commit 226224162f607ffc8b42942824256e0a0285d851
Author: lily <wanglili@cenbo.cn>
Date:   Thu Sep 17 18:49:22 2026 +0800

    add:脚本初始化

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..7e77241
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,17 @@
+# local secrets / env config
+config/config.dev.yaml
+
+# generated DML dumps
+haoxian_hisense/agentic_project/check_agentic_project_name/dml_*.sql
+
+# Go
+*.exe
+*.test
+*.out
+vendor/
+
+# OS / IDE
+.DS_Store
+.idea/
+.vscode/
+*.swp
diff --git a/cmd/check_sheet_permission/main.go b/cmd/check_sheet_permission/main.go
new file mode 100644
index 0000000..82070a0
--- /dev/null
+++ b/cmd/check_sheet_permission/main.go
@@ -0,0 +1,49 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"os"
+
+	"ai_script/common_util/feishu_util"
+)
+
+func main() {
+	client, err := feishu_util.DefaultClient()
+	if err != nil {
+		fail("加载配置失败", err)
+	}
+
+	url := feishu_util.DefaultOptimizationIssueSheetURL
+	title := feishu_util.DefaultOptimizationIssueSheetTitle
+	if len(os.Args) > 1 {
+		url = os.Args[1]
+	}
+	if len(os.Args) > 2 {
+		title = os.Args[2]
+	}
+
+	ctx := context.Background()
+	meta, token, err := client.CheckReadPermission(ctx, url, title)
+	if err != nil {
+		fail("读取权限校验失败", err)
+	}
+	fmt.Printf("OK: 具备读取权限\n")
+	fmt.Printf("  spreadsheet_token: %s\n", token)
+	fmt.Printf("  sheet_id: %s\n", meta.SheetID)
+	fmt.Printf("  sheet_title: %s\n", meta.Title)
+
+	values, err := client.ReadSpreadsheetURL(ctx, url, title)
+	if err != nil {
+		fail("读取表格失败", err)
+	}
+	fmt.Printf("  rows: %d\n", len(values.Values))
+	if len(values.Values) > 0 {
+		fmt.Printf("  first_row_cols: %d\n", len(values.Values[0]))
+	}
+}
+
+func fail(msg string, err error) {
+	fmt.Fprintf(os.Stderr, "%s: %v\n", msg, err)
+	os.Exit(1)
+}
diff --git a/common_util/feishu_util/client.go b/common_util/feishu_util/client.go
new file mode 100644
index 0000000..1489e91
--- /dev/null
+++ b/common_util/feishu_util/client.go
@@ -0,0 +1,56 @@
+package feishu_util
+
+import (
+	"fmt"
+	"sync"
+
+	"ai_script/config"
+
+	lark "github.com/larksuite/oapi-sdk-go/v3"
+)
+
+var (
+	clientOnce sync.Once
+	clientInst *Client
+	clientErr  error
+)
+
+// Client 封装飞书 OpenAPI 客户端与应用凭证。
+type Client struct {
+	AppID     string
+	AppSecret string
+	Lark      *lark.Client
+}
+
+// NewClient 使用显式凭证创建客户端。
+func NewClient(appID, appSecret string) *Client {
+	return &Client{
+		AppID:     appID,
+		AppSecret: appSecret,
+		Lark:      lark.NewClient(appID, appSecret),
+	}
+}
+
+// NewClientFromConfig 从项目配置创建客户端。
+func NewClientFromConfig(cfg *config.Config) (*Client, error) {
+	if cfg == nil {
+		return nil, fmt.Errorf("config 为空")
+	}
+	if cfg.Feishu.AppID == "" || cfg.Feishu.AppSecret == "" {
+		return nil, fmt.Errorf("配置缺少 feishu.app_id / feishu.app_secret")
+	}
+	return NewClient(cfg.Feishu.AppID, cfg.Feishu.AppSecret), nil
+}
+
+// DefaultClient 懒加载：读取本地 config/config.{APP_ENV}.yaml 并创建单例客户端。
+func DefaultClient() (*Client, error) {
+	clientOnce.Do(func() {
+		cfg, err := config.Load()
+		if err != nil {
+			clientErr = err
+			return
+		}
+		clientInst, clientErr = NewClientFromConfig(cfg)
+	})
+	return clientInst, clientErr
+}
diff --git a/common_util/feishu_util/defaults.go b/common_util/feishu_util/defaults.go
new file mode 100644
index 0000000..3809f8a
--- /dev/null
+++ b/common_util/feishu_util/defaults.go
@@ -0,0 +1,7 @@
+package feishu_util
+
+// 月度优化问题汇总管理表（知识库挂载的电子表格）
+const (
+	DefaultOptimizationIssueSheetURL   = "https://mi8elatzm1x.feishu.cn/wiki/IbCUwR4bvil4E1k9k2LcCiwSn8t?sheet=QyDfOp"
+	DefaultOptimizationIssueSheetTitle = "9月各产业优化问题汇总表"
+)
diff --git a/common_util/feishu_util/sheet.go b/common_util/feishu_util/sheet.go
new file mode 100644
index 0000000..f7c396e
--- /dev/null
+++ b/common_util/feishu_util/sheet.go
@@ -0,0 +1,236 @@
+package feishu_util
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"strings"
+
+	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
+	larksheets "github.com/larksuite/oapi-sdk-go/v3/service/sheets/v3"
+)
+
+// SheetMeta 工作表元信息。
+type SheetMeta struct {
+	SheetID     string
+	Title       string
+	Index       int
+	RowCount    int
+	ColumnCount int
+}
+
+// SheetValues 读取到的单元格数据。
+type SheetValues struct {
+	SpreadsheetToken string
+	SheetID          string
+	Title            string
+	Range            string
+	Values           [][]interface{}
+}
+
+// ListSheets 列出电子表格下全部工作表。
+func (c *Client) ListSheets(ctx context.Context, spreadsheetToken string) ([]*SheetMeta, error) {
+	req := larksheets.NewQuerySpreadsheetSheetReqBuilder().
+		SpreadsheetToken(spreadsheetToken).
+		Build()
+	resp, err := c.Lark.Sheets.V3.SpreadsheetSheet.Query(ctx, req)
+	if err != nil {
+		return nil, fmt.Errorf("查询工作表列表失败: %w", err)
+	}
+	if !resp.Success() {
+		return nil, fmt.Errorf("查询工作表列表失败: code=%d msg=%s log_id=%s detail=%s",
+			resp.Code, resp.Msg, resp.RequestId(), larkcore.Prettify(resp.CodeError))
+	}
+	if resp.Data == nil {
+		return nil, nil
+	}
+	out := make([]*SheetMeta, 0, len(resp.Data.Sheets))
+	for i, s := range resp.Data.Sheets {
+		meta := &SheetMeta{Index: i}
+		if s.SheetId != nil {
+			meta.SheetID = *s.SheetId
+		}
+		if s.Title != nil {
+			meta.Title = *s.Title
+		}
+		if s.GridProperties != nil {
+			if s.GridProperties.RowCount != nil {
+				meta.RowCount = *s.GridProperties.RowCount
+			}
+			if s.GridProperties.ColumnCount != nil {
+				meta.ColumnCount = *s.GridProperties.ColumnCount
+			}
+		}
+		out = append(out, meta)
+	}
+	return out, nil
+}
+
+// FindSheet 按 sheetID 或标题查找工作表；优先 sheetID，其次精确匹配标题。
+func (c *Client) FindSheet(ctx context.Context, spreadsheetToken, sheetID, title string) (*SheetMeta, error) {
+	sheets, err := c.ListSheets(ctx, spreadsheetToken)
+	if err != nil {
+		return nil, err
+	}
+	if sheetID != "" {
+		for _, s := range sheets {
+			if s.SheetID == sheetID {
+				return s, nil
+			}
+		}
+		return nil, fmt.Errorf("未找到 sheet_id=%s 的工作表", sheetID)
+	}
+	title = strings.TrimSpace(title)
+	if title != "" {
+		for _, s := range sheets {
+			if s.Title == title {
+				return s, nil
+			}
+		}
+		return nil, fmt.Errorf("未找到标题为 %q 的工作表", title)
+	}
+	return nil, fmt.Errorf("请提供 sheetID 或 sheet 标题")
+}
+
+// ReadSheetValues 读取指定工作表范围。
+// cellRange 为空时按工作表实际行列读取整表（飞书 open range 列数上限约 100）。
+func (c *Client) ReadSheetValues(ctx context.Context, spreadsheetToken, sheetID, cellRange string) (*SheetValues, error) {
+	if spreadsheetToken == "" || sheetID == "" {
+		return nil, fmt.Errorf("spreadsheetToken / sheetID 不能为空")
+	}
+	if cellRange == "" {
+		var err error
+		cellRange, err = c.defaultSheetRange(ctx, spreadsheetToken, sheetID)
+		if err != nil {
+			return nil, err
+		}
+	}
+	rangeExpr := cellRange
+	if !strings.Contains(cellRange, "!") {
+		rangeExpr = sheetID + "!" + cellRange
+	}
+
+	apiPath := fmt.Sprintf("/open-apis/sheets/v2/spreadsheets/%s/values/%s", spreadsheetToken, rangeExpr)
+	apiResp, err := c.Lark.Do(ctx, &larkcore.ApiReq{
+		HttpMethod:                http.MethodGet,
+		ApiPath:                   apiPath,
+		QueryParams:               larkcore.QueryParams{"valueRenderOption": []string{"ToString"}},
+		SupportedAccessTokenTypes: []larkcore.AccessTokenType{larkcore.AccessTokenTypeTenant},
+	})
+	if err != nil {
+		return nil, fmt.Errorf("读取表格失败: %w", err)
+	}
+
+	var raw struct {
+		Code int    `json:"code"`
+		Msg  string `json:"msg"`
+		Data struct {
+			Revision         int    `json:"revision"`
+			SpreadsheetToken string `json:"spreadsheetToken"`
+			ValueRange       struct {
+				Range  string          `json:"range"`
+				Values [][]interface{} `json:"values"`
+			} `json:"valueRange"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(apiResp.RawBody, &raw); err != nil {
+		return nil, fmt.Errorf("解析表格响应失败: %w", err)
+	}
+	if raw.Code != 0 {
+		return nil, fmt.Errorf("读取表格失败: code=%d msg=%s body=%s", raw.Code, raw.Msg, string(apiResp.RawBody))
+	}
+
+	return &SheetValues{
+		SpreadsheetToken: spreadsheetToken,
+		SheetID:          sheetID,
+		Range:            raw.Data.ValueRange.Range,
+		Values:           raw.Data.ValueRange.Values,
+	}, nil
+}
+
+// ReadSpreadsheetURL 解析飞书链接并读取指定 sheet（按 URL 中 sheet id 或标题）。
+func (c *Client) ReadSpreadsheetURL(ctx context.Context, rawURL, sheetTitle string) (*SheetValues, error) {
+	ref, err := ParseSpreadsheetURL(rawURL)
+	if err != nil {
+		return nil, err
+	}
+	token, err := c.ResolveSpreadsheetToken(ctx, ref)
+	if err != nil {
+		return nil, err
+	}
+	meta, err := c.FindSheet(ctx, token, ref.SheetID, sheetTitle)
+	if err != nil {
+		return nil, err
+	}
+	values, err := c.ReadSheetValues(ctx, token, meta.SheetID, "")
+	if err != nil {
+		return nil, err
+	}
+	values.Title = meta.Title
+	return values, nil
+}
+
+func (c *Client) defaultSheetRange(ctx context.Context, spreadsheetToken, sheetID string) (string, error) {
+	sheets, err := c.ListSheets(ctx, spreadsheetToken)
+	if err != nil {
+		return "", err
+	}
+	rows, cols := 200, 26
+	for _, s := range sheets {
+		if s.SheetID != sheetID {
+			continue
+		}
+		if s.RowCount > 0 {
+			rows = s.RowCount
+		}
+		if s.ColumnCount > 0 {
+			cols = s.ColumnCount
+		}
+		break
+	}
+	// 飞书 values 接口对形如 A:Z 的开放范围最多约 100 列
+	if cols > 100 {
+		cols = 100
+	}
+	if rows < 1 {
+		rows = 1
+	}
+	return fmt.Sprintf("A1:%s%d", colIndexToName(cols), rows), nil
+}
+
+// colIndexToName 将 1-based 列序号转为 Excel 列名（1->A, 26->Z, 27->AA）。
+func colIndexToName(index int) string {
+	if index < 1 {
+		index = 1
+	}
+	name := make([]byte, 0, 4)
+	for index > 0 {
+		index--
+		name = append([]byte{byte('A' + index%26)}, name...)
+		index /= 26
+	}
+	return string(name)
+}
+
+// CheckReadPermission 验证应用是否具备读取目标飞书表格的权限。
+// 成功时返回可访问的工作表元信息；失败时返回飞书错误详情，便于排查权限/授权。
+func (c *Client) CheckReadPermission(ctx context.Context, rawURL, sheetTitle string) (*SheetMeta, string, error) {
+	ref, err := ParseSpreadsheetURL(rawURL)
+	if err != nil {
+		return nil, "", err
+	}
+	token, err := c.ResolveSpreadsheetToken(ctx, ref)
+	if err != nil {
+		return nil, "", err
+	}
+	meta, err := c.FindSheet(ctx, token, ref.SheetID, sheetTitle)
+	if err != nil {
+		return nil, token, err
+	}
+	// 再读一小段确认内容读权限（不仅是元信息）
+	if _, err := c.ReadSheetValues(ctx, token, meta.SheetID, "A1:A1"); err != nil {
+		return meta, token, fmt.Errorf("可列举工作表，但读取内容失败（请检查 sheets 读权限及文档授权）: %w", err)
+	}
+	return meta, token, nil
+}
diff --git a/common_util/feishu_util/url.go b/common_util/feishu_util/url.go
new file mode 100644
index 0000000..d82ecfd
--- /dev/null
+++ b/common_util/feishu_util/url.go
@@ -0,0 +1,51 @@
+package feishu_util
+
+import (
+	"fmt"
+	"net/url"
+	"strings"
+)
+
+// SpreadsheetRef 解析后的飞书表格定位信息。
+type SpreadsheetRef struct {
+	// RawURL 原始链接
+	RawURL string
+	// WikiToken 知识库节点 token（URL 中 /wiki/ 后一段）；非 wiki 链接为空
+	WikiToken string
+	// SpreadsheetToken 电子表格 token；wiki 链接需再调接口换取 obj_token
+	SpreadsheetToken string
+	// SheetID URL 查询参数 sheet= 的值（可选）
+	SheetID string
+}
+
+// ParseSpreadsheetURL 解析飞书表格/知识库链接。
+// 支持：
+//   - https://xxx.feishu.cn/wiki/{wikiToken}?sheet={sheetId}
+//   - https://xxx.feishu.cn/sheets/{spreadsheetToken}?sheet={sheetId}
+func ParseSpreadsheetURL(raw string) (*SpreadsheetRef, error) {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return nil, fmt.Errorf("表格 URL 为空")
+	}
+	u, err := url.Parse(raw)
+	if err != nil {
+		return nil, fmt.Errorf("解析 URL 失败: %w", err)
+	}
+
+	ref := &SpreadsheetRef{RawURL: raw, SheetID: u.Query().Get("sheet")}
+	path := strings.Trim(u.Path, "/")
+	parts := strings.Split(path, "/")
+	if len(parts) < 2 {
+		return nil, fmt.Errorf("无法从 URL 解析 token: %s", raw)
+	}
+
+	switch parts[0] {
+	case "wiki":
+		ref.WikiToken = parts[1]
+	case "sheets":
+		ref.SpreadsheetToken = parts[1]
+	default:
+		return nil, fmt.Errorf("不支持的飞书链接类型 %q，仅支持 /wiki/ 或 /sheets/", parts[0])
+	}
+	return ref, nil
+}
diff --git a/common_util/feishu_util/url_test.go b/common_util/feishu_util/url_test.go
new file mode 100644
index 0000000..23c0c52
--- /dev/null
+++ b/common_util/feishu_util/url_test.go
@@ -0,0 +1,25 @@
+package feishu_util
+
+import "testing"
+
+func TestParseSpreadsheetURL_Wiki(t *testing.T) {
+	ref, err := ParseSpreadsheetURL(DefaultOptimizationIssueSheetURL)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if ref.WikiToken != "IbCUwR4bvil4E1k9k2LcCiwSn8t" {
+		t.Fatalf("wiki token = %s", ref.WikiToken)
+	}
+	if ref.SheetID != "QyDfOp" {
+		t.Fatalf("sheet id = %s", ref.SheetID)
+	}
+}
+
+func TestColIndexToName(t *testing.T) {
+	cases := map[int]string{1: "A", 26: "Z", 27: "AA", 52: "AZ", 53: "BA"}
+	for n, want := range cases {
+		if got := colIndexToName(n); got != want {
+			t.Fatalf("colIndexToName(%d)=%s want %s", n, got, want)
+		}
+	}
+}
diff --git a/common_util/feishu_util/wiki.go b/common_util/feishu_util/wiki.go
new file mode 100644
index 0000000..689cd61
--- /dev/null
+++ b/common_util/feishu_util/wiki.go
@@ -0,0 +1,83 @@
+package feishu_util
+
+import (
+	"context"
+	"fmt"
+
+	larkcore "github.com/larksuite/oapi-sdk-go/v3/core"
+	larkwiki "github.com/larksuite/oapi-sdk-go/v3/service/wiki/v2"
+)
+
+// WikiNode 知识库节点解析结果。
+type WikiNode struct {
+	SpaceID   string
+	NodeToken string
+	ObjToken  string
+	ObjType   string
+	Title     string
+}
+
+// ResolveWikiNode 通过知识库 node_token 获取实际云文档 obj_token。
+// 知识库 URL 中的 token 不是 spreadsheet_token，读写表格内容前必须先转换。
+func (c *Client) ResolveWikiNode(ctx context.Context, wikiToken string) (*WikiNode, error) {
+	if wikiToken == "" {
+		return nil, fmt.Errorf("wikiToken 为空")
+	}
+	req := larkwiki.NewGetNodeSpaceReqBuilder().
+		Token(wikiToken).
+		Build()
+	resp, err := c.Lark.Wiki.V2.Space.GetNode(ctx, req)
+	if err != nil {
+		return nil, fmt.Errorf("获取知识库节点失败: %w", err)
+	}
+	if !resp.Success() {
+		return nil, fmt.Errorf("获取知识库节点失败: code=%d msg=%s log_id=%s detail=%s",
+			resp.Code, resp.Msg, resp.RequestId(), larkcore.Prettify(resp.CodeError))
+	}
+	if resp.Data == nil || resp.Data.Node == nil {
+		return nil, fmt.Errorf("获取知识库节点返回空数据")
+	}
+	n := resp.Data.Node
+	node := &WikiNode{}
+	if n.SpaceId != nil {
+		node.SpaceID = *n.SpaceId
+	}
+	if n.NodeToken != nil {
+		node.NodeToken = *n.NodeToken
+	}
+	if n.ObjToken != nil {
+		node.ObjToken = *n.ObjToken
+	}
+	if n.ObjType != nil {
+		node.ObjType = *n.ObjType
+	}
+	if n.Title != nil {
+		node.Title = *n.Title
+	}
+	if node.ObjToken == "" {
+		return nil, fmt.Errorf("知识库节点未返回 obj_token")
+	}
+	return node, nil
+}
+
+// ResolveSpreadsheetToken 将 SpreadsheetRef 解析为可调用 Sheets API 的 spreadsheet_token。
+func (c *Client) ResolveSpreadsheetToken(ctx context.Context, ref *SpreadsheetRef) (string, error) {
+	if ref == nil {
+		return "", fmt.Errorf("SpreadsheetRef 为空")
+	}
+	if ref.SpreadsheetToken != "" {
+		return ref.SpreadsheetToken, nil
+	}
+	if ref.WikiToken == "" {
+		return "", fmt.Errorf("既无 SpreadsheetToken 也无 WikiToken")
+	}
+	node, err := c.ResolveWikiNode(ctx, ref.WikiToken)
+	if err != nil {
+		return "", err
+	}
+	if node.ObjType != "" && node.ObjType != "sheet" {
+		return "", fmt.Errorf("知识库节点类型为 %s，不是电子表格(sheet)", node.ObjType)
+	}
+	ref.SpreadsheetToken = node.ObjToken
+	return node.ObjToken, nil
+}
diff --git a/config/config.example.yaml b/config/config.example.yaml
new file mode 100644
index 0000000..3d9fa02
--- /dev/null
+++ b/config/config.example.yaml
@@ -0,0 +1,8 @@
+# 配置模板（可入库）。复制为 config/config.dev.yaml 后填入真实值。
+# 本地：config/config.dev.yaml（已 gitignore）
+feishu:
+  app_id: ""      # 飞书应用 App ID，如 cli_xxx
+  app_secret: ""  # 飞书应用 App Secret，勿提交仓库
+hisense_geo:
+  base_url: "https://geo-api.hisense.com"
+  token: ""       # 生产环境 x-token，勿提交仓库
diff --git a/config/load.go b/config/load.go
new file mode 100644
index 0000000..89bef1c
--- /dev/null
+++ b/config/load.go
@@ -0,0 +1,91 @@
+package config
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+
+	"gopkg.in/yaml.v3"
+)
+
+// Config 项目根目录 config/config.{env}.yaml（敏感项已 gitignore）。
+type Config struct {
+	Feishu     FeishuConfig     `yaml:"feishu"`
+	HisenseGeo HisenseGeoConfig `yaml:"hisense_geo"`
+}
+
+// FeishuConfig 飞书应用凭证。
+type FeishuConfig struct {
+	AppID     string `yaml:"app_id"`
+	AppSecret string `yaml:"app_secret"`
+}
+
+// HisenseGeoConfig 海信 GEO 生产环境接口配置。
+type HisenseGeoConfig struct {
+	BaseURL string `yaml:"base_url"` // 默认 https://geo-api.hisense.com
+	Token   string `yaml:"token"`    // x-token，勿提交仓库
+}
+
+// Load 按 APP_ENV 加载 config 目录下配置，默认 dev。
+// 查找顺序：
+//  1. CONFIG_PATH / FEISHU_CONFIG_PATH
+//  2. 当前工作目录 ./config/config.{env}.yaml
+//  3. 向上查找含 config/ 目录的项目根下的 config/config.{env}.yaml
+func Load() (*Config, error) {
+	env := os.Getenv("APP_ENV")
+	if env == "" {
+		env = "dev"
+	}
+	name := "config." + env + ".yaml"
+	rel := filepath.Join("config", name)
+
+	candidates := []string{}
+	if p := os.Getenv("CONFIG_PATH"); p != "" {
+		candidates = append(candidates, p)
+	}
+	if p := os.Getenv("FEISHU_CONFIG_PATH"); p != "" {
+		candidates = append(candidates, p)
+	}
+	candidates = append(candidates, rel)
+	if root := FindProjectRoot(); root != "" {
+		candidates = append(candidates, filepath.Join(root, rel))
+	}
+
+	var lastErr error
+	for _, path := range candidates {
+		data, err := os.ReadFile(path)
+		if err != nil {
+			lastErr = err
+			continue
+		}
+		var cfg Config
+		if err := yaml.Unmarshal(data, &cfg); err != nil {
+			return nil, fmt.Errorf("解析配置失败 (%s): %w", path, err)
+		}
+		if cfg.HisenseGeo.BaseURL == "" {
+			cfg.HisenseGeo.BaseURL = "https://geo-api.hisense.com"
+		}
+		return &cfg, nil
+	}
+	return nil, fmt.Errorf("未找到配置文件 %s（放在 config/ 目录，或设置 CONFIG_PATH）: %w", rel, lastErr)
+}
+
+// FindProjectRoot 向上查找包含 config/ 目录的项目根。
+func FindProjectRoot() string {
+	wd, err := os.Getwd()
+	if err != nil {
+		return ""
+	}
+	dir := wd
+	for {
+		cfgDir := filepath.Join(dir, "config")
+		if st, err := os.Stat(cfgDir); err == nil && st.IsDir() {
+			return dir
+		}
+		parent := filepath.Dir(dir)
+		if parent == dir {
+			return ""
+		}
+		dir = parent
+	}
+}
diff --git a/go.mod b/go.mod
new file mode 100644
index 0000000..2711790
--- /dev/null
+++ b/go.mod
@@ -0,0 +1,8 @@
+module ai_script
+
+go 1.24.8
+
+require (
+	github.com/larksuite/oapi-sdk-go/v3 v3.12.0
+	gopkg.in/yaml.v3 v3.0.1
+)
diff --git a/go.sum b/go.sum
new file mode 100644
index 0000000..5ef658d
--- /dev/null
+++ b/go.sum
@@ -0,0 +1,6 @@
+github.com/larksuite/oapi-sdk-go/v3 v3.12.0 h1:H8NP6YIgfEX0RBhKse25npdeZoiaDv8mrw9nCnCVFRc=
+github.com/larksuite/oapi-sdk-go/v3 v3.12.0/go.mod h1:F1nwLfYBSKvD8mS/OFw90Gr+V4RIugPBR44m/ThA5jg=
+gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
+gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
+gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
+gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
diff --git a/haoxian_hisense/agentic_project/check_agentic_project_name/README.md b/haoxian_hisense/agentic_project/check_agentic_project_name/README.md
new file mode 100644
index 0000000..5b755a4
--- /dev/null
+++ b/haoxian_hisense/agentic_project/check_agentic_project_name/README.md
@@ -0,0 +1,61 @@
+# check_agentic_project_name
+
+刷数脚本：检查项目名称、产业/品类、产品名称、漏出词等数据一致性。
+
+## 数据源
+
+1. 飞书表格「9月各产业优化问题汇总表」  
+   `https://mi8elatzm1x.feishu.cn/wiki/IbCUwR4bvil4E1k9k2LcCiwSn8t?sheet=QyDfOp`
+2. 生产接口  
+   `GET https://geo-api.hisense.com/api/agentic-project/project?page=1&page_size=999&listed=all`  
+   Header: `x-token`
+
+## 匹配主键（跟接口保持一致）
+
+| 飞书 | 接口 |
+|------|------|
+| 优化问题 | `project_name` |
+| 产业（= 品牌 + 品类） | `brand` + `category`（直接拼接） |
+
+接口侧额外过滤（必须同时满足）：
+
+- `listed = 1`
+- `status = 'running'`
+- `end_month = '2026-09'`
+
+匹配规则：
+
+1. 先按上述过滤筛接口项目
+2. 常规：`产业 == brand+category`（如 `海信中央空调`）
+3. 简写兼容：`产业 == brand` 且 `品类 == category`（如 ASKO / 古洛尼）
+4. 同名+同产业仍多条时：优先飞书【产品名称】/【特征露出词】与接口一致
+
+`品类` ↔ `category` 仅作校验罗列，不阻断主键匹配。
+
+## 比对项
+
+1. 飞书【优化问题】在接口中找不到
+2. 飞书【产业】无法按规则匹配接口【brand+category】（匹配失败，打印完整对照）
+3. 飞书【产业】与接口【brand+category】字面不一致（主键已匹配，如简写 `ASKO` vs `ASKO全品类`）
+4. 飞书【品类】与接口【category】不一致（主键已匹配）
+5. 飞书【产品名称】与接口 `product_name` 不一致 → 输出更新 SQL
+6. 飞书【特征露出词】与接口 `exposure_word` 不一致 → 输出更新 SQL
+
+## 运行
+
+```bash
+go run ./haoxian_hisense/agentic_project/check_agentic_project_name
+```
+
+## 配置
+
+`config/config.dev.yaml`（已 gitignore）：
+
+```yaml
+feishu:
+  app_id: ""
+  app_secret: ""
+hisense_geo:
+  base_url: "https://geo-api.hisense.com"
+  token: ""   # x-token，勿提交仓库
+```
diff --git a/haoxian_hisense/agentic_project/check_agentic_project_name/main.go b/haoxian_hisense/agentic_project/check_agentic_project_name/main.go
new file mode 100644
index 0000000..655c79a
--- /dev/null
+++ b/haoxian_hisense/agentic_project/check_agentic_project_name/main.go
@@ -0,0 +1,585 @@
+package main
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"log"
+	"net/http"
+	"sort"
+	"strings"
+	"time"
+	"unicode"
+
+	"ai_script/common_util/feishu_util"
+	"ai_script/config"
+)
+
+func main() {
+	// 加载项目根目录 config/config.{env}.yaml（含飞书凭证与 GEO Token）
+	cfg, err := config.Load()
+	if err != nil {
+		log.Fatalf("加载配置失败: %v", err)
+	}
+
+	ctx := context.Background()
+
+	// ---------------------------------------------------------------------------
+	// 1、拉取飞书表格中「9月各产业优化问题汇总表」sheet 数据
+	//    https://mi8elatzm1x.feishu.cn/wiki/IbCUwR4bvil4E1k9k2LcCiwSn8t?sheet=QyDfOp
+	// ---------------------------------------------------------------------------
+	feishuRows, err := fetchFeishuOptimizationRows(ctx, cfg)
+	if err != nil {
+		log.Fatalf("拉取飞书表格失败: %v", err)
+	}
+	log.Printf("[1] 飞书表格有效行数: %d", len(feishuRows))
+
+	// ---------------------------------------------------------------------------
+	// 2、拉取生产环境项目表数据
+	//    GET {base_url}/api/agentic-project/project?page=1&page_size=999&listed=all
+	//    Header: x-token
+	// ---------------------------------------------------------------------------
+	projects, err := fetchProdProjects(ctx, cfg)
+	if err != nil {
+		log.Fatalf("拉取生产环境项目失败: %v", err)
+	}
+	log.Printf("[2] 生产环境项目数: %d", len(projects))
+
+	// 按 project_name 分组，匹配时用 产业 ↔ brand+category（与接口保持一致）
+	projectsByName := indexProjectsByName(projects)
+
+	// ---------------------------------------------------------------------------
+	// 3、比对差异点
+	//    匹配主键：飞书【优化问题+产业】 ↔ 接口【project_name+(brand+category)】
+	//    品类 ↔ category 仅作校验罗列，不阻断主键匹配
+	// ---------------------------------------------------------------------------
+	compareAndLogDiffs(feishuRows, projectsByName)
+}
+
+// feishuRow 飞书「9月各产业优化问题汇总表」解析后的一行。
+type feishuRow struct {
+	RowNum        int    // 表格中的物理行号（1-based，含表头）
+	Industry      string // 产业 = 品牌+品类 → 对齐接口 brand+category
+	Category      string // 品类 → 对齐接口 category
+	ProductName   string // 产品名称
+	ExposureWord  string // 特征露出词
+	OptimQuestion string // 优化问题 → 对齐接口 project_name
+}
+
+// prodProject 生产环境 agentic_project 列表项。
+type prodProject struct {
+	ID           string
+	CompanyID    string // company_id
+	ProjectName  string // project_name
+	Brand        string // brand
+	Category     string // category
+	ProductName  string // product_name（数组拼接）
+	ExposureWord string // exposure_word
+	Status       string // status: running / completed / ...
+	Listed       int    // listed
+	EndMonth     string // end_month，如 2026-09
+}
+
+// 匹配时额外过滤条件（跟业务月度表对齐）
+const (
+	requiredListed   = 1
+	requiredStatus   = "running"
+	requiredEndMonth = "2026-09"
+)
+
+// eligibleProject 是否满足 listed/status/end_month 过滤。
+func eligibleProject(p prodProject) bool {
+	return p.Listed == requiredListed &&
+		strings.EqualFold(p.Status, requiredStatus) &&
+		p.EndMonth == requiredEndMonth
+}
+
+// apiIndustry 接口侧「产业」等价字段：brand + category（无连接符）。
+func (p prodProject) apiIndustry() string {
+	return p.Brand + p.Category
+}
+
+// industryMatched 飞书产业与接口是否对齐（跟接口保持一致）：
+// 1) 产业 == brand+category（常规：海信中央空调）
+// 2) 产业 == brand 且 品类 == category（简写：ASKO / 古洛尼）
+func industryMatched(row feishuRow, api prodProject) bool {
+	if foldText(row.Industry) == foldText(api.apiIndustry()) {
+		return true
+	}
+	if foldText(row.Industry) == foldText(api.Brand) && foldText(row.Category) == foldText(api.Category) {
+		return true
+	}
+	return false
+}
+
+// findMatchedProject 按优化问题找同名项目，再按产业规则对齐接口。
+// 仅考虑 listed=1 AND status=running AND end_month=2026-09；
+// 同名+同产业仍多条时，优先产品名称/露出词一致。
+func findMatchedProject(row feishuRow, projectsByName map[string][]prodProject) (prodProject, bool) {
+	cands := projectsByName[foldText(row.OptimQuestion)]
+	matched := make([]prodProject, 0, 2)
+	for _, api := range cands {
+		if !eligibleProject(api) {
+			continue
+		}
+		if industryMatched(row, api) {
+			matched = append(matched, api)
+		}
+	}
+	if len(matched) == 0 {
+		return prodProject{}, false
+	}
+	if len(matched) == 1 {
+		return matched[0], true
+	}
+
+	best := matched[0]
+	bestScore := matchScore(row, best)
+	for _, api := range matched[1:] {
+		if s := matchScore(row, api); s > bestScore {
+			best = api
+			bestScore = s
+		}
+	}
+	ids := make([]string, 0, len(matched))
+	for _, m := range matched {
+		ids = append(ids, m.ID)
+	}
+	log.Printf("[匹配消歧] 优化问题=%q 产业=%q 候选id=%v，选用 id=%s score=%d（飞书产品=%q 露出词=%q）",
+		row.OptimQuestion, row.Industry, ids, best.ID, bestScore, row.ProductName, row.ExposureWord)
+	return best, true
+}
+
+// matchScore 同名同产业多候选时的打分（越高越优先）。
+func matchScore(row feishuRow, api prodProject) int {
+	score := 0
+	if sameDelimitedText(row.ProductName, api.ProductName) {
+		score += 100
+	}
+	if sameDelimitedText(row.ExposureWord, api.ExposureWord) {
+		score += 50
+	}
+	return score
+}
+
+// fetchFeishuOptimizationRows 步骤1：读取飞书 sheet 并解析为结构化行。
+func fetchFeishuOptimizationRows(ctx context.Context, cfg *config.Config) ([]feishuRow, error) {
+	client, err := feishu_util.NewClientFromConfig(cfg)
+	if err != nil {
+		return nil, err
+	}
+	values, err := client.ReadSpreadsheetURL(ctx,
+		feishu_util.DefaultOptimizationIssueSheetURL,
+		feishu_util.DefaultOptimizationIssueSheetTitle,
+	)
+	if err != nil {
+		return nil, err
+	}
+	return parseFeishuSheet(values.Values)
+}
+
+// parseFeishuSheet 根据表头定位列并解析数据行。
+func parseFeishuSheet(values [][]interface{}) ([]feishuRow, error) {
+	headerIdx := -1
+	colIndustry, colCategory := -1, -1
+	colProduct, colExposure, colQuestion := -1, -1, -1
+
+	for i, row := range values {
+		cells := rowToStrings(row)
+		for j, cell := range cells {
+			switch strings.TrimSpace(cell) {
+			case "产业":
+				colIndustry = j
+			case "品类":
+				colCategory = j
+			case "产品名称":
+				colProduct = j
+			case "特征露出词":
+				colExposure = j
+			case "优化问题":
+				colQuestion = j
+			}
+		}
+		if colIndustry >= 0 && colCategory >= 0 && colProduct >= 0 && colExposure >= 0 && colQuestion >= 0 {
+			headerIdx = i
+			break
+		}
+		colIndustry, colCategory, colProduct, colExposure, colQuestion = -1, -1, -1, -1, -1
+	}
+	if headerIdx < 0 {
+		return nil, fmt.Errorf("未在飞书表格中找到表头列：产业 / 品类 / 产品名称 / 特征露出词 / 优化问题")
+	}
+
+	out := make([]feishuRow, 0)
+	for i := headerIdx + 1; i < len(values); i++ {
+		cells := rowToStrings(values[i])
+		question := cellAt(cells, colQuestion)
+		if question == "" {
+			// 空行 / 产业分隔空行跳过
+			continue
+		}
+		out = append(out, feishuRow{
+			RowNum:        i + 1,
+			Industry:      cellAt(cells, colIndustry),
+			Category:      cellAt(cells, colCategory),
+			ProductName:   cellAt(cells, colProduct),
+			ExposureWord:  cellAt(cells, colExposure),
+			OptimQuestion: question,
+		})
+	}
+	return out, nil
+}
+
+// fetchProdProjects 步骤2：拉取 agentic-project 列表（含 brand/category/exposure_word）。
+func fetchProdProjects(ctx context.Context, cfg *config.Config) ([]prodProject, error) {
+	if cfg.HisenseGeo.Token == "" {
+		return nil, fmt.Errorf("配置缺少 hisense_geo.token")
+	}
+	base := strings.TrimRight(cfg.HisenseGeo.BaseURL, "/")
+	url := fmt.Sprintf("%s/api/agentic-project/project?page=1&page_size=999&listed=all", base)
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return nil, err
+	}
+	req.Header.Set("Accept", "application/json")
+	req.Header.Set("Content-Type", "application/json; charset=utf-8")
+	req.Header.Set("Origin", "https://geo.hisense.com")
+	req.Header.Set("Referer", "https://geo.hisense.com/")
+	req.Header.Set("x-token", cfg.HisenseGeo.Token)
+
+	httpClient := &http.Client{Timeout: 60 * time.Second}
+	resp, err := httpClient.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+	body, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return nil, err
+	}
+	if resp.StatusCode != http.StatusOK {
+		return nil, fmt.Errorf("agentic-project 接口 HTTP %d: %s", resp.StatusCode, truncate(string(body), 300))
+	}
+
+	var raw struct {
+		Code    int    `json:"code"`
+		Message string `json:"message"`
+		Data    struct {
+			Total int `json:"total"`
+			List  []struct {
+				ID           json.Number `json:"id"`
+				CompanyID    json.Number `json:"company_id"`
+				ProjectName  string      `json:"project_name"`
+				Brand        string      `json:"brand"`
+				Category     string      `json:"category"`
+				ProductName  []string    `json:"product_name"`
+				ExposureWord string      `json:"exposure_word"`
+				Status       string      `json:"status"`
+				Listed       int         `json:"listed"`
+				EndMonth     string      `json:"end_month"`
+			} `json:"list"`
+		} `json:"data"`
+	}
+	if err := json.Unmarshal(body, &raw); err != nil {
+		return nil, fmt.Errorf("解析 agentic-project 响应失败: %w", err)
+	}
+	if raw.Code != 0 {
+		return nil, fmt.Errorf("agentic-project 接口业务错误 code=%d message=%s", raw.Code, raw.Message)
+	}
+
+	out := make([]prodProject, 0, len(raw.Data.List))
+	nonCompany13 := 0
+	eligible := 0
+	for _, w := range raw.Data.List {
+		companyID := w.CompanyID.String()
+		if companyID == "" {
+			companyID = "13"
+		}
+		if companyID != "13" {
+			nonCompany13++
+		}
+		p := prodProject{
+			ID:           w.ID.String(),
+			CompanyID:    companyID,
+			ProjectName:  strings.TrimSpace(w.ProjectName),
+			Brand:        strings.TrimSpace(w.Brand),
+			Category:     strings.TrimSpace(w.Category),
+			ProductName:  joinProductNames(w.ProductName),
+			ExposureWord: strings.TrimSpace(w.ExposureWord),
+			Status:       strings.TrimSpace(w.Status),
+			Listed:       w.Listed,
+			EndMonth:     strings.TrimSpace(w.EndMonth),
+		}
+		if eligibleProject(p) {
+			eligible++
+		}
+		out = append(out, p)
+	}
+	log.Printf("[2] 满足 listed=%d AND status=%s AND end_month=%s 的项目: %d / %d",
+		requiredListed, requiredStatus, requiredEndMonth, eligible, len(out))
+	if nonCompany13 > 0 {
+		log.Printf("[2] 警告: 存在 company_id != 13 的项目 %d 条", nonCompany13)
+	} else {
+		log.Printf("[2] 全部项目 company_id = 13")
+	}
+	if raw.Data.Total > len(out) {
+		log.Printf("[2] 警告: 接口 total=%d，本页仅返回 %d 条，可能未拉全", raw.Data.Total, len(out))
+	}
+	return out, nil
+}
+
+func indexProjectsByName(projects []prodProject) map[string][]prodProject {
+	m := make(map[string][]prodProject)
+	for _, p := range projects {
+		key := foldText(p.ProjectName)
+		if key == "" {
+			continue
+		}
+		m[key] = append(m[key], p)
+	}
+	return m
+}
+
+type industryCategoryMismatch struct {
+	row     feishuRow
+	apiList []prodProject // 同名 project_name 下的候选项目
+}
+
+type matchedPair struct {
+	row feishuRow
+	api prodProject
+}
+
+// compareAndLogDiffs 按接口语义匹配，并输出差异与更新 SQL。
+func compareAndLogDiffs(rows []feishuRow, projectsByName map[string][]prodProject) {
+	var (
+		missing       []feishuRow
+		fieldMismatch []industryCategoryMismatch
+		matched       []matchedPair
+		categoryOnly  []matchedPair // 主键已匹配，但品类 ≠ category
+	)
+
+	var industryLiteralDiff []matchedPair // 主键已匹配，但飞书产业 ≠ brand+category（多为简写兼容）
+
+	for _, row := range rows {
+		if api, ok := findMatchedProject(row, projectsByName); ok {
+			matched = append(matched, matchedPair{row: row, api: api})
+			if foldText(row.Industry) != foldText(api.apiIndustry()) {
+				industryLiteralDiff = append(industryLiteralDiff, matchedPair{row: row, api: api})
+			}
+			if foldText(row.Category) != foldText(api.Category) {
+				categoryOnly = append(categoryOnly, matchedPair{row: row, api: api})
+			}
+			continue
+		}
+		cands := projectsByName[foldText(row.OptimQuestion)]
+		if len(cands) == 0 {
+			missing = append(missing, row)
+			continue
+		}
+		fieldMismatch = append(fieldMismatch, industryCategoryMismatch{row: row, apiList: cands})
+	}
+
+	// ---------- 3.0 主键匹配概况 ----------
+	log.Printf("========== 主键匹配概况 ==========")
+	log.Printf("匹配主键: 飞书【优化问题+产业】 ↔ 接口【project_name+(brand+category)】")
+	log.Printf("接口过滤: listed=%d AND status=%s AND end_month=%s", requiredListed, requiredStatus, requiredEndMonth)
+	log.Printf("说明: 飞书【产业】应对齐 brand+category；简写兼容(产业=brand)时会列入「产业字面不一致」")
+	log.Printf("飞书有效行=%d, 主键匹配=%d, 产业匹配失败=%d, 产业字面不一致(已匹配)=%d, 优化问题缺失=%d, 品类不一致(已匹配)=%d",
+		len(rows), len(matched), len(fieldMismatch), len(industryLiteralDiff), len(missing), len(categoryOnly))
+	if len(missing) == 0 && len(fieldMismatch) == 0 {
+		log.Printf("结论: 主键（优化问题+产业）全部对应上")
+	}
+
+	// ---------- 3.1 优化问题在接口中找不到 ----------
+	log.Printf("========== 3.1 飞书【优化问题】在接口中找不到（共 %d 条）==========", len(missing))
+	for _, row := range missing {
+		log.Printf("[缺失] 行=%d 优化问题=%q 产业=%q 品类=%q",
+			row.RowNum, row.OptimQuestion, row.Industry, row.Category)
+	}
+
+	// ---------- 3.1b 产业匹配失败（同名存在但无法按规则对齐） ----------
+	log.Printf("========== 3.1b 飞书【产业】无法匹配接口【brand+category】（共 %d 条）==========", len(fieldMismatch))
+	for _, d := range fieldMismatch {
+		log.Printf("[产业匹配失败] 行=%d 优化问题=%q", d.row.RowNum, d.row.OptimQuestion)
+		log.Printf("  飞书: 产业=%q 品类=%q", d.row.Industry, d.row.Category)
+		for _, api := range d.apiList {
+			log.Printf("  接口: id=%s project_name=%q brand=%q category=%q → brand+category=%q product_name=%q exposure_word=%q",
+				api.ID, api.ProjectName, api.Brand, api.Category, api.apiIndustry(), api.ProductName, api.ExposureWord)
+		}
+	}
+
+	// ---------- 3.1b2 主键已匹配，但飞书产业字面 ≠ brand+category ----------
+	log.Printf("========== 3.1b2 飞书【产业】与接口【brand+category】字面不一致（主键已匹配，共 %d 条）==========", len(industryLiteralDiff))
+	for _, d := range industryLiteralDiff {
+		log.Printf("[产业字面不一致] 行=%d id=%s 优化问题=%q", d.row.RowNum, d.api.ID, d.row.OptimQuestion)
+		log.Printf("  飞书: 产业=%q 品类=%q", d.row.Industry, d.row.Category)
+		log.Printf("  接口: brand=%q category=%q → brand+category=%q",
+			d.api.Brand, d.api.Category, d.api.apiIndustry())
+	}
+
+	// ---------- 3.1c 品类不一致（主键已匹配） ----------
+	log.Printf("========== 3.1c 飞书【品类】与接口【category】不一致（主键已匹配，共 %d 条）==========", len(categoryOnly))
+	for _, d := range categoryOnly {
+		log.Printf("[品类不一致] 行=%d 优化问题=%q id=%s", d.row.RowNum, d.row.OptimQuestion, d.api.ID)
+		log.Printf("  飞书: 产业=%q 品类=%q", d.row.Industry, d.row.Category)
+		log.Printf("  接口: brand=%q category=%q → brand+category=%q",
+			d.api.Brand, d.api.Category, d.api.apiIndustry())
+	}
+
+	// ---------- 3.2 产品名称不一致 + 更新 SQL ----------
+	var productDiff []matchedPair
+	for _, m := range matched {
+		if !sameDelimitedText(m.row.ProductName, m.api.ProductName) {
+			productDiff = append(productDiff, m)
+		}
+	}
+	log.Printf("========== 3.2 飞书【产品名称】与接口 product_name 不一致（共 %d 条）==========", len(productDiff))
+	for _, d := range productDiff {
+		log.Printf("[产品名称不一致] 行=%d id=%s company_id=%s 优化问题=%q 产业=%q 品类=%q\n  飞书=%q\n  接口=%q",
+			d.row.RowNum, d.api.ID, d.api.CompanyID, d.row.OptimQuestion, d.row.Industry, d.row.Category,
+			d.row.ProductName, d.api.ProductName)
+	}
+	log.Printf("---------- 3.2 产品名称更新 SQL（共 %d 条）----------", len(productDiff))
+	for _, d := range productDiff {
+		fmt.Println(buildUpdateSQL("product_name", d.row.ProductName, d.api.ID, d.api.CompanyID))
+	}
+
+	// ---------- 3.3 特征露出词不一致 + 更新 SQL ----------
+	var exposureDiff []matchedPair
+	for _, m := range matched {
+		if !sameDelimitedText(m.row.ExposureWord, m.api.ExposureWord) {
+			exposureDiff = append(exposureDiff, m)
+		}
+	}
+	log.Printf("========== 3.3 飞书【特征露出词】与接口 exposure_word 不一致（共 %d 条）==========", len(exposureDiff))
+	for _, d := range exposureDiff {
+		log.Printf("[特征露出词不一致] 行=%d id=%s company_id=%s 优化问题=%q 产业=%q 品类=%q\n  飞书=%q\n  接口=%q",
+			d.row.RowNum, d.api.ID, d.api.CompanyID, d.row.OptimQuestion, d.row.Industry, d.row.Category,
+			d.row.ExposureWord, d.api.ExposureWord)
+	}
+	log.Printf("---------- 3.3 特征露出词更新 SQL（共 %d 条）----------", len(exposureDiff))
+	for _, d := range exposureDiff {
+		fmt.Println(buildUpdateSQL("exposure_word", d.row.ExposureWord, d.api.ID, d.api.CompanyID))
+	}
+
+	log.Printf("比对完成: 飞书行=%d, 主键匹配=%d, 产业匹配失败=%d, 产业字面不一致=%d, 品类不一致=%d, 优化问题缺失=%d, 产品名称不一致=%d, 特征露出词不一致=%d",
+		len(rows), len(matched), len(fieldMismatch), len(industryLiteralDiff), len(categoryOnly), len(missing), len(productDiff), len(exposureDiff))
+}
+
+// buildUpdateSQL 生成以飞书值为准的回写 SQL（带 company_id 条件）。
+func buildUpdateSQL(column, value, id, companyID string) string {
+	if companyID == "" {
+		companyID = "13"
+	}
+	return fmt.Sprintf("update `agentic`.agentic_project set %s = '%s' where id = %s and company_id = %s",
+		column, escapeSQLString(value), id, companyID)
+}
+
+func escapeSQLString(s string) string {
+	return strings.ReplaceAll(s, "'", "''")
+}
+
+func joinProductNames(names []string) string {
+	parts := make([]string, 0, len(names))
+	for _, n := range names {
+		n = strings.TrimSpace(n)
+		if n != "" {
+			parts = append(parts, n)
+		}
+	}
+	return strings.Join(parts, "｜")
+}
+
+// sameDelimitedText 比较以 ｜/| 分隔的多值文本（忽略顺序与首尾空白）。
+func sameDelimitedText(a, b string) bool {
+	return delimitedSetKey(a) == delimitedSetKey(b)
+}
+
+func delimitedSetKey(s string) string {
+	s = strings.ReplaceAll(s, "|", "｜")
+	parts := strings.Split(s, "｜")
+	norm := make([]string, 0, len(parts))
+	seen := make(map[string]struct{})
+	for _, p := range parts {
+		p = normalizeText(p)
+		if p == "" {
+			continue
+		}
+		if _, ok := seen[p]; ok {
+			continue
+		}
+		seen[p] = struct{}{}
+		norm = append(norm, p)
+	}
+	sort.Strings(norm)
+	return strings.Join(norm, "｜")
+}
+
+func normalizeText(s string) string {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return ""
+	}
+	var b strings.Builder
+	b.Grow(len(s))
+	prevSpace := false
+	for _, r := range s {
+		if unicode.IsSpace(r) {
+			if !prevSpace {
+				b.WriteRune(' ')
+				prevSpace = true
+			}
+			continue
+		}
+		prevSpace = false
+		b.WriteRune(r)
+	}
+	return strings.TrimSpace(b.String())
+}
+
+// foldText 归一化后转小写，避免 Vidda/VIdda 等大小写差异影响匹配。
+func foldText(s string) string {
+	return strings.ToLower(normalizeText(s))
+}
+
+func rowToStrings(row []interface{}) []string {
+	out := make([]string, len(row))
+	for i, v := range row {
+		out[i] = stringifyCell(v)
+	}
+	return out
+}
+
+func cellAt(cells []string, idx int) string {
+	if idx < 0 || idx >= len(cells) {
+		return ""
+	}
+	return strings.TrimSpace(cells[idx])
+}
+
+func stringifyCell(v interface{}) string {
+	if v == nil {
+		return ""
+	}
+	switch t := v.(type) {
+	case string:
+		return t
+	case float64:
+		if t == float64(int64(t)) {
+			return fmt.Sprintf("%d", int64(t))
+		}
+		return fmt.Sprintf("%v", t)
+	default:
+		return strings.TrimSpace(fmt.Sprint(t))
+	}
+}
+
+func truncate(s string, n int) string {
+	if len(s) <= n {
+		return s
+	}
+	return s[:n] + "..."
+}
diff --git a/haoxian_hisense/local/月度优化问题汇总管理表 - 9月各产业优化问题汇总表.xlsx b/haoxian_hisense/local/月度优化问题汇总管理表 - 9月各产业优化问题汇总表.xlsx
new file mode 100644
index 0000000..779c573
Binary files /dev/null and b/haoxian_hisense/local/月度优化问题汇总管理表 - 9月各产业优化问题汇总表.xlsx differ