diff --git a/.changes/1009-clitrack-telemetry.md b/.changes/released/1.0.59-beta.2/1009-clitrack-telemetry.md
similarity index 100%
rename from .changes/1009-clitrack-telemetry.md
rename to .changes/released/1.0.59-beta.2/1009-clitrack-telemetry.md
diff --git a/.changes/1019-help-feedback-entry.md b/.changes/released/1.0.59-beta.2/1019-help-feedback-entry.md
similarity index 100%
rename from .changes/1019-help-feedback-entry.md
rename to .changes/released/1.0.59-beta.2/1019-help-feedback-entry.md
diff --git a/.changes/954-chat-im-id-flags.md b/.changes/released/1.0.59-beta.2/954-chat-im-id-flags.md
similarity index 100%
rename from .changes/954-chat-im-id-flags.md
rename to .changes/released/1.0.59-beta.2/954-chat-im-id-flags.md
diff --git a/.changes/chat-reference-card-verification.md b/.changes/released/1.0.59-beta.2/chat-reference-card-verification.md
similarity index 100%
rename from .changes/chat-reference-card-verification.md
rename to .changes/released/1.0.59-beta.2/chat-reference-card-verification.md
diff --git a/.changes/schema-catalog-json-projection.md b/.changes/released/1.0.59-beta.2/schema-catalog-json-projection.md
similarity index 100%
rename from .changes/schema-catalog-json-projection.md
rename to .changes/released/1.0.59-beta.2/schema-catalog-json-projection.md
diff --git a/.changes/wiki-shortcut-alignment.md b/.changes/released/1.0.59-beta.2/wiki-shortcut-alignment.md
similarity index 100%
rename from .changes/wiki-shortcut-alignment.md
rename to .changes/released/1.0.59-beta.2/wiki-shortcut-alignment.md
diff --git a/.changes/sheet-source-range-dropdowns.md b/.changes/sheet-source-range-dropdowns.md
new file mode 100644
index 00000000..91159e61
--- /dev/null
+++ b/.changes/sheet-source-range-dropdowns.md
@@ -0,0 +1,6 @@
+---
+category: Added
+---
+
+- **Sheet SourceRange dropdowns** — supports range-backed dropdowns across direct, cell, and batch write paths, with structured readback for valid and invalid references. Batch `set-dropdown` now rejects unsupported top-level `colors` / `source-colors`; Inline colors belong in `options[].color`, while SourceRange color writes remain unsupported.
+- **Sheet read completion metadata** — documents and preserves returned ranges, truncation reasons, and partial-read status for large range and CSV reads.
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index ded760c7..6dc6406c 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -529,12 +529,22 @@ jobs:
     # shards at full-suite scope; release-scripts is included because its
     # dedicated job only runs at full-suite or release-sensitive scope, and
     # dropping it here would stop testing test/scripts changes entirely.
+    # internal/app is carried by one shard per bounded partition rather than a
+    # single app shard: the partitions used to run end to end inside one job,
+    # where the Schema partition alone owned most of the wall clock. The
+    # app-<partition> names are pinned to the helper's partition set by
+    # TestCIAppRacePartitionMatrixMatchesHelper, so a partition can never lose
+    # its job silently.
     timeout-minutes: 20
     strategy:
       fail-fast: false
       matrix:
         shard:
-          - app
+          - app-schema
+          - app-a-b
+          - app-c
+          - app-d-r
+          - app-s-z-example-fuzz
           - generators
           - helpers
           - cli
@@ -576,9 +586,16 @@ jobs:
           TEST_SHARD: ${{ matrix.shard }}
         run: |
           set -euo pipefail
+          # Every app partition shard tests the same single internal/app
+          # package, so the impacted-package query uses the base shard name and
+          # the partition only selects which tests run.
+          package_shard="$TEST_SHARD"
+          case "$TEST_SHARD" in
+            app-*) package_shard=app ;;
+          esac
           package_output="$(
             ./scripts/ci/changed-test-packages.sh \
-              list-shard "$TEST_SHARD" "$TEST_BASE_REF" "$TEST_HEAD_REF"
+              list-shard "$package_shard" "$TEST_BASE_REF" "$TEST_HEAD_REF"
           )"
           if [ -z "$package_output" ]; then
             echo "No buildable Go package in shard $TEST_SHARD is affected by this revision." \
@@ -618,14 +635,19 @@ jobs:
               exit 1
             }
           done
-          if [ "$TEST_SHARD" = "app" ]; then
-            # A single long-lived app test process retains every constructed
-            # command tree in framework registries. Isolate Schema assembly and
-            # bounded name ranges so each process releases that state on exit.
-            test "${#packages[@]}" -eq 1
-            ./scripts/ci/run-app-race-tests.sh run "${packages[0]}"
-            exit 0
-          fi
+          case "$TEST_SHARD" in
+            app-*)
+              # A single long-lived app test process retains every constructed
+              # command tree in framework registries. Each partition is its own
+              # job, so that state is released when the process exits and the
+              # partitions run concurrently instead of end to end. The helper
+              # still verifies that the partition patterns cover every top-level
+              # test exactly once before running the one it was asked for.
+              test "${#packages[@]}" -eq 1
+              ./scripts/ci/run-app-race-tests.sh run "${packages[0]}" "${TEST_SHARD#app-}"
+              exit 0
+              ;;
+          esac
           if [ "$TEST_SHARD" = "release-scripts" ]; then
             # Mirror the dedicated release-contract job: these suites shell out
             # to archive tooling and are not race-instrumented there.
@@ -646,14 +668,21 @@ jobs:
     needs: lint
     if: ${{ needs.lint.outputs.changelog_only != 'true' && needs.lint.outputs.docs_only != 'true' && needs.lint.outputs.full_suite == 'true' }}
     runs-on: ubuntu-latest
-    # app runs several independently bounded processes; cli/smoke need headroom
-    # beyond go test -timeout for setup + assembly.
+    # internal/app is split across one shard per bounded partition so the
+    # partitions run concurrently and each releases its framework registries
+    # when the process exits; cli/smoke need headroom beyond go test -timeout for
+    # setup + assembly. The app-<partition> names are pinned to the helper's
+    # partition set by TestCIAppRacePartitionMatrixMatchesHelper.
     timeout-minutes: 20
     strategy:
       fail-fast: false
       matrix:
         shard:
-          - app
+          - app-schema
+          - app-a-b
+          - app-c
+          - app-d-r
+          - app-s-z-example-fuzz
           - generators
           - helpers
           - cli
@@ -679,18 +708,30 @@ jobs:
           TEST_SHARD: ${{ matrix.shard }}
         run: |
           set -euo pipefail
-          package_output="$(./scripts/ci/test-packages.sh list "$TEST_SHARD")"
+          # Every app partition shard tests the same single internal/app
+          # package, so the package query uses the base shard name and the
+          # partition only selects which tests run.
+          package_shard="$TEST_SHARD"
+          case "$TEST_SHARD" in
+            app-*) package_shard=app ;;
+          esac
+          package_output="$(./scripts/ci/test-packages.sh list "$package_shard")"
           test -n "$package_output"
           mapfile -t packages <<< "$package_output"
           test "${#packages[@]}" -gt 0
-          if [ "$TEST_SHARD" = "app" ]; then
-            # A single long-lived app test process retains every constructed
-            # command tree in framework registries. Isolate Schema assembly and
-            # bounded name ranges so each process releases that state on exit.
-            test "${#packages[@]}" -eq 1
-            ./scripts/ci/run-app-race-tests.sh run "${packages[0]}"
-            exit 0
-          fi
+          case "$TEST_SHARD" in
+            app-*)
+              # A single long-lived app test process retains every constructed
+              # command tree in framework registries. Each partition is its own
+              # job, so that state is released when the process exits and the
+              # partitions run concurrently instead of end to end. The helper
+              # still verifies that the partition patterns cover every top-level
+              # test exactly once before running the one it was asked for.
+              test "${#packages[@]}" -eq 1
+              ./scripts/ci/run-app-race-tests.sh run "${packages[0]}" "${TEST_SHARD#app-}"
+              exit 0
+              ;;
+          esac
           # cli/smoke own heavy NewRootCommand / Schema assembly under -race;
           # give them a dedicated package timeout on slower hosted runners.
           timeout_budget=12m
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 0a1f2dfc..9afe33b3 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,37 @@ The format is inspired by [Keep a Changelog](https://keepachangelog.com/) and th
 
 ## [Unreleased]
 
+## [1.0.59-beta.2] - 2026-08-17
+
+### Added
+
+- **Privacy-safe CLI telemetry** (#1009) — reports reviewed command outcomes and profile identity dimensions while excluding command arguments, output, paths, device fingerprints, and automatic system dimensions; `DO_NOT_TRACK=1` disables reporting.
+
+- **Feedback survey entry in root help** (#1019) — `dws --help` now closes with a Feedback section linking the user-experience survey form.
+
+- **Wiki Shortcut workflows** — publishes 20 reviewed space, member, node, and
+  activity shortcuts with strict collection validation, cursor handling,
+  write-terminal evidence, safe read-backs where the backend supports them,
+  task-oriented routing, and documented backend
+  boundaries.
+
+### Changed
+
+- **Chat IM ID flags** (#954) — standardizes chat command entry points on `--conversation-id` for conversation IDs and `--message-id` for message IDs, so help, Schema, and Agent recommendations use the same canonical flags.
+- **Legacy chat flag compatibility** (#954) — keeps older chat IM ID flags such as `--group`, `--id`, `--chat`, `--open-conversation-id`, `--msg-id`, and `--open-message-id` working as compatibility aliases where applicable, while hiding migrated aliases from recommended help and Schema surfaces.
+- **Chat group bots target flag** (#954) — keeps `dws chat group bots` on the visible `--group` flag; this command does not register `--group-name`, and `--group` accepts either an openConversationId or a uniquely resolved group name.
+
+- **Faster Schema Catalog assembly** — projects typed values into payload JSON
+  without re-running a validation scan over documents `json.Marshal` has just
+  produced, cutting roughly a third of the projection work across the full tool
+  set. Untrusted JSON input keeps its existing validation.
+
+### Fixed
+
+- **Chat card update evidence** — distinguishes an accepted update request from an independently verified visible update, preserving the real `bizId` and warning callers not to repeat an unverified write.
+- **Chat command guidance** — splits message and group references by task and explains that `--from` is ambiguous between sender and time-range intent.
+
+
 ## [1.0.59-beta.1] - 2026-08-14
 
 ### Added
diff --git a/Formula/dingtalk-workspace-cli-beta.rb b/Formula/dingtalk-workspace-cli-beta.rb
index 98fc206f..1d76aa1f 100644
--- a/Formula/dingtalk-workspace-cli-beta.rb
+++ b/Formula/dingtalk-workspace-cli-beta.rb
@@ -1,33 +1,33 @@
 class DingtalkWorkspaceCliBeta < Formula
   desc "Automate DingTalk workspace tasks from the terminal (beta channel)"
   homepage "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli"
-  version "1.0.59-beta.1"
+  version "1.0.59-beta.2"
   license "Apache-2.0"
   keg_only "it is the beta channel and conflicts with dingtalk-workspace-cli"
 
   on_macos do
     if Hardware::CPU.arm?
-      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.1/dws-darwin-arm64.tar.gz"
-      sha256 "36a30f3496e0f759c15c0b09f67dbd23b8ecdfff2eebe572f88125b26485830f"
+      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.2/dws-darwin-arm64.tar.gz"
+      sha256 "7f11218d3222f3e93c3b1e94b3a004c061eb0a297b206447ea95fe6b2b1ec674"
     else
-      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.1/dws-darwin-amd64.tar.gz"
-      sha256 "e7a04906380efd8da88cd112e6a512bb6470a3956dc370150037ed6e314db445"
+      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.2/dws-darwin-amd64.tar.gz"
+      sha256 "72f06a334cf29d23123639fabe13bf2951765066136e47ccc6453667c382f8c2"
     end
   end
 
   on_linux do
     if Hardware::CPU.arm?
-      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.1/dws-linux-arm64.tar.gz"
-      sha256 "f59ab055f3e841e4cebc964ae3ef969668475548abaaf8bede44afdca9a3e28d"
+      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.2/dws-linux-arm64.tar.gz"
+      sha256 "3edbfabb7718b53914a2d9efe7b1152e9ebd70765ff0b6f19ad3125e4a2458ed"
     else
-      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.1/dws-linux-amd64.tar.gz"
-      sha256 "2c8f919489d958c7d49262615e81faac70a9fbcae2d589ab54a0bb3c5700a057"
+      url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.2/dws-linux-amd64.tar.gz"
+      sha256 "e1c610070a9c1b3763656cbd53c818290f199097adc07cb9fe629ed765c5e1e0"
     end
   end
 
   resource "skills" do
-    url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.1/dws-skills.zip"
-    sha256 "25f4a7e1d01fa4d771d79201b34b11ee8a24182bdcdc94bfb98d2bd5845bed3b"
+    url "https://github.com/DingTalk-Real-AI/dingtalk-workspace-cli/releases/download/v1.0.59-beta.2/dws-skills.zip"
+    sha256 "aa2854651eaa2c857b526aaec73fd1358d31b11999fe7fd3e99e5060b2522a1f"
   end
 
   def install
diff --git a/internal/helpers/sheet_batch.go b/internal/helpers/sheet_batch.go
index 218a88fd..7ce6e17d 100644
--- a/internal/helpers/sheet_batch.go
+++ b/internal/helpers/sheet_batch.go
@@ -60,6 +60,11 @@ func translateBatchOp(op map[string]any) (map[string]any, error) {
 	if !ok {
 		return nil, fmt.Errorf("unsupported toolName %q: must be a CLI command name (e.g. \"range clear\", \"range update\", \"merge-cells\"). Run 'dws sheet batch-update --help' for the full list", toolName)
 	}
+	if toolName == "set-dropdown" {
+		if err := validateBatchSetDropdownInput(input); err != nil {
+			return nil, err
+		}
+	}
 
 	return map[string]any{
 		"toolName": mapping.mcpTool,
@@ -67,6 +72,33 @@ func translateBatchOp(op map[string]any) (map[string]any, error) {
 	}, nil
 }
 
+func validateBatchSetDropdownInput(input map[string]any) error {
+	if _, exists := input["source-colors"]; exists {
+		return fmt.Errorf("set-dropdown: 顶层 source-colors 不受支持；Inline 颜色请写入 options[].color，SourceRange 颜色写入暂不支持")
+	}
+	if _, exists := input["colors"]; exists {
+		return fmt.Errorf("set-dropdown: 顶层 colors 不受支持；Inline 颜色请写入 options[].color，SourceRange 颜色写入暂不支持")
+	}
+
+	options, hasOptions := input["options"]
+	hasOptions = hasOptions && options != nil
+	sourceRange := batchStr(input, "source-range")
+	sourceSheetID := batchStr(input, "source-sheet-id")
+	hasSourceRange := sourceRange != ""
+	if hasOptions == hasSourceRange {
+		return fmt.Errorf("set-dropdown: options 与 source-range 必须且只能指定一个")
+	}
+	if hasSourceRange != (sourceSheetID != "") {
+		return fmt.Errorf("set-dropdown: source-range 与 source-sheet-id 必须同时指定")
+	}
+	if hasSourceRange {
+		if err := validateDropdownSourceRangeInput(sourceSheetID, sourceRange); err != nil {
+			return fmt.Errorf("set-dropdown: %w", err)
+		}
+	}
+	return nil
+}
+
 // ── BuildXxxArgs: CLI flag → MCP param 转换函数 ──────────────────────────────────
 // 每个函数接收 CLI flag 名（kebab-case）的 map，输出 MCP 参数名（camelCase）的 map。
 // 目前集中放在此文件；后续拆分命令文件时可移到各命令所在文件。
@@ -215,7 +247,15 @@ func BuildSetDropdownArgs(input map[string]any) map[string]any {
 	args := map[string]any{
 		"sheetId": batchStr(input, "sheet-id"),
 		"range":   batchStr(input, "range"),
-		"options": input["options"],
+	}
+	if options, ok := input["options"]; ok && options != nil {
+		args["options"] = options
+	}
+	if sourceRange := batchStr(input, "source-range"); sourceRange != "" {
+		args["sourceRange"] = map[string]any{
+			"sheetId":    batchStr(input, "source-sheet-id"),
+			"a1Notation": sourceRange,
+		}
 	}
 	if v, ok := input["multi-select"]; ok {
 		args["enableMultiSelect"] = v
@@ -417,6 +457,8 @@ func newBatchUpdateCmd() *cobra.Command {
 
 toolName 使用 CLI 命令名（与原子命令一致），input 的键用 CLI flag 名去掉 --。
 CLI 层自动翻译为 MCP toolName + 参数名，无需记忆 MCP 参数名。
+source-range 的语义按 toolName 隔离：set-dropdown 中表示下拉候选项来源，
+range fill/copy-to/move-to 中表示待填充、复制或移动的数据源区域。
 
 支持的 CLI 命令名:
   range clear / range update / merge-cells / unmerge-cells / update-dimension
@@ -425,6 +467,8 @@ CLI 层自动翻译为 MCP toolName + 参数名，无需记忆 MCP 参数名。
   set-dropdown / delete-dropdown / csv-put / delete-float-image
 其中 csv-put 与独立命令语义一致：CSV 字段值以 = 开头时按公式解析，
 前加单引号（例如 "'=1+1"）时写入以 = 开头的字面文本。
+set-dropdown 不接受顶层 colors/source-colors；Inline 颜色应写在 options[].color，
+SourceRange 颜色写入暂不支持。
 
 注意：batch-update 中 group-dimension 适合默认展开分组；需要 --group-state fold 时请使用独立
 dws sheet group-dimension 命令。
diff --git a/internal/helpers/sheet_batch_test.go b/internal/helpers/sheet_batch_test.go
index f7038d08..edb85887 100644
--- a/internal/helpers/sheet_batch_test.go
+++ b/internal/helpers/sheet_batch_test.go
@@ -23,7 +23,15 @@ func TestCrossPlatformCoverageSheetBatchOperationTranslationCoversEveryMapping(t
 		"group-state": "fold",
 	}
 	for name, mapping := range batchOpDispatch {
-		got, err := translateBatchOp(map[string]any{"toolName": name, "input": input})
+		opInput := input
+		if name == "set-dropdown" {
+			opInput = make(map[string]any, len(input))
+			for key, value := range input {
+				opInput[key] = value
+			}
+			delete(opInput, "source-range")
+		}
+		got, err := translateBatchOp(map[string]any{"toolName": name, "input": opInput})
 		if err != nil {
 			t.Errorf("translateBatchOp(%q): %v", name, err)
 			continue
diff --git a/internal/helpers/sheet_data.go b/internal/helpers/sheet_data.go
index 3cdb77b0..be6e9d2d 100644
--- a/internal/helpers/sheet_data.go
+++ b/internal/helpers/sheet_data.go
@@ -388,7 +388,10 @@ range update 与合并区域冲突时返回 MERGED_CELLS_CONFLICT 的行为。
   csv            带 [row=N] 行号前缀的 CSV 文本
   colIndices     列字母映射数组（定位列用 colIndices[j]，禁止手数逗号）
   rowIndices     行号映射数组
-  hasMore        是否因 maxChars 截断
+  hasMore        目标范围是否还有未返回数据
+  truncationReasons  部分返回原因：max_cells / max_chars
+  resolvedRange  未传 --range 时底层解析出的完整目标范围
+  returnedRange  本次实际完整返回的范围
 
 取值模式（--value-render-option）：
   formatted_value  格式化后的展示值（默认），如 ¥1,000.00、2025-06-01
@@ -398,9 +401,14 @@ range update 与合并区域冲突时返回 MERGED_CELLS_CONFLICT 的行为。
 与 range read 的区别：
   - CSV 格式 token 消耗约为 JSON 的 1/3
   - 支持选择取值模式
-  - 自动防爆（max-chars 截断 + has_more 标志）
+  - 自动防爆（30,000 单元格 / max-chars 上限 + hasMore 标志）
   - [row=N] 前缀防止行号计算错误
 
+当 hasMore=true 时本命令不会自动续读；请结合目标范围和
+returnedRange 从下一行显式传 --range。max_cells 不能通过增大
+--max-chars 解决。收到 forbidden.document.sizeOverLimit 表示工作簿整体
+无法装载，应创建更小副本或拆分工作簿，缩小 --range 不能解决。
+
 注意：csv-get 不返回合并单元格结构。查看合并范围请使用
 dws sheet info --node NODE_ID --sheet-id SHEET_ID --format json，并读取 mergedRanges。`,
 		Example: `  dws sheet csv-get --node NODE_ID
diff --git a/internal/helpers/sheet_dimension.go b/internal/helpers/sheet_dimension.go
index 4ce8d7ba..4afaf750 100644
--- a/internal/helpers/sheet_dimension.go
+++ b/internal/helpers/sheet_dimension.go
@@ -6,6 +6,7 @@ import (
 	"strconv"
 	"strings"
 
+	"github.com/DingTalk-Real-AI/dingtalk-workspace-cli/internal/cli"
 	"github.com/DingTalk-Real-AI/dingtalk-workspace-cli/internal/corecmd/contract"
 	"github.com/spf13/cobra"
 )
@@ -18,6 +19,23 @@ func sizeTypeEnumHint(dimension string) string {
 	return "pixel / standard"
 }
 
+func validateDropdownSourceRangeInput(sourceSheetID, sourceRange string) error {
+	if strings.TrimSpace(sourceSheetID) == "" {
+		return fmt.Errorf("--source-sheet-id 不能为空")
+	}
+	sourceRange = strings.TrimSpace(sourceRange)
+	if sourceRange == "" {
+		return fmt.Errorf("--source-range 不能为空")
+	}
+	if strings.Contains(sourceRange, "!") {
+		return fmt.Errorf("--source-range 不能包含工作表前缀；请通过 --source-sheet-id 指定来源工作表")
+	}
+	if strings.HasPrefix(sourceRange, "=") || strings.Contains(sourceRange, ",") {
+		return fmt.Errorf("--source-range 必须是单一连续区域，不能是公式、表达式或多区域引用")
+	}
+	return nil
+}
+
 // newDimensionCmds creates dimension-related commands: insert/delete/update/move/add-dimension,
 // merge-cells, unmerge-cells, and dropdown commands (set/get/delete-dropdown).
 func newDimensionCmds() []*cobra.Command {
@@ -775,15 +793,24 @@ sheetId 支持传入工作表 ID 或工作表名称，可通过 sheet list 获
 		Short: "设置下拉列表",
 		Long: `在钉钉表格的指定单元格范围内设置下拉列表。
 
-设置后，用户可以在这些单元格中从预定义的选项列表中选择值。
-支持自定义每个选项的颜色和是否允许多选。
+Inline 模式通过 --options 提供静态选项，支持选项颜色。
+SourceRange 模式通过 --source-sheet-id + --source-range 引用同一工作簿内的来源区域，
+支持跨工作表、整行和整列引用；读取配置时不会展开来源区域的当前值或颜色。
+两种模式都支持 --multi-select，并且 --options 与 --source-range 必须且只能指定一个。
 如果目标范围已存在下拉列表，会被新的配置覆盖。
 
 nodeId 支持传入文档链接 URL 或文档 ID（dentryUuid），系统自动识别。
 sheetId 支持传入工作表 ID 或工作表名称，可通过 sheet list 获取。
 
 --options 为 JSON 数组，每个元素包含 value（必填）和 color（可选）。
-选项值不能包含英文逗号。`,
+选项值不能包含英文逗号。
+
+--source-range 只接受不含工作表前缀的 A1 区域，例如 T1:T3、T:T 或 1:3；
+来源工作表由 --source-sheet-id 单独指定。SourceRange 颜色写入暂不支持。
+
+已验证行为：工作表重命名、在引用前插入行/列、删除引用前的行会自动调整引用并保持 valid；
+已验证的 move-dimension 场景会使引用变为 invalid。列删除、删除整个来源区域或来源工作表等未覆盖场景不能预设结果；
+结构操作后应先回读 sourceRangeStatus，只有 invalid 时才重新选择来源并写入。`,
 		Example: `  # 设置单选下拉列表
   dws sheet set-dropdown --node NODE_ID --sheet-id SHEET_ID --range "A2:A100" \
     --options '[{"value":"选项1"},{"value":"选项2"},{"value":"选项3"}]'
@@ -793,32 +820,63 @@ sheetId 支持传入工作表 ID 或工作表名称，可通过 sheet list 获
     --options '[{"value":"高","color":"#ff0000"},{"value":"中","color":"#ffaa00"},{"value":"低","color":"#00ff00"}]' \
     --multi-select
 
+  # 引用同一工作簿内另一工作表的区域作为候选项来源
+  dws sheet set-dropdown --node NODE_ID --sheet-id TARGET_SHEET_ID --range "C2:C100" \
+    --source-sheet-id SOURCE_SHEET_ID --source-range "T1:T3"
+
   # 使用文档链接 URL
   dws sheet set-dropdown --node "https://alidocs.dingtalk.com/i/nodes/<DOC_UUID>" \
     --sheet-id SHEET_ID --range "C1:C10" --options '[{"value":"是"},{"value":"否"}]'`,
 		RunE: func(cmd *cobra.Command, args []string) error {
+			if err := validateRequiredFlags(cmd, "node", "sheet-id", "range"); err != nil {
+				return err
+			}
 			optionsStr := mustGetFlag(cmd, "options")
-			var options []map[string]any
-			if err := json.Unmarshal([]byte(optionsStr), &options); err != nil {
-				return fmt.Errorf("--options JSON 解析失败: %w", err)
+			sourceSheetID := mustGetFlag(cmd, "source-sheet-id")
+			sourceRange := mustGetFlag(cmd, "source-range")
+			hasOptions := optionsStr != ""
+			hasSourceRange := sourceRange != ""
+			if hasOptions == hasSourceRange {
+				return fmt.Errorf("--options 与 --source-range 必须且只能指定一个")
 			}
-			if len(options) == 0 {
-				return fmt.Errorf("--options 至少包含 1 个选项")
+			if hasSourceRange && sourceSheetID == "" {
+				return fmt.Errorf("使用 --source-range 时必须同时指定 --source-sheet-id")
 			}
-			for i, opt := range options {
-				val, ok := opt["value"].(string)
-				if !ok || val == "" {
-					return fmt.Errorf("--options[%d] 缺少必填的 value 字段或 value 为空", i)
-				}
-				if strings.Contains(val, ",") {
-					return fmt.Errorf("--options[%d].value 不能包含英文逗号: %q", i, val)
-				}
+			if !hasSourceRange && sourceSheetID != "" {
+				return fmt.Errorf("使用 --source-sheet-id 时必须同时指定 --source-range")
 			}
+
 			toolArgs := map[string]any{
 				"nodeId":  mustGetFlag(cmd, "node"),
 				"sheetId": mustGetFlag(cmd, "sheet-id"),
 				"range":   mustGetFlag(cmd, "range"),
-				"options": options,
+			}
+			if hasOptions {
+				var options []map[string]any
+				if err := json.Unmarshal([]byte(optionsStr), &options); err != nil {
+					return fmt.Errorf("--options JSON 解析失败: %w", err)
+				}
+				if len(options) == 0 {
+					return fmt.Errorf("--options 至少包含 1 个选项")
+				}
+				for i, opt := range options {
+					val, ok := opt["value"].(string)
+					if !ok || val == "" {
+						return fmt.Errorf("--options[%d] 缺少必填的 value 字段或 value 为空", i)
+					}
+					if strings.Contains(val, ",") {
+						return fmt.Errorf("--options[%d].value 不能包含英文逗号: %q", i, val)
+					}
+				}
+				toolArgs["options"] = options
+			} else {
+				if err := validateDropdownSourceRangeInput(sourceSheetID, sourceRange); err != nil {
+					return err
+				}
+				toolArgs["sourceRange"] = map[string]any{
+					"sheetId":    sourceSheetID,
+					"a1Notation": sourceRange,
+				}
 			}
 			if multiSelect, _ := cmd.Flags().GetBool("multi-select"); multiSelect {
 				toolArgs["enableMultiSelect"] = true
@@ -839,36 +897,49 @@ sheetId 支持传入工作表 ID 或工作表名称，可通过 sheet list 获
 				CLIPath:        "sheet set-dropdown",
 				PrimaryCLIPath: "sheet set-dropdown",
 			},
-			Description: "为指定范围设置下拉列表（可多选、可带颜色）。",
+			Description: "为指定范围设置 Inline 或 SourceRange 下拉列表（可多选；颜色仅 Inline 支持）。",
 			Interface: &contract.InterfaceSpec{
 				Mode:         "mcp",
 				Availability: "available",
 				Ref:          &contract.InterfaceRefSpec{ProductID: "sheet", RPCName: "set_dropdown_lists"},
 			},
 			Selection: contract.SelectionSpec{
-				AgentSummary: "为指定范围设置下拉列表（可多选、可带颜色）。",
+				AgentSummary: "为指定范围设置 Inline 或 SourceRange 下拉列表；SourceRange 可跨同一工作簿内的工作表。",
 				UseWhen:      []string{"需要给单元格配置可选值下拉约束时"},
 				AvoidWhen:    []string{"查看已有下拉用 get-dropdown；移除下拉用 delete-dropdown"},
-				Examples:     []string{"dws sheet set-dropdown --node <NODE_ID> --sheet-id <SHEET_ID> --range \"A2:A100\" --options '[{\"value\":\"选项1\"},{\"value\":\"选项2\"}]'"},
+				Examples: []string{
+					"dws sheet set-dropdown --node <NODE_ID> --sheet-id <SHEET_ID> --range \"A2:A100\" --options '[{\"value\":\"选项1\"},{\"value\":\"选项2\"}]'",
+					"dws sheet set-dropdown --node <NODE_ID> --sheet-id <TARGET_SHEET_ID> --range \"B2:B100\" --source-sheet-id <SOURCE_SHEET_ID> --source-range \"T1:T3\"",
+				},
 			},
 			Parameters: []contract.ParamDecl{
 				{Name: "multi-select", Property: "enableMultiSelect"},
 				{Name: "node", Property: "nodeId"},
+				{Name: "source-sheet-id", Property: "sourceRange.sheetId", Required: boolPtr(false), RequiredWhen: "--source-range is provided"},
+				{Name: "source-range", Property: "sourceRange.a1Notation", Required: boolPtr(false), RequiredWhen: "exactly one of --options or --source-range must be provided"},
 			},
 		},
 	})
 	setDropdownCmd.Flags().String("node", "", "表格文档 ID 或 URL (必填)")
 	setDropdownCmd.Flags().String("sheet-id", "", "工作表 ID 或名称 (必填)")
 	setDropdownCmd.Flags().String("range", "", "目标单元格范围，A1 表示法，如 A2:A100 (必填)")
-	setDropdownCmd.Flags().String("options", "", `下拉选项 JSON 数组 (必填)，如 '[{"value":"选项1","color":"#ff0000"}]'`)
+	setDropdownCmd.Flags().String("options", "", `Inline 下拉选项 JSON 数组，与 --source-range 二选一，如 '[{"value":"选项1","color":"#ff0000"}]'`)
+	setDropdownCmd.Flags().String("source-sheet-id", "", "SourceRange 来源工作表 ID；使用 --source-range 时必填")
+	setDropdownCmd.Flags().String("source-range", "", "SourceRange 来源区域，与 --options 二选一；不含工作表前缀，如 T1:T3、T:T 或 1:3")
 	setDropdownCmd.Flags().Bool("multi-select", false, "是否允许多选（默认单选）")
+	cli.AnnotateRuntimeFlagFormat(setDropdownCmd, "source-range", "a1-range")
+	setDropdownCmd.MarkFlagsOneRequired("options", "source-range")
+	setDropdownCmd.MarkFlagsMutuallyExclusive("options", "source-range")
+	setDropdownCmd.MarkFlagsRequiredTogether("source-sheet-id", "source-range")
 
 	getDropdownCmd := &cobra.Command{
 		Use:   "get-dropdown",
 		Short: "获取下拉列表配置",
 		Long: `查询钉钉表格指定范围内的下拉列表配置。
 
-返回范围内所有单元格的下拉列表配置信息，包括选项值和颜色。
+Inline 配置返回 conditionValues 和 options；SourceRange 配置返回
+sourceType/sourceRangeStatus/enableMultiSelect，且仅在 sourceRangeStatus="valid" 时返回
+sourceRange。invalid 结果仍保留配置组，但省略 sourceRange；两种状态都不展开来源区域的候选值或颜色。
 如果范围内存在多个不同的下拉列表配置，会分别返回每组配置及其覆盖的单元格列表。
 如果范围内没有设置下拉列表，返回空。
 
diff --git a/internal/helpers/sheet_dropdown_source_range_test.go b/internal/helpers/sheet_dropdown_source_range_test.go
new file mode 100644
index 00000000..086b7da6
--- /dev/null
+++ b/internal/helpers/sheet_dropdown_source_range_test.go
@@ -0,0 +1,248 @@
+package helpers
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/DingTalk-Real-AI/dingtalk-workspace-cli/internal/corecmd/runtimeannotate"
+)
+
+func TestCrossPlatformCoverageSetDropdownSourceRangeCommand(t *testing.T) {
+	caller := &scriptedToolCaller{steps: []scriptedToolStep{{text: `{"success":true}`}}}
+	installScriptedCaller(t, caller)
+	err := executeDimensionCoverage(t, "set-dropdown",
+		"--node", "node-1",
+		"--sheet-id", "target-sheet",
+		"--range", "B2:B100",
+		"--source-sheet-id", "source-sheet",
+		"--source-range", "$t$1:$t$3",
+		"--multi-select",
+	)
+	if err != nil {
+		t.Fatalf("set SourceRange dropdown: %v", err)
+	}
+	if caller.tool != "set_dropdown_lists" {
+		t.Fatalf("tool = %q, want set_dropdown_lists", caller.tool)
+	}
+	if _, exists := caller.args["options"]; exists {
+		t.Fatalf("SourceRange request must omit options: %#v", caller.args)
+	}
+	source, ok := caller.args["sourceRange"].(map[string]any)
+	if !ok {
+		t.Fatalf("sourceRange = %#v", caller.args["sourceRange"])
+	}
+	if source["sheetId"] != "source-sheet" || source["a1Notation"] != "$t$1:$t$3" {
+		t.Fatalf("sourceRange = %#v", source)
+	}
+	if caller.args["enableMultiSelect"] != true {
+		t.Fatalf("enableMultiSelect = %#v", caller.args["enableMultiSelect"])
+	}
+}
+
+func TestCrossPlatformCoverageSetDropdownInlineCommandStillOmitsSourceRange(t *testing.T) {
+	caller := &scriptedToolCaller{steps: []scriptedToolStep{{text: `{"success":true}`}}}
+	installScriptedCaller(t, caller)
+	err := executeDimensionCoverage(t, "set-dropdown",
+		"--node", "node-1",
+		"--sheet-id", "sheet-1",
+		"--range", "A1:A3",
+		"--options", `[{"value":"one","color":"#ff0000"}]`,
+	)
+	if err != nil {
+		t.Fatalf("set inline dropdown: %v", err)
+	}
+	if _, exists := caller.args["sourceRange"]; exists {
+		t.Fatalf("inline request must omit sourceRange: %#v", caller.args)
+	}
+	options, ok := caller.args["options"].([]map[string]any)
+	if !ok || len(options) != 1 || options[0]["value"] != "one" {
+		t.Fatalf("options = %#v", caller.args["options"])
+	}
+}
+
+func TestCrossPlatformCoverageSetDropdownModeConstraints(t *testing.T) {
+	base := []string{"--node", "node", "--sheet-id", "sheet", "--range", "A1"}
+	tests := []struct {
+		name string
+		args []string
+		want string
+	}{
+		{name: "neither", want: "one of the flags"},
+		{name: "both", args: []string{"--options", `[{"value":"one"}]`, "--source-sheet-id", "source", "--source-range", "A1:A3"}, want: "none of the others"},
+		{name: "source sheet without range", args: []string{"--options", `[{"value":"one"}]`, "--source-sheet-id", "source"}, want: "must all be set"},
+		{name: "range without source sheet", args: []string{"--source-range", "A1:A3"}, want: "must all be set"},
+		{name: "sheet prefix", args: []string{"--source-sheet-id", "source", "--source-range", "Sheet2!A1:A3"}, want: "不能包含工作表前缀"},
+		{name: "formula", args: []string{"--source-sheet-id", "source", "--source-range", "=A1:A3"}, want: "必须是单一连续区域"},
+		{name: "multi region", args: []string{"--source-sheet-id", "source", "--source-range", "A1:A3,C1:C3"}, want: "必须是单一连续区域"},
+		{name: "blank source range", args: []string{"--source-sheet-id", "source", "--source-range", " \t "}, want: "--source-range 不能为空"},
+		{name: "blank source sheet", args: []string{"--source-sheet-id", " \t ", "--source-range", "A1:A3"}, want: "--source-sheet-id 不能为空"},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			caller := &scriptedToolCaller{}
+			installScriptedCaller(t, caller)
+			args := append(append([]string{}, base...), tc.args...)
+			err := executeDimensionCoverage(t, "set-dropdown", args...)
+			if err == nil || !strings.Contains(err.Error(), tc.want) {
+				t.Fatalf("error = %v, want contains %q", err, tc.want)
+			}
+			if caller.calls != 0 {
+				t.Fatalf("calls = %d, validation must fail before request", caller.calls)
+			}
+		})
+	}
+}
+
+func TestCrossPlatformCoverageSetDropdownRunEValidation(t *testing.T) {
+	tests := []struct {
+		name    string
+		options string
+		source  string
+		rangeID string
+		want    string
+	}{
+		{name: "neither mode", want: "必须且只能指定一个"},
+		{name: "source range without sheet", rangeID: "A1:A3", want: "必须同时指定 --source-sheet-id"},
+		{name: "source sheet without range", options: `[{"value":"one"}]`, source: "source", want: "必须同时指定 --source-range"},
+		{name: "invalid options JSON", options: "{", want: "JSON 解析失败"},
+		{name: "missing option value", options: `[{}]`, want: "value 为空"},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			cmd := dimensionCoverageCommand(t, "set-dropdown")
+			for name, value := range map[string]string{
+				"node": "node", "sheet-id": "sheet", "range": "A1",
+				"options": tc.options, "source-sheet-id": tc.source, "source-range": tc.rangeID,
+			} {
+				if err := cmd.Flags().Set(name, value); err != nil {
+					t.Fatalf("set --%s: %v", name, err)
+				}
+			}
+			err := cmd.RunE(cmd, nil)
+			if err == nil || !strings.Contains(err.Error(), tc.want) {
+				t.Fatalf("error = %v, want contains %q", err, tc.want)
+			}
+		})
+	}
+}
+
+func TestCrossPlatformCoverageSetDropdownPreservesSchemaConstraints(t *testing.T) {
+	cmd := dimensionCoverageCommand(t, "set-dropdown")
+	if constraints := runtimeannotate.CommandConstraints(cmd); !runtimeannotate.ConstraintsEmpty(constraints) {
+		t.Fatalf("constraints = %#v, want no new schema-level constraints", constraints)
+	}
+}
+
+func TestCrossPlatformCoverageBatchSetDropdownSourceRange(t *testing.T) {
+	got, err := translateBatchOp(map[string]any{
+		"toolName": "set-dropdown",
+		"input": map[string]any{
+			"sheet-id":        "target-sheet",
+			"range":           "B2:B100",
+			"source-sheet-id": "source-sheet",
+			"source-range":    "T:T",
+			"multi-select":    true,
+		},
+	})
+	if err != nil {
+		t.Fatalf("translate SourceRange batch op: %v", err)
+	}
+	input := got["input"].(map[string]any)
+	if _, exists := input["options"]; exists {
+		t.Fatalf("SourceRange batch input must omit options: %#v", input)
+	}
+	source := input["sourceRange"].(map[string]any)
+	if source["sheetId"] != "source-sheet" || source["a1Notation"] != "T:T" {
+		t.Fatalf("sourceRange = %#v", source)
+	}
+
+	invalid := []map[string]any{
+		{},
+		{"options": []any{map[string]any{"value": "one"}}, "source-sheet-id": "source", "source-range": "A1:A3"},
+		{"source-range": "A1:A3"},
+		{"source-sheet-id": "source", "source-range": "A1:A3", "colors": []any{"#fff"}},
+		{"source-sheet-id": "source", "source-range": "A1:A3", "source-colors": []any{"#fff"}},
+		{"source-sheet-id": "source", "source-range": "Sheet2!A1:A3"},
+	}
+	for _, value := range invalid {
+		if _, err := translateBatchOp(map[string]any{"toolName": "set-dropdown", "input": value}); err == nil {
+			t.Errorf("invalid batch SourceRange input %#v returned nil", value)
+		}
+	}
+}
+
+func TestCrossPlatformCoverageBatchSetDropdownValidationGuidance(t *testing.T) {
+	tests := []struct {
+		name  string
+		input map[string]any
+		want  string
+	}{
+		{
+			name:  "inline top-level colors",
+			input: map[string]any{"options": []any{map[string]any{"value": "one"}}, "colors": []any{"#fff"}},
+			want:  "Inline 颜色请写入 options[].color",
+		},
+		{
+			name:  "source top-level colors",
+			input: map[string]any{"source-sheet-id": "source", "source-range": "A1:A3", "source-colors": []any{"#fff"}},
+			want:  "SourceRange 颜色写入暂不支持",
+		},
+		{
+			name:  "blank source range",
+			input: map[string]any{"source-sheet-id": "source", "source-range": " \t "},
+			want:  "--source-range 不能为空",
+		},
+		{
+			name:  "blank source sheet",
+			input: map[string]any{"source-sheet-id": " \t ", "source-range": "A1:A3"},
+			want:  "--source-sheet-id 不能为空",
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			_, err := translateBatchOp(map[string]any{"toolName": "set-dropdown", "input": tc.input})
+			if err == nil || !strings.Contains(err.Error(), tc.want) {
+				t.Fatalf("error = %v, want contains %q", err, tc.want)
+			}
+		})
+	}
+}
+
+func TestCrossPlatformCoverageDropdownHelpDescribesDynamicAndInvalidContracts(t *testing.T) {
+	setDropdown := dimensionCoverageCommand(t, "set-dropdown")
+	if !strings.Contains(setDropdown.Long, "会自动调整引用并保持 valid") ||
+		!strings.Contains(setDropdown.Long, "只有 invalid 时才重新选择来源并写入") {
+		t.Fatalf("set-dropdown help does not describe verified structural behavior:\n%s", setDropdown.Long)
+	}
+	getDropdown := dimensionCoverageCommand(t, "get-dropdown")
+	if !strings.Contains(getDropdown.Long, `仅在 sourceRangeStatus="valid" 时返回`) ||
+		!strings.Contains(getDropdown.Long, "invalid 结果仍保留配置组，但省略 sourceRange") {
+		t.Fatalf("get-dropdown help does not describe conditional sourceRange readback:\n%s", getDropdown.Long)
+	}
+}
+
+func TestCrossPlatformCoverageValidateSourceRangeDataValidation(t *testing.T) {
+	valid := []any{
+		map[string]any{"type": "dropdown", "sourceRange": map[string]any{"sheetId": "source", "a1Notation": "T1:T3"}},
+		map[string]any{"type": "dropdown", "sourceRange": map[string]any{"sheetId": "missing-sheet", "a1Notation": "not-validated-locally"}, "enableMultiSelect": true},
+	}
+	for _, value := range valid {
+		if err := validateDataValidation(value, "dv"); err != nil {
+			t.Errorf("valid SourceRange data validation %#v: %v", value, err)
+		}
+	}
+
+	invalid := []any{
+		map[string]any{"type": "dropdown", "options": []any{map[string]any{"value": "one"}}, "sourceRange": map[string]any{"sheetId": "source", "a1Notation": "T1:T3"}},
+		map[string]any{"type": "dropdown", "sourceRange": "T1:T3"},
+		map[string]any{"type": "dropdown", "sourceRange": map[string]any{"a1Notation": "T1:T3"}},
+		map[string]any{"type": "dropdown", "sourceRange": map[string]any{"sheetId": "source"}},
+		map[string]any{"type": "dropdown", "sourceRange": map[string]any{"sheetId": "source", "a1Notation": "T1:T3", "colors": []any{"#fff"}}},
+		map[string]any{"type": "dropdown", "sourceRange": map[string]any{"sheetId": "source", "a1Notation": "T1:T3"}, "enableMultiSelect": "yes"},
+	}
+	for _, value := range invalid {
+		if err := validateDataValidation(value, "dv"); err == nil {
+			t.Errorf("invalid SourceRange data validation %#v returned nil", value)
+		}
+	}
+}
diff --git a/internal/helpers/sheet_range_ops.go b/internal/helpers/sheet_range_ops.go
index 5d6c8217..1c3a8fb5 100644
--- a/internal/helpers/sheet_range_ops.go
+++ b/internal/helpers/sheet_range_ops.go
@@ -26,6 +26,16 @@ func newRangeCmd() *cobra.Command {
   value           单元格值（内容由 --value-render-option 决定）
   dataValidation  数据验证配置（下拉列表/复选框），无则为 null
   hyperlink       单元格级超链接（path/sheet/range），无则省略
+顶层还会返回 rowIndices / colIndices 与完成度字段：
+  hasMore           目标范围是否还有未返回数据
+  truncationReasons 部分返回原因（max_cells）
+  resolvedRange     未传 --range 时底层解析出的完整目标范围
+  returnedRange     本次实际完整返回的范围
+
+单次最多返回 30,000 个单元格。hasMore=true 是部分成功，本命令
+不会自动续读；请结合目标范围和 returnedRange 从下一行显式传
+--range。收到 forbidden.document.sizeOverLimit 则表示工作簿整体无法装载，
+应创建更小副本或拆分工作簿，缩小 --range 不能解决。
 
 注意：range read/get 不返回合并单元格结构。查看合并范围请使用
 dws sheet info --node NODE_ID --sheet-id SHEET_ID --format json，并读取 mergedRanges。
@@ -157,7 +167,12 @@ dws sheet info --node NODE_ID --sheet-id SHEET_ID --format json，并读取 merg
   2) {"dataValidation":{"type":"none"}}             → 显式清除该单元格 DV
   3) {"dataValidation":{"type":"dropdown",...}}     → 写新 dropdown（覆盖）
      {"dataValidation":{"type":"checkbox",...}}    → 写新 checkbox（覆盖）
-  dropdown: {"dataValidation":{"type":"dropdown","options":[{"value":"选项1"}],"enableMultiSelect":false}}
+  Inline dropdown:
+    {"dataValidation":{"type":"dropdown","options":[{"value":"选项1"}],"enableMultiSelect":false}}
+  SourceRange dropdown（同一工作簿内可跨工作表；不展开来源 values，不支持 colors）：
+    {"dataValidation":{"type":"dropdown","sourceRange":{"sheetId":"SOURCE_SHEET_ID","a1Notation":"T1:T3"},"enableMultiSelect":false}}
+  dropdown 的 options 与 sourceRange 必须且只能传一个。SourceRange 支持普通区域、整行和整列；
+  a1Notation 不带工作表前缀，来源工作表单独写在 sheetId。
   checkbox: {"dataValidation":{"type":"checkbox","checked":true}}
   可与 text/richText 共存，也可单独使用（如 {dataValidation:{type:"none"}} 仅清除 DV 不写值）
 
@@ -169,6 +184,8 @@ dws sheet info --node NODE_ID --sheet-id SHEET_ID --format json，并读取 merg
   - 只设样式或批量刷整片区域样式请用 dws sheet range set-style；写值同时设置少量 cell 样式可用 cellStyles
   - 目标范围与已有合并区域冲突时，range update 会返回 MERGED_CELLS_CONFLICT；先用 sheet info 查看 mergedRanges，取消合并后写入，必要时再重新合并
   - csv-put 的合并处理不同：目标区域含合并单元格时会打散合并并写入 CSV 值或公式
+  - SourceRange 在已验证的重命名、引用前插入行/列、删除引用前行的场景会自动调整；move-dimension 及其他未覆盖的删除/移动场景后先回读 sourceRangeStatus，仅 invalid 时重新选源写入
+  - 同一 cell 的 value/style 已写入但 SourceRange 校验失败时，服务端可返回 success=true，并通过 message 明确下拉未创建；必须检查 message，必要时重新读取确认
   - 清空整片区域请用 dws sheet range clear`,
 		Example: `  # 写入文本
   dws sheet range update --node NODE_ID --sheet-id SHEET_ID --range "A1:B2" \
diff --git a/internal/helpers/sheet_read_completion_test.go b/internal/helpers/sheet_read_completion_test.go
new file mode 100644
index 00000000..dca4bfdc
--- /dev/null
+++ b/internal/helpers/sheet_read_completion_test.go
@@ -0,0 +1,202 @@
+package helpers
+
+import (
+	"bytes"
+	"encoding/json"
+	"errors"
+	"strings"
+	"testing"
+)
+
+func executeSheetReadCompletionCommand(t *testing.T, caller *scriptedToolCaller, args ...string) (*bytes.Buffer, error) {
+	t.Helper()
+	installScriptedCaller(t, caller)
+	installSheetProductArgs(t)
+
+	stdout := &bytes.Buffer{}
+	deps.Out.w = stdout
+
+	cmd := newSheetCommand()
+	cmd.SilenceErrors = true
+	cmd.SilenceUsage = true
+	cmd.SetArgs(args)
+	return stdout, cmd.Execute()
+}
+
+func decodeSheetReadOutput(t *testing.T, stdout *bytes.Buffer) map[string]any {
+	t.Helper()
+	var got map[string]any
+	if err := json.Unmarshal(stdout.Bytes(), &got); err != nil {
+		t.Fatalf("decode sheet read output %q: %v", stdout.String(), err)
+	}
+	return got
+}
+
+func assertSheetReadCompletionMetadata(t *testing.T, got map[string]any) {
+	t.Helper()
+	if hasMore, ok := got["hasMore"].(bool); !ok || !hasMore {
+		t.Fatalf("hasMore = %#v, want true", got["hasMore"])
+	}
+	if got["resolvedRange"] != "A1:A30001" {
+		t.Fatalf("resolvedRange = %#v", got["resolvedRange"])
+	}
+	if got["returnedRange"] != "A1:A30000" {
+		t.Fatalf("returnedRange = %#v", got["returnedRange"])
+	}
+	reasons, ok := got["truncationReasons"].([]any)
+	if !ok || len(reasons) == 0 || reasons[0] != "max_cells" {
+		t.Fatalf("truncationReasons = %#v", got["truncationReasons"])
+	}
+}
+
+// csv-get intentionally projects the MCP response rather than inventing a
+// client-side page model. Keep all completion metadata visible and make one
+// request only, even when the server reports a partial success.
+func TestSheetCSVGetPreservesCompletionMetadataWithoutAutoPagination(t *testing.T) {
+	const response = `{"csv":"[row=1]1\n","colIndices":["A"],"rowIndices":[1],"hasMore":true,"truncationReasons":["max_cells","max_chars"],"resolvedRange":"A1:A30001","returnedRange":"A1:A30000","message":"partial"}`
+	caller := &scriptedToolCaller{
+		format: "raw",
+		steps:  []scriptedToolStep{{text: response}},
+	}
+
+	stdout, err := executeSheetReadCompletionCommand(t, caller,
+		"csv-get", "--node", "NODE")
+	if err != nil {
+		t.Fatalf("csv-get: %v", err)
+	}
+	if caller.calls != 1 || caller.tool != "get_range_as_csv" {
+		t.Fatalf("calls/tool = %d/%q, want one get_range_as_csv call", caller.calls, caller.tool)
+	}
+	if _, exists := caller.args["range"]; exists {
+		t.Fatalf("omitted range was unexpectedly synthesized by the CLI: %#v", caller.args["range"])
+	}
+	if got := strings.TrimSpace(stdout.String()); got != response {
+		t.Fatalf("raw csv-get response changed:\n got: %s\nwant: %s", got, response)
+	}
+
+	got := decodeSheetReadOutput(t, stdout)
+	assertSheetReadCompletionMetadata(t, got)
+	reasons := got["truncationReasons"].([]any)
+	if len(reasons) != 2 || reasons[1] != "max_chars" {
+		t.Fatalf("truncationReasons = %#v, want max_cells and max_chars", reasons)
+	}
+}
+
+// range read parses the get_cell_infos envelope to remove empty per-cell
+// schema shells. That cleanup must not discard new top-level completion fields.
+func TestSheetRangeReadPreservesCompletionMetadataWithoutAutoPagination(t *testing.T) {
+	const response = `{"cells":[[{"value":"1","dataValidation":{"type":null},"hyperlink":{"url":null}}]],"colIndices":["A"],"rowIndices":[1],"hasMore":true,"truncationReasons":["max_cells"],"resolvedRange":"A1:A30001","returnedRange":"A1:A30000","message":"partial"}`
+	caller := &scriptedToolCaller{
+		format: "json",
+		steps:  []scriptedToolStep{{text: response}},
+	}
+
+	stdout, err := executeSheetReadCompletionCommand(t, caller,
+		"range", "read", "--node", "NODE")
+	if err != nil {
+		t.Fatalf("range read: %v", err)
+	}
+	if caller.calls != 1 || caller.tool != "get_cell_infos" {
+		t.Fatalf("calls/tool = %d/%q, want one get_cell_infos call", caller.calls, caller.tool)
+	}
+	if _, exists := caller.args["range"]; exists {
+		t.Fatalf("omitted range was unexpectedly synthesized by the CLI: %#v", caller.args["range"])
+	}
+
+	got := decodeSheetReadOutput(t, stdout)
+	assertSheetReadCompletionMetadata(t, got)
+	cells := got["cells"].([]any)
+	row := cells[0].([]any)
+	cell := row[0].(map[string]any)
+	if _, exists := cell["dataValidation"]; exists {
+		t.Fatalf("empty dataValidation shell was not removed: %#v", cell)
+	}
+	if _, exists := cell["hyperlink"]; exists {
+		t.Fatalf("empty hyperlink shell was not removed: %#v", cell)
+	}
+}
+
+// sheet +read calls get_cell_infos through this direct passthrough instead of
+// callMCPToolCellInfos. Lock that third read surface to the same raw envelope
+// contract and to one request only.
+func TestSheetCellInfosDirectPassthroughPreservesCompletionMetadata(t *testing.T) {
+	const response = `{"cells":[[{"value":"1"}]],"colIndices":["A"],"rowIndices":[1],"hasMore":true,"truncationReasons":["max_cells"],"resolvedRange":"A1:A30001","returnedRange":"A1:A30000","message":"partial"}`
+	caller := &scriptedToolCaller{
+		format: "raw",
+		steps:  []scriptedToolStep{{text: response}},
+	}
+	installScriptedCaller(t, caller)
+	stdout := &bytes.Buffer{}
+	deps.Out.w = stdout
+
+	if err := CallMCPToolOnServer("sheet", "get_cell_infos", map[string]any{"nodeId": "NODE"}); err != nil {
+		t.Fatalf("direct get_cell_infos: %v", err)
+	}
+	if caller.calls != 1 || caller.server != "sheet" || caller.tool != "get_cell_infos" {
+		t.Fatalf("calls/server/tool = %d/%q/%q", caller.calls, caller.server, caller.tool)
+	}
+	if got := strings.TrimSpace(stdout.String()); got != response {
+		t.Fatalf("raw get_cell_infos response changed:\n got: %s\nwant: %s", got, response)
+	}
+	assertSheetReadCompletionMetadata(t, decodeSheetReadOutput(t, stdout))
+}
+
+// Large-CP failure is not a pageable success. Both the raw csv-get path and
+// both get_cell_infos paths must retain the backend code and safe user message
+// exactly, matching the generic get_all_sheets error-classification path.
+func TestSheetReadPreservesWorkbookSizeOverLimitError(t *testing.T) {
+	const response = `{"success":false,"errorCode":"forbidden.document.sizeOverLimit","errorMessage":"The workbook data is too large to process. Use a smaller copy or split the workbook, then try again."}`
+	tests := []struct {
+		name   string
+		args   []string
+		tool   string
+		direct bool
+	}{
+		{name: "csv-get", args: []string{"csv-get", "--node", "NODE"}, tool: "get_range_as_csv"},
+		{name: "range-read", args: []string{"range", "read", "--node", "NODE"}, tool: "get_cell_infos"},
+		{name: "cell-infos-direct", tool: "get_cell_infos", direct: true},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			caller := &scriptedToolCaller{
+				format: "json",
+				steps:  []scriptedToolStep{{text: response}},
+			}
+			var stdout *bytes.Buffer
+			var err error
+			if test.direct {
+				installScriptedCaller(t, caller)
+				stdout = &bytes.Buffer{}
+				deps.Out.w = stdout
+				err = CallMCPToolOnServer("sheet", test.tool, map[string]any{"nodeId": "NODE"})
+			} else {
+				stdout, err = executeSheetReadCompletionCommand(t, caller, test.args...)
+			}
+			if err == nil {
+				t.Fatal("large-CP response was accepted as success")
+			}
+			var cliErr *CLIError
+			if !errors.As(err, &cliErr) || cliErr.Code != CodeMCPToolError {
+				t.Fatalf("error = %T %v, want MCP tool error", err, err)
+			}
+			if cliErr.Message != response {
+				t.Fatalf("backend error payload changed:\n got: %s\nwant: %s", cliErr.Message, response)
+			}
+			for _, want := range []string{
+				"forbidden.document.sizeOverLimit",
+				"The workbook data is too large to process.",
+			} {
+				if !strings.Contains(err.Error(), want) {
+					t.Errorf("error %q does not preserve %q", err, want)
+				}
+			}
+			if stdout.Len() != 0 {
+				t.Fatalf("failure wrote a success body: %q", stdout.String())
+			}
+			if caller.calls != 1 || caller.tool != test.tool {
+				t.Fatalf("calls/tool = %d/%q, want one %s call", caller.calls, caller.tool, test.tool)
+			}
+		})
+	}
+}
diff --git a/internal/helpers/sheet_validate.go b/internal/helpers/sheet_validate.go
index e40a7cb3..f32e6836 100644
--- a/internal/helpers/sheet_validate.go
+++ b/internal/helpers/sheet_validate.go
@@ -239,22 +239,48 @@ func validateDataValidation(dvRaw any, path string) error {
 	}
 	switch dvType {
 	case "dropdown":
-		optionsRaw, ok := dv["options"]
-		if !ok {
-			return fmt.Errorf("%s: type=dropdown 必须包含 options 数组", path)
+		optionsRaw, hasOptions := dv["options"]
+		sourceRangeRaw, hasSourceRange := dv["sourceRange"]
+		hasOptions = hasOptions && optionsRaw != nil
+		hasSourceRange = hasSourceRange && sourceRangeRaw != nil
+		if hasOptions == hasSourceRange {
+			return fmt.Errorf("%s: type=dropdown 的 options 与 sourceRange 必须且只能包含一个", path)
 		}
-		options, ok := optionsRaw.([]any)
-		if !ok || len(options) == 0 {
-			return fmt.Errorf("%s.options 必须为非空数组", path)
-		}
-		for i, opt := range options {
-			optMap, ok := opt.(map[string]any)
-			if !ok {
-				return fmt.Errorf("%s.options[%d] 必须为 object（如 {\"value\":\"选项\"}）", path, i)
+		if hasOptions {
+			options, ok := optionsRaw.([]any)
+			if !ok || len(options) == 0 {
+				return fmt.Errorf("%s.options 必须为非空数组", path)
 			}
-			val, ok := optMap["value"].(string)
-			if !ok || val == "" {
-				return fmt.Errorf("%s.options[%d].value 必须为非空字符串", path, i)
+			for i, opt := range options {
+				optMap, ok := opt.(map[string]any)
+				if !ok {
+					return fmt.Errorf("%s.options[%d] 必须为 object（如 {\"value\":\"选项\"}）", path, i)
+				}
+				val, ok := optMap["value"].(string)
+				if !ok || val == "" {
+					return fmt.Errorf("%s.options[%d].value 必须为非空字符串", path, i)
+				}
+			}
+		} else {
+			sourceRange, ok := sourceRangeRaw.(map[string]any)
+			if !ok {
+				return fmt.Errorf("%s.sourceRange 必须为 object", path)
+			}
+			if _, exists := sourceRange["colors"]; exists {
+				return fmt.Errorf("%s.sourceRange.colors 暂不支持", path)
+			}
+			sheetID, ok := sourceRange["sheetId"].(string)
+			if !ok || strings.TrimSpace(sheetID) == "" {
+				return fmt.Errorf("%s.sourceRange.sheetId 必须为非空字符串", path)
+			}
+			a1Notation, ok := sourceRange["a1Notation"].(string)
+			if !ok || strings.TrimSpace(a1Notation) == "" {
+				return fmt.Errorf("%s.sourceRange.a1Notation 必须为非空字符串", path)
+			}
+		}
+		if multi, exists := dv["enableMultiSelect"]; exists {
+			if _, ok := multi.(bool); !ok {
+				return fmt.Errorf("%s.enableMultiSelect 必须为 boolean", path)
 			}
 		}
 	case "checkbox":
diff --git a/scripts/ci/run-app-race-tests.sh b/scripts/ci/run-app-race-tests.sh
index 9922e9ab..3f586dd0 100755
--- a/scripts/ci/run-app-race-tests.sh
+++ b/scripts/ci/run-app-race-tests.sh
@@ -6,16 +6,35 @@ ROOT="$(CDPATH= cd -- "$(dirname -- "$0")/../.." && pwd)"
 usage() {
 	printf '%s\n' \
 		"usage: $0 verify <app-package>" \
-		"       $0 run <app-package>" >&2
+		"       $0 run <app-package> [partition]" \
+		"       $0 list-partitions" >&2
 	exit 2
 }
 
-[ "$#" -eq 2 ] || usage
-mode="$1"
-app_package="$2"
+# Single source of truth for the partition set. CI runs one job per partition and
+# pins its shard names to this list, so a name that appears here without a
+# dispatch entry below fails closed rather than silently skipping tests.
+APP_PARTITIONS='schema a-b c d-r s-z-example-fuzz'
 
+mode="${1:-}"
+partition=""
 case "$mode" in
-	verify|run) ;;
+	list-partitions)
+		[ "$#" -eq 1 ] || usage
+		for name in $APP_PARTITIONS; do
+			printf '%s\n' "$name"
+		done
+		exit 0
+		;;
+	verify)
+		[ "$#" -eq 2 ] || usage
+		app_package="$2"
+		;;
+	run)
+		[ "$#" -eq 2 ] || [ "$#" -eq 3 ] || usage
+		app_package="$2"
+		partition="${3:-}"
+		;;
 	*) usage ;;
 esac
 
@@ -70,21 +89,49 @@ if [ "$unmatched_count" -ne 0 ]; then
 	exit 1
 fi
 
-for partition in \
+# The loop variable is deliberately not named "partition": that name holds the
+# partition requested on the command line, and run mode still executes this
+# discovery pass before dispatching.
+classified=''
+for spec in \
 	"schema:$schema_count" \
 	"a-b:$ab_count" \
 	"c:$c_count" \
 	"d-r:$dr_count" \
 	"s-z-example-fuzz:$sz_count"
 do
-	name="${partition%%:*}"
-	count="${partition#*:}"
+	name="${spec%%:*}"
+	count="${spec#*:}"
+	classified="$classified $name"
 	if [ "$count" -eq 0 ]; then
 		printf 'app race partition %s is empty\n' "$name" >&2
 		exit 1
 	fi
 done
 
+# The counters above and APP_PARTITIONS must describe the same set in both
+# directions. A counted partition missing from APP_PARTITIONS would never be
+# dispatched by any CI job, and a dispatchable partition with no counter would
+# escape the exact-coverage check above.
+for name in $APP_PARTITIONS; do
+	case " $classified " in
+		*" $name "*) ;;
+		*)
+			printf 'app partition %s has no coverage counter\n' "$name" >&2
+			exit 1
+			;;
+	esac
+done
+for name in $classified; do
+	case " $APP_PARTITIONS " in
+		*" $name "*) ;;
+		*)
+			printf 'counted partition %s is not a dispatchable app partition\n' "$name" >&2
+			exit 1
+			;;
+	esac
+done
+
 total_count="$(wc -l < "$tests" | tr -d ' ')"
 assigned_count=$((schema_count + ab_count + c_count + dr_count + sz_count))
 if [ "$assigned_count" -ne "$total_count" ]; then
@@ -101,26 +148,73 @@ fi
 
 run_partition() {
 	name="$1"
-	run_pattern="$2"
-	skip_pattern="${3:-}"
+	instrumentation="$2"
+	run_pattern="$3"
+	skip_pattern="${4:-}"
 
-	printf 'running internal/app race partition %s\n' "$name"
+	# Fail closed on an unrecognized mode: a typo must not silently drop race
+	# instrumentation from a partition that is supposed to carry it.
+	case "$instrumentation" in
+		race|no-race) ;;
+		*)
+			printf 'unknown instrumentation %s for app partition %s\n' \
+				"$instrumentation" "$name" >&2
+			exit 1
+			;;
+	esac
+
+	printf 'running internal/app %s partition %s\n' "$instrumentation" "$name"
+	set -- -v -count=1 -timeout=15m -run "$run_pattern"
 	if [ -n "$skip_pattern" ]; then
-		go test -v -race -count=1 -timeout=15m \
-			-run "$run_pattern" -skip "$skip_pattern" "$app_package"
-	else
-		go test -v -race -count=1 -timeout=15m \
-			-run "$run_pattern" "$app_package"
+		set -- "$@" -skip "$skip_pattern"
 	fi
+	if [ "$instrumentation" = race ]; then
+		set -- -race "$@"
+	fi
+	go test "$@" "$app_package"
 }
 
 # Schema assembly has the largest transient memory footprint. Run those tests
 # in a fresh process, then keep each remaining name range in its own process so
 # command trees retained by process-global registries are released between
 # partitions. The complementary run/skip patterns preserve the full test set.
+#
+# The schema partition runs uninstrumented. Its tests assert structural
+# Schema-to-Cobra contracts over a single goroutine: none of them call
+# t.Parallel or start a goroutine, so the race detector has no concurrent access
+# to observe here. The process-global lazy metadata that does need race coverage
+# (schema_source_root's atomic.Value, the parameter-binding lazy loaders) is
+# exercised by internal/cli's concurrent tests, which stay instrumented. The
+# instrumentation is not free on this partition: its shared sync.Once Catalog
+# build is allocation-heavy, and -race made the partition roughly 11x slower
+# (26s -> 291s locally, 357s in CI) without being able to report anything.
 schema_pattern='^Test.*Schema'
-run_partition schema "$schema_pattern"
-run_partition a-b '^Test[A-B]' "$schema_pattern"
-run_partition c '^TestC' "$schema_pattern"
-run_partition d-r '^Test[D-R]' "$schema_pattern"
-run_partition s-z-example-fuzz '^(Test[S-Z]|Example|Fuzz)' "$schema_pattern"
+
+# Dispatch table for the partition set declared in APP_PARTITIONS. CI passes one
+# partition per job so they run concurrently; running without a partition keeps
+# the original end-to-end behaviour for local use and for any caller that wants
+# the whole package in one invocation.
+run_named_partition() {
+	case "$1" in
+		schema) run_partition schema no-race "$schema_pattern" ;;
+		a-b) run_partition a-b race '^Test[A-B]' "$schema_pattern" ;;
+		c) run_partition c race '^TestC' "$schema_pattern" ;;
+		d-r) run_partition d-r race '^Test[D-R]' "$schema_pattern" ;;
+		s-z-example-fuzz)
+			run_partition s-z-example-fuzz race '^(Test[S-Z]|Example|Fuzz)' "$schema_pattern"
+			;;
+		*)
+			printf 'unknown app partition: %s\n' "$1" >&2
+			exit 1
+			;;
+	esac
+}
+
+if [ -n "$partition" ]; then
+	run_named_partition "$partition"
+	exit 0
+fi
+
+for name in $APP_PARTITIONS; do
+	run_named_partition "$name"
+done
diff --git a/skills/mono/references/products/sheet/sheet-batch-operations.md b/skills/mono/references/products/sheet/sheet-batch-operations.md
index 8d922bec..66fafc90 100644
--- a/skills/mono/references/products/sheet/sheet-batch-operations.md
+++ b/skills/mono/references/products/sheet/sheet-batch-operations.md
@@ -83,7 +83,8 @@ Example:
     {"toolName":"range update","input":{"sheet-id":"Sheet1","range":"A1","values":[[{"type":"text","text":"hello"}]]}},
     {"toolName":"merge-cells","input":{"sheet-id":"Sheet1","range":"A1:B1","merge-type":"mergeAll"}},
     {"toolName":"update-dimension","input":{"sheet-id":"Sheet1","dimension":"ROWS","start-index":"1","length":1,"pixel-size":40}},
-    {"toolName":"group-dimension","input":{"sheet-id":"Sheet1","range":"3:7","group-state":"expand"}}
+    {"toolName":"group-dimension","input":{"sheet-id":"Sheet1","range":"3:7","group-state":"expand"}},
+    {"toolName":"set-dropdown","input":{"sheet-id":"Sheet1","range":"C2:C100","source-sheet-id":"SourceSheet","source-range":"T1:T3"}}
   ]'
   dws sheet batch-update --node <NODE_ID> --continue-on-error --operations '[...]'
 Flags:
@@ -101,6 +102,9 @@ Notes:
   - operations 最多 20 条
   - 当需要对多个区域执行相同清除时，优先使用 `range batch-clear`（更简洁）
   - `csv-put` 子操作与独立命令语义一致：CSV 字段值以 `=` 开头时按公式解析；前加单引号时写入以 `=` 开头的字面文本
+  - `set-dropdown` 的 `input` 中，Inline 使用 `options`；SourceRange 使用 `source-sheet-id` + `source-range`，两种模式必须且只能选一个。顶层 `colors` / `source-colors` 会被拒绝；Inline 颜色写在 `options[].color`，SourceRange 颜色写入暂不支持
+  - `set-dropdown` SourceRange 在已验证的重命名、引用前插入行/列、删除引用前行的场景会自动调整；已验证的 `move-dimension` 会使其变为 `invalid`。其他未覆盖删除/移动场景后先回读，仅 `invalid` 时重新选源写入
+  - `source-range` 按 `toolName` 解释：`set-dropdown` 中是下拉候选项来源；`range fill` / `range copy-to` / `range move-to` 中是待填充、复制或移动的数据源区域
   - 典型场景：先插入行列再写入数据、先清除再写入、批量合并+调整行高列宽
   - `group-dimension` 在 batch 中只适合默认展开分组；需要 `--group-state fold` 时请使用独立 `dws sheet group-dimension`
   - `table-put` 不支持放进 batch-update；结构化 table 请用独立 `dws sheet table-put`
diff --git a/skills/mono/references/products/sheet/sheet-dropdown.md b/skills/mono/references/products/sheet/sheet-dropdown.md
index 0b135aa4..05b5f0c0 100644
--- a/skills/mono/references/products/sheet/sheet-dropdown.md
+++ b/skills/mono/references/products/sheet/sheet-dropdown.md
@@ -7,6 +7,7 @@
 用户说"设置下拉列表/下拉选项/下拉菜单/添加下拉/配置下拉":
 - 设置下拉列表 → `set-dropdown`
 - 设置多选下拉 → `set-dropdown --multi-select`
+- 引用单元格区域作为候选项 → `set-dropdown --source-sheet-id ... --source-range ...`
 
 用户说"查看下拉列表/获取下拉配置/下拉列表有哪些选项":
 - 获取下拉列表配置 → `get-dropdown`
@@ -29,18 +30,25 @@ Example:
   dws sheet set-dropdown --node <NODE_ID> --sheet-id <SHEET_ID> --range "B2:B50" \
     --options '[{"value":"高","color":"#ff0000"},{"value":"中","color":"#ffaa00"},{"value":"低","color":"#00ff00"}]' \
     --multi-select
+
+  # 引用同一工作簿内另一工作表的区域作为候选项来源
+  dws sheet set-dropdown --node <NODE_ID> --sheet-id <TARGET_SHEET_ID> --range "C2:C100" \
+    --source-sheet-id <SOURCE_SHEET_ID> --source-range "T1:T3"
 Flags:
-      --node string         表格文档 ID 或 URL (必填)
-      --sheet-id string     工作表 ID 或名称 (必填)
-      --range string        目标单元格范围，A1 表示法，如 A2:A100 (必填)
-      --options string      下拉选项 JSON 数组 (必填)，如 '[{"value":"选项1","color":"#ff0000"}]'
-      --multi-select        是否允许多选（默认单选）
+      --node string              表格文档 ID 或 URL (必填)
+      --sheet-id string          工作表 ID 或名称 (必填)
+      --range string             目标单元格范围，A1 表示法，如 A2:A100 (必填)
+      --options string           Inline 下拉选项 JSON 数组，与 --source-range 二选一
+      --source-sheet-id string   SourceRange 来源工作表 ID，与 --source-range 同时指定
+      --source-range string      SourceRange 来源区域，与 --options 二选一；不带工作表前缀
+      --multi-select             是否允许多选（默认单选）
 ```
 
-在指定单元格范围内设置下拉列表。设置后用户可从预定义选项中选择值。
-- **用途**：为单元格配置下拉列表，支持自定义选项颜色和多选。
+在指定单元格范围内设置下拉列表。Inline 模式直接存储选项；SourceRange 模式引用同一工作簿内的来源区域，可跨工作表，并支持普通区域、整行和整列。
+- **用途**：为单元格配置静态选项或区域来源下拉，两种模式都支持多选；颜色仅 Inline 支持。
 - **场景**：规范数据输入，如状态选择（完成/进行中/待处理）、优先级（高/中/低）等。
-- **注意**：选项值不能包含英文逗号；如果目标范围已存在下拉列表，会被新配置覆盖。
+- **注意**：`--options` 与 `--source-range` 必须且只能指定一个。`--source-range` 只写 `T1:T3`、`T:T`、`1:3` 这类 A1 区域，来源工作表通过 `--source-sheet-id` 单独指定；不接受工作表前缀、公式或多区域。SourceRange 颜色写入暂不支持。
+- **结构操作行为**：已验证的工作表重命名、在引用前插入行/列、删除引用前的行会自动调整引用并保持 `valid`；已验证的 `move-dimension` 场景会使其变为 `invalid`。列删除、删除整个来源区域或来源工作表等场景未覆盖，不能预设结果；结构操作后先回读 `sourceRangeStatus`，仅在 `invalid` 时重新选择来源并写入。
 
 ### 获取下拉列表配置
 ```
@@ -55,10 +63,10 @@ Flags:
       --range string      查询范围，A1 表示法，如 A1:A100 (必填)
 ```
 
-查询指定范围内的下拉列表配置信息，包括选项值、颜色和是否多选。
+查询指定范围内的下拉列表配置信息。
 - **用途**：查看单元格已设置的下拉列表选项和配置。
 - **场景**：在修改下拉列表前先查询现有配置；确认下拉列表是否设置成功。
-- **返回**：`dataValidations` 数组，相同选项的单元格聚合为一组，每组包含 `conditionValues`（选项值）、`ranges`（覆盖范围）、`options`（含 `enableMultiSelect` 和 `colorValueMap`）。范围内无下拉列表时 `hasDropdown` 为 false。
+- **返回**：`dataValidations` 数组由底层按配置聚合。Inline 组返回 `sourceType:"inline"`、`conditionValues`、`ranges` 和 `options`；SourceRange 组始终返回 `sourceType:"sourceRange"`、`sourceRangeStatus:"valid"/"invalid"`、`enableMultiSelect` 和 `ranges`，仅在 `sourceRangeStatus:"valid"` 时返回 `sourceRange:{sheetId,a1Notation}`。`invalid` 时仍保留配置组，但省略 `sourceRange`，不得依赖旧坐标修复。SourceRange 不会展开候选值，因此不返回 `conditionValues`、`options` 或颜色。范围内无下拉列表时 `hasDropdown` 为 false。
 
 ### 删除下拉列表
 ```
@@ -81,14 +89,15 @@ Flags:
 
 | 操作 | 从返回中提取 | 用于 |
 |------|-------------|------|
-| `set-dropdown` | `range` 实际设置范围、`optionCount` 选项数量、`enableMultiSelect` 是否多选 | 确认下拉列表设置成功 |
-| `get-dropdown` | `hasDropdown` 是否存在下拉、`dataValidations` 下拉配置列表（含 `conditionValues`、`ranges`、`options`） | 查看已有下拉配置 |
+| `set-dropdown` | `range` 实际设置范围、`enableMultiSelect` 是否多选；仅 Inline 模式返回 `optionCount` | 确认下拉列表设置成功 |
+| `get-dropdown` | `hasDropdown`、`dataValidations`；按 `sourceType` 区分 Inline 与 SourceRange | 查看已有下拉配置 |
 | `delete-dropdown` | `range` 实际删除范围 | 确认下拉列表删除完成 |
 | `list` | 工作表的 `sheetId` | info / range read / range update / find 的 --sheet-id |
 
 ## 注意事项
 
 - ★ **`--sheet-id` 获取规范（强制）**：`sheetId` 未知时必须先通过 `dws sheet list --node <NODE_ID> --format json` 查询，禁止凭空编造（如臆测为 `Sheet1`、`sheet1`、`0`、`default` 等）
-- `set-dropdown` 在指定范围内设置下拉列表，`--options` 为 JSON 数组，每个元素包含 `value`（必填）和 `color`（可选，`#RRGGBB` 格式）。选项值不能包含英文逗号。`--multi-select` 启用多选模式。如果目标范围已存在下拉列表，会被新配置覆盖
-- `get-dropdown` 查询指定范围内的下拉列表配置，返回 `dataValidations` 数组，相同选项的单元格聚合为一组。无下拉列表时 `hasDropdown` 为 false
+- `set-dropdown` 的 Inline 模式使用 `--options`，每个元素包含 `value`（必填）和 `color`（可选，`#RRGGBB`）；SourceRange 模式使用 `--source-sheet-id` + `--source-range`。两种模式均可用 `--multi-select`，并会覆盖目标范围已有下拉
+- SourceRange 在已验证的重命名、引用前插入行/列、删除引用前行的场景会自动调整；已验证的 `move-dimension` 会使其变为 `invalid`。其他未覆盖删除/移动场景后先回读，仅 `invalid` 时重新选源写入；颜色写入暂不支持
+- `get-dropdown` 查询指定范围内的下拉配置，聚合发生在底层服务。SourceRange 即使无效也保留一组并以 `sourceRangeStatus:"invalid"` 表示，但省略 `sourceRange`，不回退展开候选值
 - `delete-dropdown` 删除指定范围内的下拉列表配置，单元格恢复为普通文本格式。已填写的值不会被清除。目标范围不存在下拉列表时操作仍返回成功
diff --git a/skills/mono/references/products/sheet/sheet-read-data.md b/skills/mono/references/products/sheet/sheet-read-data.md
index 0e2ae054..f8008d0e 100644
--- a/skills/mono/references/products/sheet/sheet-read-data.md
+++ b/skills/mono/references/products/sheet/sheet-read-data.md
@@ -13,7 +13,7 @@
 
 | 读取目的 | 推荐命令 | 说明 |
 |---------|---------|------|
-| 快速查看纯值、数据分析、大表分批读取 | `csv-get` | CSV 格式，token 消耗约为 JSON 的 1/3，内置 maxChars 防爆 |
+| 快速查看纯值、数据分析、大表分批读取 | `csv-get` | CSV 格式，token 消耗约为 JSON 的 1/3，内置 30,000 单元格和 maxChars 防爆 |
 | 按 table/dataframe 协议读取 | `table-get` | 返回 `columns` / `data` / `dtypes` / `formats`；默认首行为表头 |
 | 查看数据验证配置（下拉/复选框） | `range read` | 返回 per-cell 结构，含 dataValidation |
 | 查看单元格样式（背景色/字体/对齐等） | `range read` | 返回 per-cell 结构，含 cellStyles（仅显式设置的样式） |
@@ -52,7 +52,10 @@ Flags:
 - `csv` — CSV 文本，每逻辑行前加 `[row=N]` 前缀标注真实表格行号。行号一律从此前缀读取，禁止手算
 - `colIndices` — 列字母映射数组（如 `["A","B","C"]`）。定位列字母用 `colIndices[j]`，禁止手数逗号
 - `rowIndices` — 行号映射数组（如 `[1,2,3]`）
-- `hasMore` — 是否因 maxChars 截断。为 true 时需要调整 `--range` 继续分页读取
+- `hasMore` — 完整目标范围是否还有未返回数据。`true` 是部分成功，不能宣称已读完
+- `truncationReasons` — 部分返回原因。`max_cells` 表示命中单次 30,000 单元格上限；`max_chars` 表示命中 CSV 字符上限，两者可同时出现
+- `resolvedRange` — 仅在未传 `--range` 时返回，表示底层解析出的完整目标范围
+- `returnedRange` — 本次实际完整返回的精确范围；它不是服务端续读游标
 
 `csv-get` 不返回合并单元格结构。若 CSV 中出现合并区域的非左上角单元格为空，不能据此判断该区域"无内容"；需要先用 `dws sheet info --node <NODE_ID> --sheet-id <SHEET_ID> --format json` 读取 `mergedRanges`，再结合左上角单元格理解合并区域语义。
 
@@ -63,10 +66,15 @@ Flags:
 | `raw_value` | 原始值（如 1000、45808） | 数据处理、计算 |
 | `formula` | 公式文本（如 =SUM(A1:A10)），无公式时回退原始值 | 查看/复制公式 |
 
-**大表分批读取**：当 `hasMore=true` 或数据量很大时，按行窗口分批：
-- 先通过 `info` 获取 `nonEmptyRange.range`，或用 `nonEmptyRange.lastRow` / `nonEmptyRange.lastColumn` 确定 A1 边界
-- 分批读取：`--range "A1:J500"`、`--range "A501:J1000"` ……
-- 单次建议 ≤5000 单元格
+**大表分批读取**：`csv-get` 不会自动发起后续请求。当 `hasMore=true` 时：
+- 显式传了 `--range` 时，该范围是完整目标；未传时，以 `resolvedRange` 为完整目标
+- 根据 `returnedRange` 的结束行，从下一行开始显式构造新的 `--range`；例如目标是 `A1:A30001`、本次返回 `A1:A30000`，则继续读 `A30001:A30001`
+- `max_cells` 不能通过增大 `--max-chars` 解决；`max_chars` 且没有 `returnedRange` 时，增大 `--max-chars` 或缩小列范围
+- 上述完成度协议覆盖未传 `--range` 和显式有限矩形范围；`A:A`、`1:1` 等非有限范围可能不返回完成度字段，建议先用 `sheet info` 的 `nonEmptyRange` 换算成有限范围再读
+- 即使未命中硬上限，单次仍建议控制在 5000 单元格以内，减少超时和 token 消耗
+
+**工作簿过大失败与部分成功不同**：收到顶层
+`errorCode=forbidden.document.sizeOverLimit` 和对应 `errorMessage` 时，表示工作簿整体无法装载。应停止读取并提示创建更小副本或拆分工作簿；缩小 `--range` 不能解决这类错误。不要把它当成 `hasMore=true` 的可续读结果。
 
 ### 以 table/dataframe 协议读取结构化数据
 ```
@@ -130,6 +138,8 @@ Flags:
 ```
 
 **返回字段说明**：
+- `rowIndices` / `colIndices` — 本次返回的真实行号与列字母映射
+- `hasMore` / `truncationReasons` / `resolvedRange` / `returnedRange` — 语义与 `csv-get` 相同；`range read` 的 `truncationReasons` 主要为 `max_cells`
 - `cells` — 二维数组，第一维为行，第二维为列。每个元素为 per-cell 对象，字段如下：
 
 | 字段 | 类型 | 是否必有 | 说明 |
@@ -146,10 +156,16 @@ Flags:
 
 | type | 字段 | 说明 |
 |------|------|------|
-| `dropdown` | `options: [{value: string, color?: string}]` | 下拉选项列表 |
-| `dropdown` | `enableMultiSelect: boolean` | 是否允许多选 |
+| `dropdown` Inline | `sourceType: "inline"` | 静态选项模式 |
+| `dropdown` Inline | `options: [{value: string, color?: string}]` | 下拉选项列表 |
+| `dropdown` Inline | `enableMultiSelect: boolean` | 是否允许多选 |
+| `dropdown` SourceRange | `sourceType: "sourceRange"` | 区域来源模式 |
+| `dropdown` SourceRange | `sourceRange: {sheetId, a1Notation}` | 仅 `sourceRangeStatus:"valid"` 时返回的来源工作表和规范化 A1 区域；`invalid` 时省略 |
+| `dropdown` SourceRange | `sourceRangeStatus: "valid" / "invalid"` | 底层引用是否可解析；无效引用仍返回配置，但不返回 `sourceRange` |
 | `checkbox` | `checked: boolean` | 当前勾选状态 |
 
+SourceRange 模式不动态拉取或展开来源区域的候选值，因此不返回 `options` 或颜色。`invalid` 时只能依赖 `sourceRangeStatus` 判定需要重新选源，不得假定仍能从 `sourceRange` 取回旧坐标。
+
 **hyperlink 结构**：
 
 | type | 字段 | 说明 |
@@ -216,7 +232,7 @@ Flags:
 
 **公式校验建议**：写公式后不要只看写入返回结果。先用 `formula` 模式确认公式文本已落表，再运行 `formula-verify` 聚合扫描错误；关键业务数值继续用 `raw_value` 抽样对账。详见 [sheet-formula](./sheet-formula.md)。
 
-**超时处理建议**：读取大范围数据时若出现超时或响应过慢，请主动缩小 `--range` 查询范围，**建议单次读取的单元格数量控制在 5000 个以内**（例如 50 行 × 100 列、100 行 × 50 列）。对于大表可采用分页读取策略：
+**超时处理建议**：读取大范围数据时若出现超时或响应过慢，请主动缩小 `--range` 查询范围，**建议单次读取的单元格数量控制在 5000 个以内**（例如 50 行 × 100 列、100 行 × 50 列）。对于大表可采用分批读取策略，每次都必须检查 `hasMore`：
 - 先通过 `info` 获取 `nonEmptyRange.range`，或用 `nonEmptyRange.lastRow` / `nonEmptyRange.lastColumn` 确定 A1 边界
 - 按行分批读取，如 `A1:J500`、`A501:J1000`、`A1001:J1500` ……
 - 避免不传 `--range` 直接读取整个大工作表
@@ -232,7 +248,7 @@ dws sheet list --node <NODE_ID> --format json
 # 2. 查看工作表详情（行列数、最后非空位置、mergedRanges 等）
 dws sheet info --node <NODE_ID> --sheet-id <SHEET_ID> --format json
 
-# 3. 读取全部数据
+# 3. 读取数据（只有 hasMore=false 才代表目标范围已完整返回）
 dws sheet range read --node <NODE_ID> --sheet-id <SHEET_ID> --format json
 
 # 4. 读取指定区域
@@ -251,6 +267,8 @@ dws sheet range read --node <NODE_ID> --sheet-id <SHEET_ID> --range "A1:D10" --f
 - ★ **`--sheet-id` 获取规范（强制）**：`sheetId` 未知时必须先通过 `dws sheet list --node <NODE_ID> --format json` 查询真实的 `sheetId` / 工作表名称后再调用，禁止凭空编造（如臆测为 `Sheet1`、`sheet1`、`0`、`default` 等）；用户仅给出工作表名称时，也应通过 `list` 校验该名称是否存在，避免名称大小写或拼写不一致导致失败
 - `range read` 不传 `--range` 时默认读取整个工作表的全部非空数据
 - `range read` 的 `--range` 支持 `Sheet1!A1:D10` 格式直接指定工作表（此时忽略 `--sheet-id`）
+- ★ `csv-get` 和 `range read` 都必须检查 `hasMore`；`true` 时结合目标范围与 `returnedRange` 显式分批，CLI 不会自动续读
+- ★ `forbidden.document.sizeOverLimit` 是工作簿装载失败，不是 30,000 单元格部分成功；缩小 `--range` 不能修复
 - ★ `csv-get` / `range read` / `range get` 不返回合并单元格结构；查看合并范围必须用 `sheet info` 的 `mergedRanges`
 - `table-get` 返回 table/dataframe 结构，适合需要 `columns` / `data` / `dtypes` / `formats` 的场景；它不返回 per-cell 元数据，也不替代 `range read`
 - ★ 大整数和长数字标识符回读校验：精确 ID 应是字符串 / `object`；若看到 `float64`，说明它已经按数值路径处理，超过 `9007199254740991` 时可能已丢精度
diff --git a/skills/mono/references/products/sheet/sheet-write-data.md b/skills/mono/references/products/sheet/sheet-write-data.md
index eaa5e695..9050fb43 100644
--- a/skills/mono/references/products/sheet/sheet-write-data.md
+++ b/skills/mono/references/products/sheet/sheet-write-data.md
@@ -386,9 +386,14 @@ Flags:
 **dropdown（下拉列表）**：
 ```json
 { "type": "text", "text": "High", "dataValidation": { "type": "dropdown", "options": [{"value":"High","color":"#00ff00"},{"value":"Low","color":"#ff0000"}], "enableMultiSelect": false } }
+{ "dataValidation": { "type": "dropdown", "sourceRange": {"sheetId":"SOURCE_SHEET_ID","a1Notation":"T1:T3"}, "enableMultiSelect": false } }
 ```
-- `options`：必填，`[{value, color?}]` 数组
+- `options` 与 `sourceRange` 必须且只能提供一个
+- Inline：`options` 为非空 `[{value, color?}]` 数组
+- SourceRange：`sourceRange` 为 `{sheetId,a1Notation}`；可引用同一工作簿内其他工作表，支持普通区域、整行和整列。`a1Notation` 不带工作表前缀，不接受公式或多区域；颜色写入暂不支持
 - `enableMultiSelect`：可选，是否多选，默认 false
+- SourceRange 在已验证的工作表重命名、引用前插入行/列、删除引用前行的场景会自动调整；已验证的 `move-dimension` 会使其变为 `invalid`。其他未覆盖删除/移动场景后先回读 `sourceRangeStatus`，仅 `invalid` 时重新选源写入
+- 复合 cell 写入中，如果值或样式已成功落盘但 SourceRange 校验失败，服务端可返回 `success:true` 并通过 `message` 说明下拉未创建；必须检查 `message` 并按需回读
 
 **checkbox（复选框）**：
 ```json
diff --git a/skills/multi/dingtalk-misc/references/sheet/sheet-batch-operations.md b/skills/multi/dingtalk-misc/references/sheet/sheet-batch-operations.md
index 8d922bec..66fafc90 100644
--- a/skills/multi/dingtalk-misc/references/sheet/sheet-batch-operations.md
+++ b/skills/multi/dingtalk-misc/references/sheet/sheet-batch-operations.md
@@ -83,7 +83,8 @@ Example:
     {"toolName":"range update","input":{"sheet-id":"Sheet1","range":"A1","values":[[{"type":"text","text":"hello"}]]}},
     {"toolName":"merge-cells","input":{"sheet-id":"Sheet1","range":"A1:B1","merge-type":"mergeAll"}},
     {"toolName":"update-dimension","input":{"sheet-id":"Sheet1","dimension":"ROWS","start-index":"1","length":1,"pixel-size":40}},
-    {"toolName":"group-dimension","input":{"sheet-id":"Sheet1","range":"3:7","group-state":"expand"}}
+    {"toolName":"group-dimension","input":{"sheet-id":"Sheet1","range":"3:7","group-state":"expand"}},
+    {"toolName":"set-dropdown","input":{"sheet-id":"Sheet1","range":"C2:C100","source-sheet-id":"SourceSheet","source-range":"T1:T3"}}
   ]'
   dws sheet batch-update --node <NODE_ID> --continue-on-error --operations '[...]'
 Flags:
@@ -101,6 +102,9 @@ Notes:
   - operations 最多 20 条
   - 当需要对多个区域执行相同清除时，优先使用 `range batch-clear`（更简洁）
   - `csv-put` 子操作与独立命令语义一致：CSV 字段值以 `=` 开头时按公式解析；前加单引号时写入以 `=` 开头的字面文本
+  - `set-dropdown` 的 `input` 中，Inline 使用 `options`；SourceRange 使用 `source-sheet-id` + `source-range`，两种模式必须且只能选一个。顶层 `colors` / `source-colors` 会被拒绝；Inline 颜色写在 `options[].color`，SourceRange 颜色写入暂不支持
+  - `set-dropdown` SourceRange 在已验证的重命名、引用前插入行/列、删除引用前行的场景会自动调整；已验证的 `move-dimension` 会使其变为 `invalid`。其他未覆盖删除/移动场景后先回读，仅 `invalid` 时重新选源写入
+  - `source-range` 按 `toolName` 解释：`set-dropdown` 中是下拉候选项来源；`range fill` / `range copy-to` / `range move-to` 中是待填充、复制或移动的数据源区域
   - 典型场景：先插入行列再写入数据、先清除再写入、批量合并+调整行高列宽
   - `group-dimension` 在 batch 中只适合默认展开分组；需要 `--group-state fold` 时请使用独立 `dws sheet group-dimension`
   - `table-put` 不支持放进 batch-update；结构化 table 请用独立 `dws sheet table-put`
diff --git a/skills/multi/dingtalk-misc/references/sheet/sheet-dropdown.md b/skills/multi/dingtalk-misc/references/sheet/sheet-dropdown.md
index 0b135aa4..05b5f0c0 100644
--- a/skills/multi/dingtalk-misc/references/sheet/sheet-dropdown.md
+++ b/skills/multi/dingtalk-misc/references/sheet/sheet-dropdown.md
@@ -7,6 +7,7 @@
 用户说"设置下拉列表/下拉选项/下拉菜单/添加下拉/配置下拉":
 - 设置下拉列表 → `set-dropdown`
 - 设置多选下拉 → `set-dropdown --multi-select`
+- 引用单元格区域作为候选项 → `set-dropdown --source-sheet-id ... --source-range ...`
 
 用户说"查看下拉列表/获取下拉配置/下拉列表有哪些选项":
 - 获取下拉列表配置 → `get-dropdown`
@@ -29,18 +30,25 @@ Example:
   dws sheet set-dropdown --node <NODE_ID> --sheet-id <SHEET_ID> --range "B2:B50" \
     --options '[{"value":"高","color":"#ff0000"},{"value":"中","color":"#ffaa00"},{"value":"低","color":"#00ff00"}]' \
     --multi-select
+
+  # 引用同一工作簿内另一工作表的区域作为候选项来源
+  dws sheet set-dropdown --node <NODE_ID> --sheet-id <TARGET_SHEET_ID> --range "C2:C100" \
+    --source-sheet-id <SOURCE_SHEET_ID> --source-range "T1:T3"
 Flags:
-      --node string         表格文档 ID 或 URL (必填)
-      --sheet-id string     工作表 ID 或名称 (必填)
-      --range string        目标单元格范围，A1 表示法，如 A2:A100 (必填)
-      --options string      下拉选项 JSON 数组 (必填)，如 '[{"value":"选项1","color":"#ff0000"}]'
-      --multi-select        是否允许多选（默认单选）
+      --node string              表格文档 ID 或 URL (必填)
+      --sheet-id string          工作表 ID 或名称 (必填)
+      --range string             目标单元格范围，A1 表示法，如 A2:A100 (必填)
+      --options string           Inline 下拉选项 JSON 数组，与 --source-range 二选一
+      --source-sheet-id string   SourceRange 来源工作表 ID，与 --source-range 同时指定
+      --source-range string      SourceRange 来源区域，与 --options 二选一；不带工作表前缀
+      --multi-select             是否允许多选（默认单选）
 ```
 
-在指定单元格范围内设置下拉列表。设置后用户可从预定义选项中选择值。
-- **用途**：为单元格配置下拉列表，支持自定义选项颜色和多选。
+在指定单元格范围内设置下拉列表。Inline 模式直接存储选项；SourceRange 模式引用同一工作簿内的来源区域，可跨工作表，并支持普通区域、整行和整列。
+- **用途**：为单元格配置静态选项或区域来源下拉，两种模式都支持多选；颜色仅 Inline 支持。
 - **场景**：规范数据输入，如状态选择（完成/进行中/待处理）、优先级（高/中/低）等。
-- **注意**：选项值不能包含英文逗号；如果目标范围已存在下拉列表，会被新配置覆盖。
+- **注意**：`--options` 与 `--source-range` 必须且只能指定一个。`--source-range` 只写 `T1:T3`、`T:T`、`1:3` 这类 A1 区域，来源工作表通过 `--source-sheet-id` 单独指定；不接受工作表前缀、公式或多区域。SourceRange 颜色写入暂不支持。
+- **结构操作行为**：已验证的工作表重命名、在引用前插入行/列、删除引用前的行会自动调整引用并保持 `valid`；已验证的 `move-dimension` 场景会使其变为 `invalid`。列删除、删除整个来源区域或来源工作表等场景未覆盖，不能预设结果；结构操作后先回读 `sourceRangeStatus`，仅在 `invalid` 时重新选择来源并写入。
 
 ### 获取下拉列表配置
 ```
@@ -55,10 +63,10 @@ Flags:
       --range string      查询范围，A1 表示法，如 A1:A100 (必填)
 ```
 
-查询指定范围内的下拉列表配置信息，包括选项值、颜色和是否多选。
+查询指定范围内的下拉列表配置信息。
 - **用途**：查看单元格已设置的下拉列表选项和配置。
 - **场景**：在修改下拉列表前先查询现有配置；确认下拉列表是否设置成功。
-- **返回**：`dataValidations` 数组，相同选项的单元格聚合为一组，每组包含 `conditionValues`（选项值）、`ranges`（覆盖范围）、`options`（含 `enableMultiSelect` 和 `colorValueMap`）。范围内无下拉列表时 `hasDropdown` 为 false。
+- **返回**：`dataValidations` 数组由底层按配置聚合。Inline 组返回 `sourceType:"inline"`、`conditionValues`、`ranges` 和 `options`；SourceRange 组始终返回 `sourceType:"sourceRange"`、`sourceRangeStatus:"valid"/"invalid"`、`enableMultiSelect` 和 `ranges`，仅在 `sourceRangeStatus:"valid"` 时返回 `sourceRange:{sheetId,a1Notation}`。`invalid` 时仍保留配置组，但省略 `sourceRange`，不得依赖旧坐标修复。SourceRange 不会展开候选值，因此不返回 `conditionValues`、`options` 或颜色。范围内无下拉列表时 `hasDropdown` 为 false。
 
 ### 删除下拉列表
 ```
@@ -81,14 +89,15 @@ Flags:
 
 | 操作 | 从返回中提取 | 用于 |
 |------|-------------|------|
-| `set-dropdown` | `range` 实际设置范围、`optionCount` 选项数量、`enableMultiSelect` 是否多选 | 确认下拉列表设置成功 |
-| `get-dropdown` | `hasDropdown` 是否存在下拉、`dataValidations` 下拉配置列表（含 `conditionValues`、`ranges`、`options`） | 查看已有下拉配置 |
+| `set-dropdown` | `range` 实际设置范围、`enableMultiSelect` 是否多选；仅 Inline 模式返回 `optionCount` | 确认下拉列表设置成功 |
+| `get-dropdown` | `hasDropdown`、`dataValidations`；按 `sourceType` 区分 Inline 与 SourceRange | 查看已有下拉配置 |
 | `delete-dropdown` | `range` 实际删除范围 | 确认下拉列表删除完成 |
 | `list` | 工作表的 `sheetId` | info / range read / range update / find 的 --sheet-id |
 
 ## 注意事项
 
 - ★ **`--sheet-id` 获取规范（强制）**：`sheetId` 未知时必须先通过 `dws sheet list --node <NODE_ID> --format json` 查询，禁止凭空编造（如臆测为 `Sheet1`、`sheet1`、`0`、`default` 等）
-- `set-dropdown` 在指定范围内设置下拉列表，`--options` 为 JSON 数组，每个元素包含 `value`（必填）和 `color`（可选，`#RRGGBB` 格式）。选项值不能包含英文逗号。`--multi-select` 启用多选模式。如果目标范围已存在下拉列表，会被新配置覆盖
-- `get-dropdown` 查询指定范围内的下拉列表配置，返回 `dataValidations` 数组，相同选项的单元格聚合为一组。无下拉列表时 `hasDropdown` 为 false
+- `set-dropdown` 的 Inline 模式使用 `--options`，每个元素包含 `value`（必填）和 `color`（可选，`#RRGGBB`）；SourceRange 模式使用 `--source-sheet-id` + `--source-range`。两种模式均可用 `--multi-select`，并会覆盖目标范围已有下拉
+- SourceRange 在已验证的重命名、引用前插入行/列、删除引用前行的场景会自动调整；已验证的 `move-dimension` 会使其变为 `invalid`。其他未覆盖删除/移动场景后先回读，仅 `invalid` 时重新选源写入；颜色写入暂不支持
+- `get-dropdown` 查询指定范围内的下拉配置，聚合发生在底层服务。SourceRange 即使无效也保留一组并以 `sourceRangeStatus:"invalid"` 表示，但省略 `sourceRange`，不回退展开候选值
 - `delete-dropdown` 删除指定范围内的下拉列表配置，单元格恢复为普通文本格式。已填写的值不会被清除。目标范围不存在下拉列表时操作仍返回成功
diff --git a/skills/multi/dingtalk-misc/references/sheet/sheet-read-data.md b/skills/multi/dingtalk-misc/references/sheet/sheet-read-data.md
index 0e2ae054..f8008d0e 100644
--- a/skills/multi/dingtalk-misc/references/sheet/sheet-read-data.md
+++ b/skills/multi/dingtalk-misc/references/sheet/sheet-read-data.md
@@ -13,7 +13,7 @@
 
 | 读取目的 | 推荐命令 | 说明 |
 |---------|---------|------|
-| 快速查看纯值、数据分析、大表分批读取 | `csv-get` | CSV 格式，token 消耗约为 JSON 的 1/3，内置 maxChars 防爆 |
+| 快速查看纯值、数据分析、大表分批读取 | `csv-get` | CSV 格式，token 消耗约为 JSON 的 1/3，内置 30,000 单元格和 maxChars 防爆 |
 | 按 table/dataframe 协议读取 | `table-get` | 返回 `columns` / `data` / `dtypes` / `formats`；默认首行为表头 |
 | 查看数据验证配置（下拉/复选框） | `range read` | 返回 per-cell 结构，含 dataValidation |
 | 查看单元格样式（背景色/字体/对齐等） | `range read` | 返回 per-cell 结构，含 cellStyles（仅显式设置的样式） |
@@ -52,7 +52,10 @@ Flags:
 - `csv` — CSV 文本，每逻辑行前加 `[row=N]` 前缀标注真实表格行号。行号一律从此前缀读取，禁止手算
 - `colIndices` — 列字母映射数组（如 `["A","B","C"]`）。定位列字母用 `colIndices[j]`，禁止手数逗号
 - `rowIndices` — 行号映射数组（如 `[1,2,3]`）
-- `hasMore` — 是否因 maxChars 截断。为 true 时需要调整 `--range` 继续分页读取
+- `hasMore` — 完整目标范围是否还有未返回数据。`true` 是部分成功，不能宣称已读完
+- `truncationReasons` — 部分返回原因。`max_cells` 表示命中单次 30,000 单元格上限；`max_chars` 表示命中 CSV 字符上限，两者可同时出现
+- `resolvedRange` — 仅在未传 `--range` 时返回，表示底层解析出的完整目标范围
+- `returnedRange` — 本次实际完整返回的精确范围；它不是服务端续读游标
 
 `csv-get` 不返回合并单元格结构。若 CSV 中出现合并区域的非左上角单元格为空，不能据此判断该区域"无内容"；需要先用 `dws sheet info --node <NODE_ID> --sheet-id <SHEET_ID> --format json` 读取 `mergedRanges`，再结合左上角单元格理解合并区域语义。
 
@@ -63,10 +66,15 @@ Flags:
 | `raw_value` | 原始值（如 1000、45808） | 数据处理、计算 |
 | `formula` | 公式文本（如 =SUM(A1:A10)），无公式时回退原始值 | 查看/复制公式 |
 
-**大表分批读取**：当 `hasMore=true` 或数据量很大时，按行窗口分批：
-- 先通过 `info` 获取 `nonEmptyRange.range`，或用 `nonEmptyRange.lastRow` / `nonEmptyRange.lastColumn` 确定 A1 边界
-- 分批读取：`--range "A1:J500"`、`--range "A501:J1000"` ……
-- 单次建议 ≤5000 单元格
+**大表分批读取**：`csv-get` 不会自动发起后续请求。当 `hasMore=true` 时：
+- 显式传了 `--range` 时，该范围是完整目标；未传时，以 `resolvedRange` 为完整目标
+- 根据 `returnedRange` 的结束行，从下一行开始显式构造新的 `--range`；例如目标是 `A1:A30001`、本次返回 `A1:A30000`，则继续读 `A30001:A30001`
+- `max_cells` 不能通过增大 `--max-chars` 解决；`max_chars` 且没有 `returnedRange` 时，增大 `--max-chars` 或缩小列范围
+- 上述完成度协议覆盖未传 `--range` 和显式有限矩形范围；`A:A`、`1:1` 等非有限范围可能不返回完成度字段，建议先用 `sheet info` 的 `nonEmptyRange` 换算成有限范围再读
+- 即使未命中硬上限，单次仍建议控制在 5000 单元格以内，减少超时和 token 消耗
+
+**工作簿过大失败与部分成功不同**：收到顶层
+`errorCode=forbidden.document.sizeOverLimit` 和对应 `errorMessage` 时，表示工作簿整体无法装载。应停止读取并提示创建更小副本或拆分工作簿；缩小 `--range` 不能解决这类错误。不要把它当成 `hasMore=true` 的可续读结果。
 
 ### 以 table/dataframe 协议读取结构化数据
 ```
@@ -130,6 +138,8 @@ Flags:
 ```
 
 **返回字段说明**：
+- `rowIndices` / `colIndices` — 本次返回的真实行号与列字母映射
+- `hasMore` / `truncationReasons` / `resolvedRange` / `returnedRange` — 语义与 `csv-get` 相同；`range read` 的 `truncationReasons` 主要为 `max_cells`
 - `cells` — 二维数组，第一维为行，第二维为列。每个元素为 per-cell 对象，字段如下：
 
 | 字段 | 类型 | 是否必有 | 说明 |
@@ -146,10 +156,16 @@ Flags:
 
 | type | 字段 | 说明 |
 |------|------|------|
-| `dropdown` | `options: [{value: string, color?: string}]` | 下拉选项列表 |
-| `dropdown` | `enableMultiSelect: boolean` | 是否允许多选 |
+| `dropdown` Inline | `sourceType: "inline"` | 静态选项模式 |
+| `dropdown` Inline | `options: [{value: string, color?: string}]` | 下拉选项列表 |
+| `dropdown` Inline | `enableMultiSelect: boolean` | 是否允许多选 |
+| `dropdown` SourceRange | `sourceType: "sourceRange"` | 区域来源模式 |
+| `dropdown` SourceRange | `sourceRange: {sheetId, a1Notation}` | 仅 `sourceRangeStatus:"valid"` 时返回的来源工作表和规范化 A1 区域；`invalid` 时省略 |
+| `dropdown` SourceRange | `sourceRangeStatus: "valid" / "invalid"` | 底层引用是否可解析；无效引用仍返回配置，但不返回 `sourceRange` |
 | `checkbox` | `checked: boolean` | 当前勾选状态 |
 
+SourceRange 模式不动态拉取或展开来源区域的候选值，因此不返回 `options` 或颜色。`invalid` 时只能依赖 `sourceRangeStatus` 判定需要重新选源，不得假定仍能从 `sourceRange` 取回旧坐标。
+
 **hyperlink 结构**：
 
 | type | 字段 | 说明 |
@@ -216,7 +232,7 @@ Flags:
 
 **公式校验建议**：写公式后不要只看写入返回结果。先用 `formula` 模式确认公式文本已落表，再运行 `formula-verify` 聚合扫描错误；关键业务数值继续用 `raw_value` 抽样对账。详见 [sheet-formula](./sheet-formula.md)。
 
-**超时处理建议**：读取大范围数据时若出现超时或响应过慢，请主动缩小 `--range` 查询范围，**建议单次读取的单元格数量控制在 5000 个以内**（例如 50 行 × 100 列、100 行 × 50 列）。对于大表可采用分页读取策略：
+**超时处理建议**：读取大范围数据时若出现超时或响应过慢，请主动缩小 `--range` 查询范围，**建议单次读取的单元格数量控制在 5000 个以内**（例如 50 行 × 100 列、100 行 × 50 列）。对于大表可采用分批读取策略，每次都必须检查 `hasMore`：
 - 先通过 `info` 获取 `nonEmptyRange.range`，或用 `nonEmptyRange.lastRow` / `nonEmptyRange.lastColumn` 确定 A1 边界
 - 按行分批读取，如 `A1:J500`、`A501:J1000`、`A1001:J1500` ……
 - 避免不传 `--range` 直接读取整个大工作表
@@ -232,7 +248,7 @@ dws sheet list --node <NODE_ID> --format json
 # 2. 查看工作表详情（行列数、最后非空位置、mergedRanges 等）
 dws sheet info --node <NODE_ID> --sheet-id <SHEET_ID> --format json
 
-# 3. 读取全部数据
+# 3. 读取数据（只有 hasMore=false 才代表目标范围已完整返回）
 dws sheet range read --node <NODE_ID> --sheet-id <SHEET_ID> --format json
 
 # 4. 读取指定区域
@@ -251,6 +267,8 @@ dws sheet range read --node <NODE_ID> --sheet-id <SHEET_ID> --range "A1:D10" --f
 - ★ **`--sheet-id` 获取规范（强制）**：`sheetId` 未知时必须先通过 `dws sheet list --node <NODE_ID> --format json` 查询真实的 `sheetId` / 工作表名称后再调用，禁止凭空编造（如臆测为 `Sheet1`、`sheet1`、`0`、`default` 等）；用户仅给出工作表名称时，也应通过 `list` 校验该名称是否存在，避免名称大小写或拼写不一致导致失败
 - `range read` 不传 `--range` 时默认读取整个工作表的全部非空数据
 - `range read` 的 `--range` 支持 `Sheet1!A1:D10` 格式直接指定工作表（此时忽略 `--sheet-id`）
+- ★ `csv-get` 和 `range read` 都必须检查 `hasMore`；`true` 时结合目标范围与 `returnedRange` 显式分批，CLI 不会自动续读
+- ★ `forbidden.document.sizeOverLimit` 是工作簿装载失败，不是 30,000 单元格部分成功；缩小 `--range` 不能修复
 - ★ `csv-get` / `range read` / `range get` 不返回合并单元格结构；查看合并范围必须用 `sheet info` 的 `mergedRanges`
 - `table-get` 返回 table/dataframe 结构，适合需要 `columns` / `data` / `dtypes` / `formats` 的场景；它不返回 per-cell 元数据，也不替代 `range read`
 - ★ 大整数和长数字标识符回读校验：精确 ID 应是字符串 / `object`；若看到 `float64`，说明它已经按数值路径处理，超过 `9007199254740991` 时可能已丢精度
diff --git a/skills/multi/dingtalk-misc/references/sheet/sheet-write-data.md b/skills/multi/dingtalk-misc/references/sheet/sheet-write-data.md
index eaa5e695..9050fb43 100644
--- a/skills/multi/dingtalk-misc/references/sheet/sheet-write-data.md
+++ b/skills/multi/dingtalk-misc/references/sheet/sheet-write-data.md
@@ -386,9 +386,14 @@ Flags:
 **dropdown（下拉列表）**：
 ```json
 { "type": "text", "text": "High", "dataValidation": { "type": "dropdown", "options": [{"value":"High","color":"#00ff00"},{"value":"Low","color":"#ff0000"}], "enableMultiSelect": false } }
+{ "dataValidation": { "type": "dropdown", "sourceRange": {"sheetId":"SOURCE_SHEET_ID","a1Notation":"T1:T3"}, "enableMultiSelect": false } }
 ```
-- `options`：必填，`[{value, color?}]` 数组
+- `options` 与 `sourceRange` 必须且只能提供一个
+- Inline：`options` 为非空 `[{value, color?}]` 数组
+- SourceRange：`sourceRange` 为 `{sheetId,a1Notation}`；可引用同一工作簿内其他工作表，支持普通区域、整行和整列。`a1Notation` 不带工作表前缀，不接受公式或多区域；颜色写入暂不支持
 - `enableMultiSelect`：可选，是否多选，默认 false
+- SourceRange 在已验证的工作表重命名、引用前插入行/列、删除引用前行的场景会自动调整；已验证的 `move-dimension` 会使其变为 `invalid`。其他未覆盖删除/移动场景后先回读 `sourceRangeStatus`，仅 `invalid` 时重新选源写入
+- 复合 cell 写入中，如果值或样式已成功落盘但 SourceRange 校验失败，服务端可返回 `success:true` 并通过 `message` 说明下拉未创建；必须检查 `message` 并按需回读
 
 **checkbox（复选框）**：
 ```json
diff --git a/test/scripts/changelog_pr_gate_test.go b/test/scripts/changelog_pr_gate_test.go
index 849d9191..5d8ef72b 100644
--- a/test/scripts/changelog_pr_gate_test.go
+++ b/test/scripts/changelog_pr_gate_test.go
@@ -1157,7 +1157,7 @@ func TestChangelogPRFastPathWorkflowContract(t *testing.T) {
 		`test "$(git rev-parse HEAD^1)" = "$PR_BASE_SHA"`,
 		`test "$(git rev-parse HEAD^2)" = "$PR_HEAD_SHA"`,
 		`echo "TEST_HEAD_REF=$(git rev-parse HEAD)"`,
-		`list-shard "$TEST_SHARD" "$TEST_BASE_REF" "$TEST_HEAD_REF"`,
+		`list-shard "$package_shard" "$TEST_BASE_REF" "$TEST_HEAD_REF"`,
 		"needs.lint.outputs.full_suite != 'true'",
 		`name: "Test (race: ${{ matrix.shard }})"`,
 		"name: Test (workflow and release contracts)",
@@ -1219,16 +1219,20 @@ func TestChangelogPRFastPathWorkflowContract(t *testing.T) {
 	}
 	// The focused path fans the impacted set across the same shards as test-race
 	// and runs each shard the way test-race runs it, so no single job carries
-	// internal/app together with its reverse dependencies. internal/app keeps its
-	// package-level headroom through the process-isolating helper instead of one
-	// long -timeout, which is strictly stronger: every process releases the
-	// framework registries it populated. release-scripts is asserted because its
-	// dedicated job only runs at full-suite or release-sensitive scope, so losing
-	// it here would silently stop testing test/scripts changes.
+	// internal/app together with its reverse dependencies. internal/app is split
+	// further into one shard per bounded partition, which keeps its package-level
+	// headroom through the process-isolating helper instead of one long -timeout
+	// and is strictly stronger than a single app job: every partition process
+	// releases the framework registries it populated, and the partitions run
+	// concurrently rather than end to end. Each partition shard still selects the
+	// same single internal/app package, so the impacted-package query maps the
+	// shard name back to app. release-scripts is asserted because its dedicated
+	// job only runs at full-suite or release-sensitive scope, so losing it here
+	// would silently stop testing test/scripts changes.
 	for _, want := range []string{
-		`if [ "$TEST_SHARD" = "app" ]; then`,
+		`app-*) package_shard=app ;;`,
 		`test "${#packages[@]}" -eq 1`,
-		`./scripts/ci/run-app-race-tests.sh run "${packages[0]}"`,
+		`./scripts/ci/run-app-race-tests.sh run "${packages[0]}" "${TEST_SHARD#app-}"`,
 		`if [ "$TEST_SHARD" = "release-scripts" ]; then`,
 		`go test -v -count=1 -timeout=10m "${packages[@]}"`,
 		"timeout_budget=12m",
@@ -1249,14 +1253,16 @@ func TestChangelogPRFastPathWorkflowContract(t *testing.T) {
 		t.Fatal("Code Admission workflow missing race test job boundaries")
 	}
 	raceJob := admission[raceStart:raceEnd]
-	// The app shard uses independently bounded test processes so process-global
-	// command registries are released before the Schema assembly peak. Other full
-	// race shards retain the dynamic package timeout: default/floor 12m, with
-	// cli/smoke raised to 15m on slower hosted runners.
+	// internal/app is carried by one shard per bounded partition, so process-global
+	// command registries are released with each partition process and the Schema
+	// assembly peak no longer sits in front of the other partitions. Each
+	// partition shard resolves back to the same single internal/app package.
+	// Other full race shards retain the dynamic package timeout: default/floor
+	// 12m, with cli/smoke raised to 15m on slower hosted runners.
 	for _, want := range []string{
-		`if [ "$TEST_SHARD" = "app" ]; then`,
+		`app-*) package_shard=app ;;`,
 		`test "${#packages[@]}" -eq 1`,
-		`./scripts/ci/run-app-race-tests.sh run "${packages[0]}"`,
+		`./scripts/ci/run-app-race-tests.sh run "${packages[0]}" "${TEST_SHARD#app-}"`,
 		"timeout_budget=12m",
 		`if [ "$TEST_SHARD" = "cli" ] ||`,
 		`[ "$TEST_SHARD" = "smoke" ]; then`,
diff --git a/test/scripts/ci_test_package_plan_test.go b/test/scripts/ci_test_package_plan_test.go
index 0d7944f2..b8250c19 100644
--- a/test/scripts/ci_test_package_plan_test.go
+++ b/test/scripts/ci_test_package_plan_test.go
@@ -99,6 +99,74 @@ func TestCIAppRacePartitionsCoverTopLevelTestsExactlyOnce(t *testing.T) {
 	}
 }
 
+// TestCIAppRacePartitionMatrixMatchesHelper pins the workflow's app partition
+// shards to the partition set the helper actually runs. The partitions are
+// separate CI jobs now, so the helper's own "covered exactly once" check can no
+// longer prove the whole package ran: a partition the helper knows about but no
+// matrix shard dispatches would silently stop running while every job stays
+// green. Both directions are asserted so a stale matrix shard fails too.
+func TestCIAppRacePartitionMatrixMatchesHelper(t *testing.T) {
+	root := testPackagePlanRoot(t)
+	script := filepath.Join(root, "scripts", "ci", "run-app-race-tests.sh")
+	cmd := exec.Command("sh", script, "list-partitions")
+	cmd.Dir = root
+	output, err := cmd.CombinedOutput()
+	if err != nil {
+		t.Fatalf("%s list-partitions failed: %v\n%s", script, err, output)
+	}
+	partitions := strings.Fields(string(output))
+	if len(partitions) == 0 {
+		t.Fatalf("list-partitions returned no partitions: %q", output)
+	}
+
+	workflow, err := os.ReadFile(filepath.Join(root, ".github", "workflows", "ci.yml"))
+	if err != nil {
+		t.Fatalf("read ci.yml: %v", err)
+	}
+	admission := string(workflow)
+
+	for _, job := range []struct {
+		name      string
+		startMark string
+		endMark   string
+	}{
+		{"test-focused", "\n  test-focused:\n", "\n  test-race:\n"},
+		{"test-race", "\n  test-race:\n", "\n  test-release-scripts:\n"},
+	} {
+		start := strings.Index(admission, job.startMark)
+		end := strings.Index(admission, job.endMark)
+		if start < 0 || end <= start {
+			t.Fatalf("ci.yml is missing %s job boundaries", job.name)
+		}
+		body := admission[start:end]
+
+		for _, partition := range partitions {
+			want := "- app-" + partition
+			if !strings.Contains(body, want) {
+				t.Errorf("%s matrix is missing shard %q for a partition the helper runs", job.name, want)
+			}
+		}
+
+		for _, line := range strings.Split(body, "\n") {
+			shard := strings.TrimSpace(line)
+			if !strings.HasPrefix(shard, "- app-") {
+				continue
+			}
+			name := strings.TrimPrefix(shard, "- app-")
+			matched := false
+			for _, partition := range partitions {
+				if partition == name {
+					matched = true
+					break
+				}
+			}
+			if !matched {
+				t.Errorf("%s matrix shard %q has no matching helper partition", job.name, shard)
+			}
+		}
+	}
+}
+
 func TestCITestPackagePlanFailsClosedWhenGoListFails(t *testing.T) {
 	root := testPackagePlanRoot(t)
 	fakeBin := t.TempDir()