmt.Errorf("close docx output: %w", err)
+		return appError(CodeOutputWrite, "close docx output").
+			withContext("output", outputPath).
+			withCause(err)
 	}
 	if err := output.Close(); err != nil {
-		return fmt.Errorf("close docx file: %w", err)
+		return appError(CodeOutputWrite, "close docx file").
+			withContext("output", outputPath).
+			withCause(err)
 	}
 
 	return nil
@@ -364,7 +600,8 @@ func copyDocxPart(writer *zip.Writer, source *zip.File, row map[string]string) e
 
 	dest, err := writer.CreateHeader(&header)
 	if err != nil {
-		return fmt.Errorf("create docx part %s: %w", source.Name, err)
+		return appError(CodeOutputWrite, fmt.Sprintf("create docx part %s", source.Name)).
+			withCause(err)
 	}
 
 	if source.FileInfo().IsDir() {
@@ -373,13 +610,15 @@ func copyDocxPart(writer *zip.Writer, source *zip.File, row map[string]string) e
 
 	src, err := source.Open()
 	if err != nil {
-		return fmt.Errorf("open docx part %s: %w", source.Name, err)
+		return appError(CodeTemplateOpen, fmt.Sprintf("open docx part %s", source.Name)).
+			withCause(err)
 	}
 	defer src.Close()
 
 	data, err := io.ReadAll(src)
 	if err != nil {
-		return fmt.Errorf("read docx part %s: %w", source.Name, err)
+		return appError(CodeTemplateOpen, fmt.Sprintf("read docx part %s", source.Name)).
+			withCause(err)
 	}
 	if isDocxXMLPart(source.Name) {
 		rendered, err := renderDocxXML(string(data), row)
@@ -387,13 +626,15 @@ func copyDocxPart(writer *zip.Writer, source *zip.File, row map[string]string) e
 			return fmt.Errorf("render docx part %s: %w", source.Name, err)
 		}
 		if docxHasPlaceholder(rendered) {
-			return fmt.Errorf("unresolved placeholder in docx part %s", source.Name)
+			return appError(CodeTemplateUnresolved, fmt.Sprintf("unresolved placeholder in docx part %s", source.Name)).
+				withHint("Check source data values and template placeholders.")
 		}
 		data = []byte(rendered)
 	}
 
 	if _, err := dest.Write(data); err != nil {
-		return fmt.Errorf("write docx part %s: %w", source.Name, err)
+		return appError(CodeOutputWrite, fmt.Sprintf("write docx part %s", source.Name)).
+			withCause(err)
 	}
 	return nil
 }
@@ -471,12 +712,20 @@ func renderContext(vars, row map[string]string, rowNumber int) map[string]string
 func sqliteParams(params []string, vars map[string]string) ([]any, error) {
 	args := make([]any, 0, len(params))
 	for i, param := range params {
-		if err := ensurePlaceholdersAvailable(fmt.Sprintf("sqlite params[%d]", i), param, vars); err != nil {
+		if err := ensurePlaceholdersAvailableWithCode(
+			fmt.Sprintf("sqlite params[%d]", i),
+			param,
+			vars,
+			CodeDataSQLite,
+			"Provide the missing SQLite parameter with vars or --var.",
+			nil,
+		); err != nil {
 			return nil, err
 		}
 		rendered := replacePlaceholders(param, vars)
 		if hasPlaceholder(rendered) {
-			return nil, fmt.Errorf("unresolved placeholder in sqlite params[%d]", i)
+			return nil, appError(CodeDataSQLite, fmt.Sprintf("unresolved placeholder in sqlite params[%d]", i)).
+				withHint("Provide the missing SQLite parameter with vars or --var.")
 		}
 		args = append(args, rendered)
 	}
@@ -490,7 +739,10 @@ func validateTemplateFields(templatePath string, row map[string]string) error {
 	}
 	for field := range fields {
 		if _, ok := row[field]; !ok {
-			return fmt.Errorf("missing template field %q", field)
+			return appError(CodeTemplateFieldMissing, fmt.Sprintf("missing template field %q", field)).
+				withHint(fmt.Sprintf("Add column %q to the data source or pass --var %s=...", field, field)).
+				withContext("field", field).
+				withContext("template", templatePath)
 		}
 	}
 	return nil
@@ -510,7 +762,10 @@ func templatePlaceholders(templatePath string) (map[string]struct{}, error) {
 func xlsxPlaceholders(templatePath string) (map[string]struct{}, error) {
 	f, err := excelize.OpenFile(templatePath)
 	if err != nil {
-		return nil, fmt.Errorf("open xlsx template: %w", err)
+		return nil, appError(CodeTemplateOpen, "open xlsx template").
+			withHint("Check that template.path points to a readable .xlsx file.").
+			withContext("template", templatePath).
+			withCause(err)
 	}
 	defer f.Close()
 
@@ -518,7 +773,10 @@ func xlsxPlaceholders(templatePath string) (map[string]struct{}, error) {
 	for _, sheet := range f.GetSheetList() {
 		rows, err := f.GetRows(sheet)
 		if err != nil {
-			return nil, fmt.Errorf("read template sheet %q: %w", sheet, err)
+			return nil, appError(CodeTemplateOpen, fmt.Sprintf("read template sheet %q", sheet)).
+				withContext("template", templatePath).
+				withContext("sheet", sheet).
+				withCause(err)
 		}
 		for _, row := range rows {
 			for _, value := range row {
@@ -532,7 +790,10 @@ func xlsxPlaceholders(templatePath string) (map[string]struct{}, error) {
 func docxPlaceholders(templatePath string) (map[string]struct{}, error) {
 	reader, err := zip.OpenReader(templatePath)
 	if err != nil {
-		return nil, fmt.Errorf("open docx template: %w", err)
+		return nil, appError(CodeTemplateOpen, "open docx template").
+			withHint("Check that template.path points to a readable .docx file.").
+			withContext("template", templatePath).
+			withCause(err)
 	}
 	defer reader.Close()
 
@@ -543,15 +804,21 @@ func docxPlaceholders(templatePath string) (map[string]struct{}, error) {
 		}
 		rc, err := file.Open()
 		if err != nil {
-			return nil, fmt.Errorf("open docx part %s: %w", file.Name, err)
+			return nil, appError(CodeTemplateOpen, fmt.Sprintf("open docx part %s", file.Name)).
+				withContext("template", templatePath).
+				withCause(err)
 		}
 		data, readErr := io.ReadAll(rc)
 		closeErr := rc.Close()
 		if readErr != nil {
-			return nil, fmt.Errorf("read docx part %s: %w", file.Name, readErr)
+			return nil, appError(CodeTemplateOpen, fmt.Sprintf("read docx part %s", file.Name)).
+				withContext("template", templatePath).
+				withCause(readErr)
 		}
 		if closeErr != nil {
-			return nil, fmt.Errorf("close docx part %s: %w", file.Name, closeErr)
+			return nil, appError(CodeTemplateOpen, fmt.Sprintf("close docx part %s", file.Name)).
+				withContext("template", templatePath).
+				withCause(closeErr)
 		}
 		xml := string(data)
 		addPlaceholders(fields, xml)
@@ -561,11 +828,28 @@ func docxPlaceholders(templatePath string) (map[string]struct{}, error) {
 }
 
 func ensurePlaceholdersAvailable(label, text string, row map[string]string) error {
+	return ensurePlaceholdersAvailableWithCode(
+		label,
+		text,
+		row,
+		CodeTemplateFieldMissing,
+		"Add the missing field to the data source or pass it with --var.",
+		nil,
+	)
+}
+
+func ensurePlaceholdersAvailableWithCode(label, text string, row map[string]string, code ErrorCode, hint string, context map[string]string) error {
 	fields := make(map[string]struct{})
 	addPlaceholders(fields, text)
 	for field := range fields {
 		if _, ok := row[field]; !ok {
-			return fmt.Errorf("missing %s field %q", label, field)
+			err := appError(code, fmt.Sprintf("missing %s field %q", label, field)).
+				withHint(hint).
+				withContext("field", field)
+			for key, value := range context {
+				err.withContext(key, value)
+			}
+			return err
 		}
 	}
 	return nil
@@ -637,7 +921,8 @@ func renderDocxTextGroup(xml string, row map[string]string) (string, error) {
 	}
 	rendered := replaceDocxPlaceholders(joined, row)
 	if hasPlaceholder(rendered) {
-		return "", errors.New("unresolved placeholder in docx text")
+		return "", appError(CodeTemplateUnresolved, "unresolved placeholder in docx text").
+			withHint("Check source data values and template placeholders.")
 	}
 
 	var out strings.Builder
diff --git a/internal/docfill/docfill_test.go b/internal/docfill/docfill_test.go
index 5ae97b5..f09fb25 100644
--- a/internal/docfill/docfill_test.go
+++ b/internal/docfill/docfill_test.go
@@ -153,6 +153,30 @@ func TestRenderDOCXReplacesPlaceholderSplitAcrossTextNodes(t *testing.T) {
 	}
 }
 
+func TestRenderDOCXReplacesMultipleSplitPlaceholdersInSameParagraph(t *testing.T) {
+	dir := t.TempDir()
+	templatePath := filepath.Join(dir, "template.docx")
+	outputPath := filepath.Join(dir, "out.docx")
+	writeMinimalDocx(t, templatePath, `<w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:body><w:p><w:r><w:t>员工：{{na</w:t></w:r><w:r><w:t>me}}</w:t></w:r><w:r><w:t>，完成：{{do</w:t></w:r><w:r><w:t>ne}}</w:t></w:r><w:r><w:t>，备注：{{note}}</w:t></w:r></w:p></w:body></w:document>`)
+
+	err := renderTemplate(templatePath, outputPath, map[string]string{
+		"name": "Alice",
+		"done": "接口数据整理",
+		"note": "保持同段落",
+	})
+	if err != nil {
+		t.Fatalf("renderTemplate returned error: %v", err)
+	}
+
+	got := readZipEntry(t, outputPath, "word/document.xml")
+	if !strings.Contains(got, "员工：Alice，完成：接口数据整理，备注：保持同段落") {
+		t.Fatalf("document.xml = %q, want rendered split placeholders", got)
+	}
+	if strings.Contains(got, "{{") || strings.Contains(got, "}}") {
+		t.Fatalf("document.xml still has placeholder markers: %q", got)
+	}
+}
+
 func TestRunGeneratesOneOutputPerInputRow(t *testing.T) {
 	dir := t.TempDir()
 	dataPath := filepath.Join(dir, "data.xlsx")
diff --git a/internal/docfill/errors.go b/internal/docfill/errors.go
new file mode 100644
index 0000000..4eebb01
--- /dev/null
+++ b/internal/docfill/errors.go
@@ -0,0 +1,74 @@
+package docfill
+
+type ErrorCode string
+
+const (
+	CodeCLIUsage             ErrorCode = "CLI_USAGE"
+	CodeConfigRead           ErrorCode = "CONFIG_READ"
+	CodeConfigParse          ErrorCode = "CONFIG_PARSE"
+	CodeConfigValidate       ErrorCode = "CONFIG_VALIDATE"
+	CodeDataExcel            ErrorCode = "DATA_EXCEL"
+	CodeDataSQLite           ErrorCode = "DATA_SQLITE"
+	CodeDataEmpty            ErrorCode = "DATA_EMPTY"
+	CodeDataFilterField      ErrorCode = "DATA_FILTER_FIELD"
+	CodeDataFilterValue      ErrorCode = "DATA_FILTER_VALUE"
+	CodeDataFilterNoMatch    ErrorCode = "DATA_FILTER_NO_MATCH"
+	CodeDataFilterMultiMatch ErrorCode = "DATA_FILTER_MULTI_MATCH"
+	CodeTemplateOpen         ErrorCode = "TEMPLATE_OPEN"
+	CodeTemplateFieldMissing ErrorCode = "TEMPLATE_FIELD_MISSING"
+	CodeTemplateUnresolved   ErrorCode = "TEMPLATE_UNRESOLVED"
+	CodeOutputExists         ErrorCode = "OUTPUT_EXISTS"
+	CodeOutputWrite          ErrorCode = "OUTPUT_WRITE"
+	CodeInternal             ErrorCode = "INTERNAL"
+)
+
+type AppError struct {
+	Code    ErrorCode
+	Message string
+	Hint    string
+	Context map[string]string
+	Cause   error
+}
+
+func (e *AppError) Error() string {
+	if e == nil {
+		return ""
+	}
+	if e.Cause == nil {
+		return e.Message
+	}
+	return e.Message + ": " + e.Cause.Error()
+}
+
+func (e *AppError) Unwrap() error {
+	if e == nil {
+		return nil
+	}
+	return e.Cause
+}
+
+func appError(code ErrorCode, message string) *AppError {
+	return &AppError{
+		Code:    code,
+		Message: message,
+		Context: make(map[string]string),
+	}
+}
+
+func (e *AppError) withHint(hint string) *AppError {
+	e.Hint = hint
+	return e
+}
+
+func (e *AppError) withCause(cause error) *AppError {
+	e.Cause = cause
+	return e
+}
+
+func (e *AppError) withContext(key, value string) *AppError {
+	if e.Context == nil {
+		e.Context = make(map[string]string)
+	}
+	e.Context[key] = value
+	return e
+}
diff --git a/internal/docfill/p0_test.go b/internal/docfill/p0_test.go
index b4d5bb5..1053df5 100644
--- a/internal/docfill/p0_test.go
+++ b/internal/docfill/p0_test.go
@@ -2,6 +2,7 @@ package docfill
 
 import (
 	"database/sql"
+	"errors"
 	"os"
 	"path/filepath"
 	"strings"
@@ -176,6 +177,16 @@ func TestRunRefusesToOverwriteUnlessConfigured(t *testing.T) {
 	if err == nil || !strings.Contains(err.Error(), "already exists") {
 		t.Fatalf("expected already exists error, got %v", err)
 	}
+	var appErr *AppError
+	if !errors.As(err, &appErr) {
+		t.Fatalf("expected AppError, got %T", err)
+	}
+	if appErr.Code != CodeOutputExists {
+		t.Fatalf("error code = %s, want %s", appErr.Code, CodeOutputExists)
+	}
+	if appErr.Hint == "" || appErr.Context["output"] != outputPath {
+		t.Fatalf("expected hint and output context, got %#v", appErr)
+	}
 
 	writeJSON(t, configPath, map[string]any{
 		"data": map[string]any{
@@ -226,6 +237,19 @@ func TestRunFailsWhenTemplatePlaceholderIsMissing(t *testing.T) {
 	if err == nil || !strings.Contains(err.Error(), "missing template field") {
 		t.Fatalf("expected missing template field error, got %v", err)
 	}
+	var appErr *AppError
+	if !errors.As(err, &appErr) {
+		t.Fatalf("expected AppError, got %T", err)
+	}
+	if appErr.Code != CodeTemplateFieldMissing {
+		t.Fatalf("error code = %s, want %s", appErr.Code, CodeTemplateFieldMissing)
+	}
+	if appErr.Context["field"] != "missing" || appErr.Context["template"] != templatePath {
+		t.Fatalf("unexpected context: %#v", appErr.Context)
+	}
+	if !strings.Contains(appErr.Hint, "missing") {
+		t.Fatalf("unexpected hint: %q", appErr.Hint)
+	}
 }
 
 func TestRunFailsWhenRenderedTemplateStillHasPlaceholder(t *testing.T) {
@@ -259,6 +283,282 @@ func TestRunFailsWhenRenderedTemplateStillHasPlaceholder(t *testing.T) {
 	}
 }
 
+func TestRunFiltersExcelRowsByMultipleEqualsAndVars(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "tickets.xlsx")
+	templatePath := filepath.Join(dir, "ticket-template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeWorkbook(t, dataPath, "工单", [][]string{
+		{"日期", "工单编号", "客户名称", "处理结果"},
+		{"2026-06-18", "GD-001", "A 公司", "已派单"},
+		{"2026-06-18", "GD-002", "B 公司", "已完成"},
+		{"2026-06-19", "GD-001", "C 公司", "待复核"},
+	})
+	writeWorkbook(t, templatePath, "工单", [][]string{
+		{"编号", "{{工单编号}}"},
+		{"客户", "{{客户名称}}"},
+		{"结果", "{{处理结果}}"},
+	})
+	writeJSON(t, configPath, map[string]any{
+		"vars": map[string]any{
+			"date": "2026-06-17",
+		},
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "工单",
+		},
+		"filter": map[string]any{
+			"equals": map[string]any{
+				"日期":   "{{date}}",
+				"工单编号": "{{ticket_no}}",
+			},
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "工单_{{日期}}_{{工单编号}}.xlsx"),
+		},
+	})
+
+	result, err := RunWithOptions(configPath, RunOptions{
+		Vars: map[string]string{
+			"date":      "2026-06-18",
+			"ticket_no": "GD-002",
+		},
+	})
+	if err != nil {
+		t.Fatalf("RunWithOptions returned error: %v", err)
+	}
+
+	wantPath := filepath.Join(dir, "out", "工单_2026-06-18_GD-002.xlsx")
+	if len(result.Files) != 1 || result.Files[0] != wantPath {
+		t.Fatalf("result files = %#v, want only %q", result.Files, wantPath)
+	}
+	assertWorkbookCell(t, wantPath, "工单", "B2", "B 公司")
+	assertWorkbookCell(t, wantPath, "工单", "B3", "已完成")
+}
+
+func TestRunFilterAllowsManyRowsWhenConfigured(t *testing.T) {
+	dir := t.TempDir()
+	dbPath := filepath.Join(dir, "tickets.db")
+	templatePath := filepath.Join(dir, "ticket-template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	seedP0SQLite(t, dbPath, `
+		create table tickets (
+			report_date text,
+			department text,
+			ticket_no text,
+			result text
+		);
+		insert into tickets values
+			('2026-06-18', '营销部', 'GD-001', '已派单'),
+			('2026-06-18', '客服部', 'GD-002', '已完成'),
+			('2026-06-19', '营销部', 'GD-003', '待复核');
+	`)
+	writeWorkbook(t, templatePath, "工单", [][]string{
+		{"编号", "{{ticket_no}}"},
+		{"部门", "{{department}}"},
+		{"结果", "{{result}}"},
+	})
+	writeJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "sqlite",
+			"path":  dbPath,
+			"query": "select report_date, department, ticket_no, result from tickets order by ticket_no",
+		},
+		"filter": map[string]any{
+			"equals": map[string]any{
+				"report_date": "{{date}}",
+			},
+			"expect": "many",
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "工单_{{ticket_no}}_{{_row}}.xlsx"),
+		},
+	})
+
+	result, err := RunWithOptions(configPath, RunOptions{Vars: map[string]string{"date": "2026-06-18"}})
+	if err != nil {
+		t.Fatalf("RunWithOptions returned error: %v", err)
+	}
+
+	if len(result.Files) != 2 {
+		t.Fatalf("expected 2 filtered files, got %#v", result.Files)
+	}
+	assertWorkbookCell(t, filepath.Join(dir, "out", "工单_GD-001_1.xlsx"), "工单", "B3", "已派单")
+	assertWorkbookCell(t, filepath.Join(dir, "out", "工单_GD-002_2.xlsx"), "工单", "B2", "客服部")
+}
+
+func TestRunFilterFailsWhenNoRowsMatch(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "tickets.xlsx")
+	templatePath := filepath.Join(dir, "template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeWorkbook(t, dataPath, "工单", [][]string{
+		{"日期", "工单编号", "处理结果"},
+		{"2026-06-18", "GD-001", "已派单"},
+	})
+	writeWorkbook(t, templatePath, "工单", [][]string{{"{{工单编号}}", "{{处理结果}}"}})
+	writeJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "工单",
+		},
+		"filter": map[string]any{
+			"equals": map[string]any{
+				"日期":   "{{date}}",
+				"工单编号": "{{ticket_no}}",
+			},
+			"expect": "one",
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "工单_{{工单编号}}.xlsx"),
+		},
+	})
+
+	err := RunWithOptionsError(configPath, RunOptions{Vars: map[string]string{
+		"date":      "2026-06-18",
+		"ticket_no": "GD-999",
+	}})
+	assertAppErrorCode(t, err, ErrorCode("DATA_FILTER_NO_MATCH"))
+}
+
+func TestRunFilterFailsWhenOneRowExpectedButManyMatch(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "tickets.xlsx")
+	templatePath := filepath.Join(dir, "template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeWorkbook(t, dataPath, "工单", [][]string{
+		{"日期", "工单编号", "处理结果"},
+		{"2026-06-18", "GD-001", "已派单"},
+		{"2026-06-18", "GD-002", "已完成"},
+	})
+	writeWorkbook(t, templatePath, "工单", [][]string{{"{{工单编号}}", "{{处理结果}}"}})
+	writeJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "工单",
+		},
+		"filter": map[string]any{
+			"equals": map[string]any{
+				"日期": "{{date}}",
+			},
+			"expect": "one",
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "工单_{{工单编号}}.xlsx"),
+		},
+	})
+
+	err := RunWithOptionsError(configPath, RunOptions{Vars: map[string]string{"date": "2026-06-18"}})
+	assertAppErrorCode(t, err, ErrorCode("DATA_FILTER_MULTI_MATCH"))
+}
+
+func TestRunFilterFailsWhenFieldIsMissing(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "tickets.xlsx")
+	templatePath := filepath.Join(dir, "template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeWorkbook(t, dataPath, "工单", [][]string{
+		{"日期", "处理结果"},
+		{"2026-06-18", "已派单"},
+	})
+	writeWorkbook(t, templatePath, "工单", [][]string{{"{{处理结果}}"}})
+	writeJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "工单",
+		},
+		"filter": map[string]any{
+			"equals": map[string]any{
+				"工单编号": "{{ticket_no}}",
+			},
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "工单.xlsx"),
+		},
+	})
+
+	err := RunWithOptionsError(configPath, RunOptions{Vars: map[string]string{"ticket_no": "GD-001"}})
+	assertAppErrorCode(t, err, ErrorCode("DATA_FILTER_FIELD"))
+}
+
+func TestRunFilterFailsWhenVariableIsMissing(t *testing.T) {
+	dir := t.TempDir()
+	dataPath := filepath.Join(dir, "tickets.xlsx")
+	templatePath := filepath.Join(dir, "template.xlsx")
+	configPath := filepath.Join(dir, "task.json")
+
+	writeWorkbook(t, dataPath, "工单", [][]string{
+		{"日期", "工单编号", "处理结果"},
+		{"2026-06-18", "GD-001", "已派单"},
+	})
+	writeWorkbook(t, templatePath, "工单", [][]string{{"{{工单编号}}", "{{处理结果}}"}})
+	writeJSON(t, configPath, map[string]any{
+		"data": map[string]any{
+			"type":  "excel",
+			"path":  dataPath,
+			"sheet": "工单",
+		},
+		"filter": map[string]any{
+			"equals": map[string]any{
+				"工单编号": "{{ticket_no}}",
+			},
+		},
+		"template": map[string]any{
+			"path": templatePath,
+		},
+		"output": map[string]any{
+			"path": filepath.Join(dir, "out", "工单_{{工单编号}}.xlsx"),
+		},
+	})
+
+	err := RunWithOptionsError(configPath, RunOptions{})
+	assertAppErrorCode(t, err, ErrorCode("DATA_FILTER_VALUE"))
+}
+
+func RunWithOptionsError(configPath string, opts RunOptions) error {
+	_, err := RunWithOptions(configPath, opts)
+	return err
+}
+
+func assertAppErrorCode(t *testing.T, err error, want ErrorCode) {
+	t.Helper()
+
+	if err == nil {
+		t.Fatalf("expected %s error, got nil", want)
+	}
+	var appErr *AppError
+	if !errors.As(err, &appErr) {
+		t.Fatalf("expected AppError, got %T: %v", err, err)
+	}
+	if appErr.Code != want {
+		t.Fatalf("error code = %s, want %s; error=%v", appErr.Code, want, err)
+	}
+}
+
 func seedP0SQLite(t *testing.T, path, statements string) {
 	t.Helper()
 
diff --git a/manual_tests/01_excel_to_word/daily_report_template.docx b/manual_tests/01_excel_to_word/daily_report_template.docx
new file mode 100644
index 0000000..3f6f1f6
Binary files /dev/null and b/manual_tests/01_excel_to_word/daily_report_template.docx differ
diff --git a/manual_tests/01_excel_to_word/out/daily_2026-06-18_张三_测试_1.docx b/manual_tests/01_excel_to_word/out/daily_2026-06-18_张三_测试_1.docx
new file mode 100644
index 0000000..a2c17a2
Binary files /dev/null and b/manual_tests/01_excel_to_word/out/daily_2026-06-18_张三_测试_1.docx differ
diff --git a/manual_tests/01_excel_to_word/out/daily_2026-06-18_李四_2.docx b/manual_tests/01_excel_to_word/out/daily_2026-06-18_李四_2.docx
new file mode 100644
index 0000000..fd5f06c
Binary files /dev/null and b/manual_tests/01_excel_to_word/out/daily_2026-06-18_李四_2.docx differ
diff --git a/manual_tests/01_excel_to_word/source_daily_data.xlsx b/manual_tests/01_excel_to_word/source_daily_data.xlsx
new file mode 100644
index 0000000..2d9e57c
Binary files /dev/null and b/manual_tests/01_excel_to_word/source_daily_data.xlsx differ
diff --git a/manual_tests/01_excel_to_word/task.json b/manual_tests/01_excel_to_word/task.json
new file mode 100644
index 0000000..b955da5
--- /dev/null
+++ b/manual_tests/01_excel_to_word/task.json
@@ -0,0 +1,19 @@
+{
+  "vars": {
+    "date": "2026-06-18",
+    "batch": "默认班次"
+  },
+  "data": {
+    "type": "excel",
+    "path": "./source_daily_data.xlsx",
+    "sheet": "Daily",
+    "header_row": 1
+  },
+  "template": {
+    "path": "./daily_report_template.docx"
+  },
+  "output": {
+    "path": "./out/daily_{{date}}_{{name}}_{{_row}}.docx",
+    "overwrite": true
+  }
+}
\ No newline at end of file
diff --git a/manual_tests/02_sqlite_to_excel/department_summary_template.xlsx b/manual_tests/02_sqlite_to_excel/department_summary_template.xlsx
new file mode 100644
index 0000000..c2b029f
Binary files /dev/null and b/manual_tests/02_sqlite_to_excel/department_summary_template.xlsx differ
diff --git a/manual_tests/02_sqlite_to_excel/out/summary_客服部_1.xlsx b/manual_tests/02_sqlite_to_excel/out/summary_客服部_1.xlsx
new file mode 100755
index 0000000..fd4362f
Binary files /dev/null and b/manual_tests/02_sqlite_to_excel/out/summary_客服部_1.xlsx differ
diff --git a/manual_tests/02_sqlite_to_excel/out/summary_营销部_2.xlsx b/manual_tests/02_sqlite_to_excel/out/summary_营销部_2.xlsx
new file mode 100755
index 0000000..dd26c30
Binary files /dev/null and b/manual_tests/02_sqlite_to_excel/out/summary_营销部_2.xlsx differ
diff --git a/manual_tests/02_sqlite_to_excel/source_summary.db b/manual_tests/02_sqlite_to_excel/source_summary.db
new file mode 100644
index 0000000..38362de
Binary files /dev/null and b/manual_tests/02_sqlite_to_excel/source_summary.db differ
diff --git a/manual_tests/02_sqlite_to_excel/task.json b/manual_tests/02_sqlite_to_excel/task.json
new file mode 100644
index 0000000..0564aa7
--- /dev/null
+++ b/manual_tests/02_sqlite_to_excel/task.json
@@ -0,0 +1,20 @@
+{
+  "vars": {
+    "date": "2026-06-18"
+  },
+  "data": {
+    "type": "sqlite",
+    "path": "./source_summary.db",
+    "query": "select department, report_date, total_count, abnormal_count, owner from department_summary where report_date = ? order by department",
+    "params": [
+      "{{date}}"
+    ]
+  },
+  "template": {
+    "path": "./department_summary_template.xlsx"
+  },
+  "output": {
+    "path": "./out/summary_{{department}}_{{_row}}.xlsx",
+    "overwrite": true
+  }
+}
\ No newline at end of file
diff --git a/manual_tests/03_excel_to_excel/audit_template.xlsx b/manual_tests/03_excel_to_excel/audit_template.xlsx
new file mode 100644
index 0000000..178c064
Binary files /dev/null and b/manual_tests/03_excel_to_excel/audit_template.xlsx differ
diff --git a/manual_tests/03_excel_to_excel/out/audit_2026-06-18_王五_1.xlsx b/manual_tests/03_excel_to_excel/out/audit_2026-06-18_王五_1.xlsx
new file mode 100755
index 0000000..1897d2d
Binary files /dev/null and b/manual_tests/03_excel_to_excel/out/audit_2026-06-18_王五_1.xlsx differ
diff --git a/manual_tests/03_excel_to_excel/out/audit_2026-06-18_赵六_2.xlsx b/manual_tests/03_excel_to_excel/out/audit_2026-06-18_赵六_2.xlsx
new file mode 100755
index 0000000..b89752b
Binary files /dev/null and b/manual_tests/03_excel_to_excel/out/audit_2026-06-18_赵六_2.xlsx differ
diff --git a/manual_tests/03_excel_to_excel/source_audit_data.xlsx b/manual_tests/03_excel_to_excel/source_audit_data.xlsx
new file mode 100644
index 0000000..180ab5f
Binary files /dev/null and b/manual_tests/03_excel_to_excel/source_audit_data.xlsx differ
diff --git a/manual_tests/03_excel_to_excel/task.json b/manual_tests/03_excel_to_excel/task.json
new file mode 100644
index 0000000..75e8fd7
--- /dev/null
+++ b/manual_tests/03_excel_to_excel/task.json
@@ -0,0 +1,18 @@
+{
+  "vars": {
+    "date": "2026-06-18"
+  },
+  "data": {
+    "type": "excel",
+    "path": "./source_audit_data.xlsx",
+    "sheet": "Audit",
+    "header_row": 2
+  },
+  "template": {
+    "path": "./audit_template.xlsx"
+  },
+  "output": {
+    "path": "./out/audit_{{date}}_{{employee}}_{{_row}}.xlsx",
+    "overwrite": true
+  }
+}
\ No newline at end of file
diff --git a/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx b/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
new file mode 100644
index 0000000..f8954c4
Binary files /dev/null and b/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx differ
diff --git a/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx b/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx
new file mode 100644
index 0000000..3d7777b
Binary files /dev/null and b/manual_tests/04_filter_work_order_to_word/source_work_orders.xlsx differ
diff --git a/manual_tests/04_filter_work_order_to_word/task.json b/manual_tests/04_filter_work_order_to_word/task.json
new file mode 100644
index 0000000..d623e8c
--- /dev/null
+++ b/manual_tests/04_filter_work_order_to_word/task.json
@@ -0,0 +1,25 @@
+{
+  "vars": {
+    "date": "2026-06-18"
+  },
+  "data": {
+    "type": "excel",
+    "path": "./source_work_orders.xlsx",
+    "sheet": "工单",
+    "header_row": 1
+  },
+  "filter": {
+    "equals": {
+      "日期": "{{date}}",
+      "工单编号": "{{ticket_no}}"
+    },
+    "expect": "one"
+  },
+  "template": {
+    "path": "./work_order_template.docx"
+  },
+  "output": {
+    "path": "./out/work_order_{{日期}}_{{工单编号}}.docx",
+    "overwrite": true
+  }
+}
\ No newline at end of file
diff --git a/manual_tests/04_filter_work_order_to_word/work_order_template.docx b/manual_tests/04_filter_work_order_to_word/work_order_template.docx
new file mode 100644
index 0000000..7f05070
Binary files /dev/null and b/manual_tests/04_filter_work_order_to_word/work_order_template.docx differ
diff --git a/manual_tests/README.md b/manual_tests/README.md
new file mode 100644
index 0000000..32ce752
--- /dev/null
+++ b/manual_tests/README.md
@@ -0,0 +1,131 @@
+# docfill 手工测试素材
+
+本目录提供四组可直接运行的本地测试素材，用于验证 `docfill` 对筛选、Word、Excel、Excel 数据源和 SQLite 数据源的支持情况。
+
+运行前进入项目根目录：
+
+```bash
+cd /Users/zhaoyilun/Documents/xls_doc
+```
+
+构建工具：
+
+```bash
+bash scripts/build.sh
+```
+
+## 案例 1：Excel 数据生成 Word 日报
+
+目录：
+
+```text
+manual_tests/01_excel_to_word
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/01_excel_to_word/task.json --var batch=早班 --verbose
+```
+
+输出示例：
+
+```text
+generated 2 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/01_excel_to_word/out/daily_2026-06-18_张三_测试_1.docx
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/01_excel_to_word/out/daily_2026-06-18_李四_2.docx
+```
+
+## 案例 2：SQLite 数据生成 Excel 汇总表
+
+目录：
+
+```text
+manual_tests/02_sqlite_to_excel
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/02_sqlite_to_excel/task.json --var date=2026-06-18 --verbose
+```
+
+输出示例：
+
+```text
+generated 2 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/02_sqlite_to_excel/out/summary_客服部_1.xlsx
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/02_sqlite_to_excel/out/summary_营销部_2.xlsx
+```
+
+## 案例 3：Excel 数据生成 Excel 质检表
+
+目录：
+
+```text
+manual_tests/03_excel_to_excel
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/03_excel_to_excel/task.json --var date=2026-06-18 --verbose
+```
+
+输出示例：
+
+```text
+generated 2 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/03_excel_to_excel/out/audit_2026-06-18_王五_1.xlsx
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/03_excel_to_excel/out/audit_2026-06-18_赵六_2.xlsx
+```
+
+## 案例 4：按日期和工单编号筛选一条工单并生成 Word
+
+目录：
+
+```text
+manual_tests/04_filter_work_order_to_word
+```
+
+输入数据内容：
+
+```text
+日期          工单编号   客户名称  系统来源   处理结果  下一步
+2026-06-18   GD-001    A 公司   PMS3.0   已派单   等待现场反馈
+2026-06-18   GD-002    B 公司   D5000    已完成   归档并同步日报
+2026-06-19   GD-001    C 公司   风控平台  待复核   补充截图
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002 --verbose
+```
+
+输出示例：
+
+```text
+generated 1 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+```
+
+查看结果：
+
+```bash
+open manual_tests/04_filter_work_order_to_word/out
+```
+
+## 重新生成素材
+
+如需重新生成全部手工测试文件，执行：
+
+```bash
+/Users/zhaoyilun/.cache/codex-runtimes/codex-primary-runtime/dependencies/python/bin/python3 scripts/make-manual-test-materials.py
+```
+
+输出示例：
+
+```text
+/Users/zhaoyilun/Documents/xls_doc/manual_tests
+```
diff --git a/manual_tests/_evaluation_rendered/daily/page-1.png b/manual_tests/_evaluation_rendered/daily/page-1.png
new file mode 100644
index 0000000..9d72199
Binary files /dev/null and b/manual_tests/_evaluation_rendered/daily/page-1.png differ
diff --git a/manual_tests/_evaluation_rendered/filter/page-1.png b/manual_tests/_evaluation_rendered/filter/page-1.png
new file mode 100644
index 0000000..a5f8576
Binary files /dev/null and b/manual_tests/_evaluation_rendered/filter/page-1.png differ
diff --git a/scripts/make-manual-test-materials.py b/scripts/make-manual-test-materials.py
new file mode 100644
index 0000000..729e133
--- /dev/null
+++ b/scripts/make-manual-test-materials.py
@@ -0,0 +1,473 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import json
+import shutil
+import sqlite3
+from pathlib import Path
+
+from docx import Document
+from docx.enum.table import WD_TABLE_ALIGNMENT, WD_CELL_VERTICAL_ALIGNMENT
+from docx.shared import Inches, Pt
+from openpyxl import Workbook
+from openpyxl.styles import Alignment, Font, PatternFill, Border, Side
+from openpyxl.utils import get_column_letter
+
+
+ROOT = Path(__file__).resolve().parents[1]
+OUT = ROOT / "manual_tests"
+
+
+def main() -> None:
+    if OUT.exists():
+        shutil.rmtree(OUT)
+    OUT.mkdir(parents=True)
+
+    build_excel_to_word(OUT / "01_excel_to_word")
+    build_sqlite_to_excel(OUT / "02_sqlite_to_excel")
+    build_excel_to_excel(OUT / "03_excel_to_excel")
+    build_filter_work_order_to_word(OUT / "04_filter_work_order_to_word")
+    write_readme()
+
+    print(OUT)
+
+
+def build_excel_to_word(path: Path) -> None:
+    path.mkdir(parents=True)
+    write_daily_source_xlsx(path / "source_daily_data.xlsx")
+    write_daily_word_template(path / "daily_report_template.docx")
+    write_json(
+        path / "task.json",
+        {
+            "vars": {"date": "2026-06-18", "batch": "默认班次"},
+            "data": {
+                "type": "excel",
+                "path": "./source_daily_data.xlsx",
+                "sheet": "Daily",
+                "header_row": 1,
+            },
+            "template": {"path": "./daily_report_template.docx"},
+            "output": {
+                "path": "./out/daily_{{date}}_{{name}}_{{_row}}.docx",
+                "overwrite": True,
+            },
+        },
+    )
+
+
+def build_sqlite_to_excel(path: Path) -> None:
+    path.mkdir(parents=True)
+    write_summary_sqlite(path / "source_summary.db")
+    write_summary_excel_template(path / "department_summary_template.xlsx")
+    write_json(
+        path / "task.json",
+        {
+            "vars": {"date": "2026-06-18"},
+            "data": {
+                "type": "sqlite",
+                "path": "./source_summary.db",
+                "query": "select department, report_date, total_count, abnormal_count, owner from department_summary where report_date = ? order by department",
+                "params": ["{{date}}"],
+            },
+            "template": {"path": "./department_summary_template.xlsx"},
+            "output": {
+                "path": "./out/summary_{{department}}_{{_row}}.xlsx",
+                "overwrite": True,
+            },
+        },
+    )
+
+
+def build_excel_to_excel(path: Path) -> None:
+    path.mkdir(parents=True)
+    write_audit_source_xlsx(path / "source_audit_data.xlsx")
+    write_audit_excel_template(path / "audit_template.xlsx")
+    write_json(
+        path / "task.json",
+        {
+            "vars": {"date": "2026-06-18"},
+            "data": {
+                "type": "excel",
+                "path": "./source_audit_data.xlsx",
+                "sheet": "Audit",
+                "header_row": 2,
+            },
+            "template": {"path": "./audit_template.xlsx"},
+            "output": {
+                "path": "./out/audit_{{date}}_{{employee}}_{{_row}}.xlsx",
+                "overwrite": True,
+            },
+        },
+    )
+
+
+def build_filter_work_order_to_word(path: Path) -> None:
+    path.mkdir(parents=True)
+    write_work_order_source_xlsx(path / "source_work_orders.xlsx")
+    write_work_order_word_template(path / "work_order_template.docx")
+    write_json(
+        path / "task.json",
+        {
+            "vars": {"date": "2026-06-18"},
+            "data": {
+                "type": "excel",
+                "path": "./source_work_orders.xlsx",
+                "sheet": "工单",
+                "header_row": 1,
+            },
+            "filter": {
+                "equals": {
+                    "日期": "{{date}}",
+                    "工单编号": "{{ticket_no}}",
+                },
+                "expect": "one",
+            },
+            "template": {"path": "./work_order_template.docx"},
+            "output": {
+                "path": "./out/work_order_{{日期}}_{{工单编号}}.docx",
+                "overwrite": True,
+            },
+        },
+    )
+
+
+def write_daily_source_xlsx(path: Path) -> None:
+    rows = [
+        ["name", "department", "done", "plan", "risk"],
+        ["张三/测试", "营销部", "完成 A 系统订单核对 128 条；同步 B 系统回款 35 笔", "跟进 3 条异常订单并补充截图", "无"],
+        ["李四", "客服部", "整理投诉工单 21 条；完成日报提交", "复核高优先级工单并同步主管", "等待业务确认 1 条"],
+    ]
+    write_table_workbook(path, "Daily", rows, title="日报源数据")
+
+
+def write_daily_word_template(path: Path) -> None:
+    doc = Document()
+    section = doc.sections[0]
+    section.top_margin = Inches(0.8)
+    section.bottom_margin = Inches(0.8)
+    section.left_margin = Inches(0.85)
+    section.right_margin = Inches(0.85)
+
+    title = doc.add_paragraph()
+    title_run = title.add_run("员工日报（测试模板）")
+    title_run.bold = True
+    title_run.font.size = Pt(18)
+
+    meta = doc.add_table(rows=4, cols=2)
+    meta.alignment = WD_TABLE_ALIGNMENT.LEFT
+    set_table_style(meta)
+    labels = ["日期", "员工", "部门", "班次"]
+    values = ["{{date}}", "{{na", "{{department}}", "{{batch}}"]
+    for i, label in enumerate(labels):
+        meta.cell(i, 0).text = label
+        if label == "员工":
+            p = meta.cell(i, 1).paragraphs[0]
+            p.add_run(values[i])
+            p.add_run("me}}")
+        else:
+            meta.cell(i, 1).text = values[i]
+
+    for heading, placeholder in [
+        ("今日完成", "{{done}}"),
+        ("明日计划", "{{plan}}"),
+        ("风险备注", "{{risk}}"),
+    ]:
+        doc.add_paragraph()
+        h = doc.add_paragraph()
+        r = h.add_run(heading)
+        r.bold = True
+        r.font.size = Pt(12)
+        p = doc.add_paragraph()
+        p.add_run(placeholder)
+
+    doc.save(path)
+
+
+def write_summary_sqlite(path: Path) -> None:
+    conn = sqlite3.connect(path)
+    try:
+        conn.executescript(
+            """
+            create table department_summary (
+                department text,
+                report_date text,
+                total_count integer,
+                abnormal_count integer,
+                owner text
+            );
+            insert into department_summary values
+                ('客服部', '2026-06-18', 76, 0, '李四'),
+                ('营销部', '2026-06-18', 128, 3, '张三'),
+                ('营销部', '2026-06-17', 115, 2, '张三');
+            """
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
+
+def write_summary_excel_template(path: Path) -> None:
+    wb = Workbook()
+    ws = wb.active
+    ws.title = "Summary"
+    ws.append(["部门日报汇总（测试模板）", ""])
+    ws.append(["部门", "{{department}}"])
+    ws.append(["日期", "{{report_date}}"])
+    ws.append(["负责人", "{{owner}}"])
+    ws.append(["处理总数", "{{total_count}}"])
+    ws.append(["异常数", "{{abnormal_count}}"])
+    style_sheet(ws, title_rows={1})
+    ws.column_dimensions["A"].width = 18
+    ws.column_dimensions["B"].width = 32
+    wb.save(path)
+
+
+def write_audit_source_xlsx(path: Path) -> None:
+    rows = [
+        ["日报质检导出"],
+        ["employee", "date", "item", "score", "comment"],
+        ["王五", "2026-06-18", "日报完整性", "92", "字段齐全，说明清楚"],
+        ["赵六", "2026-06-18", "系统截图一致性", "87", "缺少 1 张截图，需要补充"],
+    ]
+    write_table_workbook(path, "Audit", rows, title=None)
+
+
+def write_audit_excel_template(path: Path) -> None:
+    wb = Workbook()
+    ws = wb.active
+    ws.title = "AuditReport"
+    ws.append(["日报质检结果（测试模板）", ""])
+    ws.append(["日期", "{{date}}"])
+    ws.append(["员工", "{{employee}}"])
+    ws.append(["检查项", "{{item}}"])
+    ws.append(["得分", "{{score}}"])
+    ws.append(["评语", "{{comment}}"])
+    style_sheet(ws, title_rows={1})
+    ws.column_dimensions["A"].width = 16
+    ws.column_dimensions["B"].width = 42
+    wb.save(path)
+
+
+def write_work_order_source_xlsx(path: Path) -> None:
+    rows = [
+        ["日期", "工单编号", "客户名称", "系统来源", "处理结果", "下一步"],
+        ["2026-06-18", "GD-001", "A 公司", "PMS3.0", "已派单", "等待现场反馈"],
+        ["2026-06-18", "GD-002", "B 公司", "D5000", "已完成", "归档并同步日报"],
+        ["2026-06-19", "GD-001", "C 公司", "风控平台", "待复核", "补充截图"],
+    ]
+    write_table_workbook(path, "工单", rows, title="工单源数据")
+
+
+def write_work_order_word_template(path: Path) -> None:
+    doc = Document()
+    section = doc.sections[0]
+    section.top_margin = Inches(0.8)
+    section.bottom_margin = Inches(0.8)
+    section.left_margin = Inches(0.85)
+    section.right_margin = Inches(0.85)
+
+    title = doc.add_paragraph()
+    title_run = title.add_run("工单处理记录（筛选测试模板）")
+    title_run.bold = True
+    title_run.font.size = Pt(18)
+
+    table = doc.add_table(rows=6, cols=2)
+    table.alignment = WD_TABLE_ALIGNMENT.LEFT
+    set_table_style(table)
+    rows = [
+        ("日期", "{{日期}}"),
+        ("工单编号", "{{工单编号}}"),
+        ("客户名称", "{{客户名称}}"),
+        ("系统来源", "{{系统来源}}"),
+        ("处理结果", "{{处理结果}}"),
+        ("下一步", "{{下一步}}"),
+    ]
+    for idx, (label, value) in enumerate(rows):
+        table.cell(idx, 0).text = label
+        table.cell(idx, 1).text = value
+
+    doc.save(path)
+
+
+def write_table_workbook(path: Path, sheet: str, rows: list[list[str]], title: str | None) -> None:
+    wb = Workbook()
+    ws = wb.active
+    ws.title = sheet
+    for row in rows:
+        ws.append(row)
+    title_rows = {1} if title else set()
+    style_sheet(ws, title_rows=title_rows)
+    for idx, width in enumerate([18, 16, 44, 36, 28], start=1):
+        ws.column_dimensions[get_column_letter(idx)].width = width
+    wb.save(path)
+
+
+def style_sheet(ws, title_rows: set[int]) -> None:
+    header_fill = PatternFill("solid", fgColor="D9EAF7")
+    title_fill = PatternFill("solid", fgColor="1F4E79")
+    thin = Side(style="thin", color="D9E2F3")
+    border = Border(top=thin, left=thin, right=thin, bottom=thin)
+    for row in ws.iter_rows():
+        for cell in row:
+            cell.alignment = Alignment(vertical="center", wrap_text=True)
+            cell.border = border
+            if cell.row in title_rows:
+                cell.font = Font(bold=True, color="FFFFFF", size=13)
+                cell.fill = title_fill
+            elif cell.row == min(title_rows or {1}) + 1 or (not title_rows and cell.row == 1):
+                cell.font = Font(bold=True)
+                cell.fill = header_fill
+    ws.freeze_panes = "A2"
+
+
+def set_table_style(table) -> None:
+    for row in table.rows:
+        for cell in row.cells:
+            cell.vertical_alignment = WD_CELL_VERTICAL_ALIGNMENT.CENTER
+            for paragraph in cell.paragraphs:
+                for run in paragraph.runs:
+                    run.font.size = Pt(10.5)
+
+
+def write_json(path: Path, data: dict) -> None:
+    path.write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")
+
+
+def write_readme() -> None:
+    text = """# docfill 手工测试素材
+
+本目录提供四组可直接运行的本地测试素材，用于验证 `docfill` 对筛选、Word、Excel、Excel 数据源和 SQLite 数据源的支持情况。
+
+运行前进入项目根目录：
+
+```bash
+cd /Users/zhaoyilun/Documents/xls_doc
+```
+
+构建工具：
+
+```bash
+bash scripts/build.sh
+```
+
+## 案例 1：Excel 数据生成 Word 日报
+
+目录：
+
+```text
+manual_tests/01_excel_to_word
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/01_excel_to_word/task.json --var batch=早班 --verbose
+```
+
+输出示例：
+
+```text
+generated 2 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/01_excel_to_word/out/daily_2026-06-18_张三_测试_1.docx
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/01_excel_to_word/out/daily_2026-06-18_李四_2.docx
+```
+
+## 案例 2：SQLite 数据生成 Excel 汇总表
+
+目录：
+
+```text
+manual_tests/02_sqlite_to_excel
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/02_sqlite_to_excel/task.json --var date=2026-06-18 --verbose
+```
+
+输出示例：
+
+```text
+generated 2 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/02_sqlite_to_excel/out/summary_客服部_1.xlsx
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/02_sqlite_to_excel/out/summary_营销部_2.xlsx
+```
+
+## 案例 3：Excel 数据生成 Excel 质检表
+
+目录：
+
+```text
+manual_tests/03_excel_to_excel
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/03_excel_to_excel/task.json --var date=2026-06-18 --verbose
+```
+
+输出示例：
+
+```text
+generated 2 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/03_excel_to_excel/out/audit_2026-06-18_王五_1.xlsx
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/03_excel_to_excel/out/audit_2026-06-18_赵六_2.xlsx
+```
+
+## 案例 4：按日期和工单编号筛选一条工单并生成 Word
+
+目录：
+
+```text
+manual_tests/04_filter_work_order_to_word
+```
+
+输入数据内容：
+
+```text
+日期          工单编号   客户名称  系统来源   处理结果  下一步
+2026-06-18   GD-001    A 公司   PMS3.0   已派单   等待现场反馈
+2026-06-18   GD-002    B 公司   D5000    已完成   归档并同步日报
+2026-06-19   GD-001    C 公司   风控平台  待复核   补充截图
+```
+
+运行命令：
+
+```bash
+dist/docfill-darwin-arm64 run -c manual_tests/04_filter_work_order_to_word/task.json --var date=2026-06-18 --var ticket_no=GD-002 --verbose
+```
+
+输出示例：
+
+```text
+generated 1 file(s)
+- /Users/zhaoyilun/Documents/xls_doc/manual_tests/04_filter_work_order_to_word/out/work_order_2026-06-18_GD-002.docx
+```
+
+查看结果：
+
+```bash
+open manual_tests/04_filter_work_order_to_word/out
+```
+
+## 重新生成素材
+
+如需重新生成全部手工测试文件，执行：
+
+```bash
+/Users/zhaoyilun/.cache/codex-runtimes/codex-primary-runtime/dependencies/python/bin/python3 scripts/make-manual-test-materials.py
+```
+
+输出示例：
+
+```text
+/Users/zhaoyilun/Documents/xls_doc/manual_tests
+```
+"""
+    (OUT / "README.md").write_text(text, encoding="utf-8")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/smoke-fixtures/main.go b/scripts/smoke-fixtures/main.go
index c2aa8cd..1762084 100644
--- a/scripts/smoke-fixtures/main.go
+++ b/scripts/smoke-fixtures/main.go
@@ -25,6 +25,10 @@ func main() {
 		fmt.Fprintf(os.Stderr, "write excel-to-word fixtures: %v\n", err)
 		os.Exit(1)
 	}
+	if err := writeExcelFilterToWord(filepath.Join(*base, "excel-filter-to-word")); err != nil {
+		fmt.Fprintf(os.Stderr, "write excel-filter-to-word fixtures: %v\n", err)
+		os.Exit(1)
+	}
 	if err := writeSQLiteToExcel(filepath.Join(*base, "sqlite-to-excel")); err != nil {
 		fmt.Fprintf(os.Stderr, "write sqlite-to-excel fixtures: %v\n", err)
 		os.Exit(1)
@@ -45,6 +49,21 @@ func writeExcelToWord(dir string) error {
 	return writeDocx(filepath.Join(dir, "template.docx"), `<w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:body><w:p><w:r><w:t>员工：{{na</w:t></w:r><w:r><w:t>me}}</w:t></w:r></w:p><w:p><w:r><w:t>日期：{{date}}</w:t></w:r></w:p><w:p><w:r><w:t>完成：{{done}}</w:t></w:r></w:p><w:p><w:r><w:t>计划：{{plan}}</w:t></w:r></w:p></w:body></w:document>`)
 }
 
+func writeExcelFilterToWord(dir string) error {
+	if err := os.MkdirAll(dir, 0o755); err != nil {
+		return err
+	}
+	if err := writeWorkbook(filepath.Join(dir, "data.xlsx"), "Tickets", [][]string{
+		{"date", "ticket_no", "customer", "result"},
+		{"2026-06-18", "GD-001", "A 公司", "已派单"},
+		{"2026-06-18", "GD-002", "B 公司", "已完成"},
+		{"2026-06-19", "GD-001", "C 公司", "待复核"},
+	}); err != nil {
+		return err
+	}
+	return writeDocx(filepath.Join(dir, "template.docx"), `<w:document xmlns:w="http://schemas.openxmlformats.org/wordprocessingml/2006/main"><w:body><w:p><w:r><w:t>工单：{{ticket_no}}</w:t></w:r></w:p><w:p><w:r><w:t>日期：{{date}}</w:t></w:r></w:p><w:p><w:r><w:t>客户：{{customer}}</w:t></w:r></w:p><w:p><w:r><w:t>结果：{{result}}</w:t></w:r></w:p></w:body></w:document>`)
+}
+
 func writeSQLiteToExcel(dir string) error {
 	if err := os.MkdirAll(dir, 0o755); err != nil {
 		return err
diff --git a/scripts/smoke.sh b/scripts/smoke.sh
index 7ac00c4..5837ea0 100755
--- a/scripts/smoke.sh
+++ b/scripts/smoke.sh
@@ -17,10 +17,13 @@ cp -R examples "$TMP/examples"
 go run ./scripts/smoke-fixtures --base "$TMP/examples"
 
 "$TMP/docfill" run -c "$TMP/examples/excel-to-word/task.json" --var date=2026-06-18
+"$TMP/docfill" run -c "$TMP/examples/excel-filter-to-word/task.json" --var date=2026-06-18 --var ticket_no=GD-002
 "$TMP/docfill" run -c "$TMP/examples/sqlite-to-excel/task.json" --var date=2026-06-18
 
 test -f "$TMP/examples/excel-to-word/out/daily_2026-06-18_张三_1.docx"
 test -f "$TMP/examples/excel-to-word/out/daily_2026-06-18_李四_2.docx"
+test -f "$TMP/examples/excel-filter-to-word/out/ticket_2026-06-18_GD-002.docx"
+test ! -f "$TMP/examples/excel-filter-to-word/out/ticket_2026-06-18_GD-001.docx"
 
 xlsx_count="$(find "$TMP/examples/sqlite-to-excel/out" -type f -name '*.xlsx' | wc -l | tr -d ' ')"
 if [[ "$xlsx_count" != "2" ]]; then