feat: add im-markdown output for doc fetch (#1550)

* feat: add docs im-markdown fetch format * refactor: tune docs im-markdown conversion * test: expand docs im-markdown conversion coverage * refactor: simplify docs im-markdown handlers * test: cover docs im-markdown edge cases * fix: expand doc im markdown tag downgrades * fix: preserve blockquote paragraph breaks * fix: handle im markdown nested tables and urls * docs: document im markdown skill usage * test: cover doc im markdown fetch * test: strengthen doc fetch error coverage * fix: fetch doc skill typo
2026-07-03 14:02:43 +08:00 · 2026-06-25 11:49:58 +08:00
parent 644c3c77dd
commit 1c92ed8841
8 changed files with 2706 additions and 31 deletions
--- a/shortcuts/doc/docs_fetch_im_markdown.go
+++ b/shortcuts/doc/docs_fetch_im_markdown.go
@@ -0,0 +1,861 @@
+// Copyright (c) 2026 Lark Technologies Pte. Ltd.
+// SPDX-License-Identifier: MIT
+
+package doc
+
+import (
+	"fmt"
+	"html"
+	"net/url"
+	"regexp"
+	"strings"
+	"unicode/utf8"
+)
+
+type imMarkdownContext struct {
+	baseURL         string
+	blockquoteDepth int
+}
+
+type imMarkdownHandleFunc func(segment, inner string, attrs map[string]string, imCtx imMarkdownContext) string
+
+type imMarkdownTagHandler struct {
+	closeRE *regexp.Regexp
+	handle  imMarkdownHandleFunc
+}
+
+func registerIMMarkdownHandler(tag string, handle imMarkdownHandleFunc) {
+	imMarkdownHandlers[tag] = imMarkdownTagHandler{
+		closeRE: regexp.MustCompile(`(?is)<(/?)` + regexp.QuoteMeta(tag) + `(?:\s[^<>]*?)?\s*/?>`),
+		handle:  handle,
+	}
+}
+
+var (
+	imMarkdownTagStartRE  = regexp.MustCompile(`(?s)<([A-Za-z][A-Za-z0-9:_-]*)(?:\s[^<>]*?)?\s*/?>`)
+	imMarkdownAttrRE      = regexp.MustCompile(`([A-Za-z_:][A-Za-z0-9_:.-]*)\s*=\s*(?:"([^"]*)"|'([^']*)')`)
+	imMarkdownRowTagRE    = regexp.MustCompile(`(?is)<(/?)tr\b[^>]*?\s*/?>`)
+	imMarkdownCellTagRE   = regexp.MustCompile(`(?is)<(/?)t[dh]\b[^>]*?\s*/?>`)
+	imMarkdownCellBreakRE = regexp.MustCompile(`(?i)<br\s*/?>`)
+	imMarkdownAnyTagRE    = regexp.MustCompile(`(?s)</?([A-Za-z][A-Za-z0-9:_-]*)(?:\s[^<>]*?)?>`)
+	imMarkdownLinkRE      = regexp.MustCompile(`(?is)<a\b[^>]*\bhref=(?:"([^"]*)"|'([^']*)')[^>]*>(.*?)</a>`)
+	imMarkdownCodeBlockRE = regexp.MustCompile(`(?is)^\s*<code(?:\s[^<>]*?)?>(.*?)</code>\s*$`)
+	imMarkdownLiOpenRE    = regexp.MustCompile(`(?is)<li(?:\s[^<>]*?)?>`)
+	imMarkdownLiCloseRE   = regexp.MustCompile(`(?is)<(/?)li(?:\s[^<>]*?)?\s*/?>`)
+)
+
+var imMarkdownHandlers = map[string]imMarkdownTagHandler{}
+
+func init() {
+	registerIMMarkdownHandler("title", handleIMMarkdownTitle)
+	for level := 1; level <= 9; level++ {
+		registerIMMarkdownHandler(fmt.Sprintf("h%d", level), handleIMMarkdownHeading(level))
+	}
+	registerIMMarkdownHandler("p", handleIMMarkdownParagraph)
+	registerIMMarkdownHandler("ul", handleIMMarkdownUnorderedList)
+	registerIMMarkdownHandler("ol", handleIMMarkdownOrderedList)
+	registerIMMarkdownHandler("li", handleIMMarkdownListItem)
+	registerIMMarkdownHandler("callout", handleIMMarkdownCallout)
+	registerIMMarkdownHandler("blockquote", handleIMMarkdownBlockquote)
+	registerIMMarkdownHandler("grid", handleIMMarkdownPassthroughContainer)
+	registerIMMarkdownHandler("column", handleIMMarkdownColumn)
+	registerIMMarkdownHandler("table", handleIMMarkdownTable)
+	registerIMMarkdownHandler("colgroup", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("col", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("pre", handleIMMarkdownPre)
+	registerIMMarkdownHandler("code", handleIMMarkdownCode)
+	registerIMMarkdownHandler("latex", handleIMMarkdownLatex)
+	registerIMMarkdownHandler("hr", handleIMMarkdownHorizontalRule)
+	registerIMMarkdownHandler("img", handleIMMarkdownImage)
+	registerIMMarkdownHandler("figure", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("source", handleIMMarkdownSource)
+	registerIMMarkdownHandler("button", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("time", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("whiteboard", handleIMMarkdownInlineCode)
+	registerIMMarkdownHandler("sheet", handleIMMarkdownSheet)
+	registerIMMarkdownHandler("task", handleIMMarkdownConditionalResourceLabel("任务", "task-id", "guid", "token", "id"))
+	registerIMMarkdownHandler("chat_card", handleIMMarkdownConditionalResourceLabel("群聊卡片", "chat-id", "chat_id", "id"))
+	registerIMMarkdownHandler("bitable", handleIMMarkdownResourceLabel("多维表格"))
+	registerIMMarkdownHandler("base_refer", handleIMMarkdownResourceLabel("多维表格"))
+	registerIMMarkdownHandler("okr", handleIMMarkdownResourceLabel("OKR"))
+	registerIMMarkdownHandler("poll", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("agenda", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("folder_manager", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("wiki_catalog", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("wiki_recent_update", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("chart_refer_host_perm", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("synced_reference", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("synced-source", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("mindnote", handleIMMarkdownDiscard)
+	registerIMMarkdownHandler("bookmark", handleIMMarkdownBookmark)
+	registerIMMarkdownHandler("cite", handleIMMarkdownCite)
+	registerIMMarkdownHandler("b", handleIMMarkdownStrong)
+	registerIMMarkdownHandler("em", handleIMMarkdownEmphasis)
+	registerIMMarkdownHandler("del", handleIMMarkdownDelete)
+	registerIMMarkdownHandler("u", handleIMMarkdownPlainInline)
+	registerIMMarkdownHandler("span", handleIMMarkdownPlainInline)
+	registerIMMarkdownHandler("a", handleIMMarkdownAnchor)
+}
+
+func isIMMarkdownFetch(runtime interface{ Str(string) string }) bool {
+	return strings.TrimSpace(runtime.Str("doc-format")) == "im-markdown"
+}
+
+func applyFetchIMMarkdown(data map[string]interface{}, docInput string) {
+	doc, ok := data["document"].(map[string]interface{})
+	if !ok {
+		return
+	}
+	content, ok := doc["content"].(string)
+	if !ok {
+		return
+	}
+	doc["content"] = convertToIMMarkdown(content, newIMMarkdownContext(docInput))
+}
+
+func newIMMarkdownContext(docInput string) imMarkdownContext {
+	base := "https://larkoffice.com"
+	raw := strings.TrimSpace(docInput)
+	if extracted, ok := imMarkdownBaseURLFromInput(raw); ok {
+		base = extracted
+	}
+	return imMarkdownContext{baseURL: base}
+}
+
+func (c imMarkdownContext) withBlockquote() imMarkdownContext {
+	c.blockquoteDepth++
+	return c
+}
+
+func (c imMarkdownContext) inBlockquote() bool {
+	return c.blockquoteDepth > 0
+}
+
+// imMarkdownBaseURLFromInput keeps the tenant host from --doc when it is a URL
+// so generated doc/sheet links point back to the same tenant. parseDocumentRef
+// intentionally strips host information, so it cannot serve this formatting path.
+func imMarkdownBaseURLFromInput(raw string) (string, bool) {
+	if raw == "" {
+		return "", false
+	}
+	if u, err := url.Parse(raw); err == nil && u.Scheme != "" && u.Host != "" {
+		return u.Scheme + "://" + u.Host, true
+	}
+	for _, marker := range []string{"/docx/", "/wiki/", "/doc/"} {
+		idx := strings.Index(raw, marker)
+		if idx <= 0 {
+			continue
+		}
+		candidate := strings.Trim(raw[:idx], "/")
+		if candidate == "" {
+			continue
+		}
+		if u, err := url.Parse(candidate); err == nil && u.Scheme != "" && u.Host != "" {
+			return u.Scheme + "://" + u.Host, true
+		}
+		if u, err := url.Parse("https://" + candidate); err == nil && u.Host != "" && strings.Contains(u.Host, ".") {
+			return "https://" + u.Host, true
+		}
+	}
+	return "", false
+}
+
+func convertToIMMarkdown(content string, imCtx imMarkdownContext) string {
+	var out strings.Builder
+	for offset := 0; offset < len(content); {
+		// Scan only to the next XML-like opening tag. Plain Markdown text between
+		// registered tags is copied unchanged, so ordinary Markdown is not re-parsed.
+		loc := imMarkdownTagStartRE.FindStringSubmatchIndex(content[offset:])
+		if loc == nil {
+			out.WriteString(content[offset:])
+			break
+		}
+		start := offset + loc[0]
+		openEnd := offset + loc[1]
+		tag := strings.ToLower(content[offset+loc[2] : offset+loc[3]])
+		handler, ok := imMarkdownHandlers[tag]
+		if !ok {
+			// Unknown tags are left intact. im-markdown only downgrades tags with
+			// explicit handlers so future server output does not get guessed at.
+			out.WriteString(content[offset:openEnd])
+			offset = openEnd
+			continue
+		}
+
+		out.WriteString(content[offset:start])
+		opening := content[start:openEnd]
+		attrs := parseIMMarkdownAttrs(opening)
+		if isSelfClosingIMMarkdownTag(opening) {
+			out.WriteString(handler.handle(opening, "", attrs, imCtx))
+			offset = openEnd
+			continue
+		}
+
+		// Use the handler's precompiled close regexp to find the matching end tag.
+		// Depth tracking keeps nested same-name containers paired correctly.
+		closeStart, closeEnd, found := findIMMarkdownClosingTag(content, openEnd, handler)
+		if !found {
+			// Malformed or truncated fragments are preserved as-is from the opening
+			// tag onward; do not drop content when the XML-ish structure is incomplete.
+			out.WriteString(content[start:])
+			break
+		}
+		segment := content[start:closeEnd]
+		inner := content[openEnd:closeStart]
+		out.WriteString(handler.handle(segment, inner, attrs, imCtx))
+		offset = closeEnd
+	}
+	return out.String()
+}
+
+func findIMMarkdownClosingTag(content string, from int, handler imMarkdownTagHandler) (int, int, bool) {
+	depth := 1
+	for _, loc := range handler.closeRE.FindAllStringSubmatchIndex(content[from:], -1) {
+		start := from + loc[0]
+		end := from + loc[1]
+		token := content[start:end]
+		if loc[2] >= 0 && content[from+loc[2]:from+loc[3]] == "/" {
+			depth--
+			if depth == 0 {
+				return start, end, true
+			}
+			continue
+		}
+		if !isSelfClosingIMMarkdownTag(token) {
+			depth++
+		}
+	}
+	return 0, 0, false
+}
+
+func parseIMMarkdownAttrs(opening string) map[string]string {
+	attrs := map[string]string{}
+	for _, match := range imMarkdownAttrRE.FindAllStringSubmatch(opening, -1) {
+		value := match[2]
+		if value == "" {
+			value = match[3]
+		}
+		attrs[strings.ToLower(match[1])] = html.UnescapeString(value)
+	}
+	return attrs
+}
+
+func isSelfClosingIMMarkdownTag(tag string) bool {
+	return strings.HasSuffix(strings.TrimSpace(tag), "/>")
+}
+
+func handleIMMarkdownTitle(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	text := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if text == "" {
+		return ""
+	}
+	return "# " + text
+}
+
+func handleIMMarkdownHeading(level int) imMarkdownHandleFunc {
+	return func(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+		text := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+		if text == "" {
+			return ""
+		}
+		markdownLevel := level
+		if markdownLevel > 6 {
+			markdownLevel = 6
+		}
+		return strings.Repeat("#", markdownLevel) + " " + text
+	}
+}
+
+func handleIMMarkdownParagraph(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if body == "" {
+		return ""
+	}
+	if imCtx.inBlockquote() {
+		return body + "\n"
+	}
+	return body
+}
+
+func handleIMMarkdownUnorderedList(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	return convertIMMarkdownListItems(inner, false, imCtx)
+}
+
+func handleIMMarkdownOrderedList(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	return convertIMMarkdownListItems(inner, true, imCtx)
+}
+
+func handleIMMarkdownListItem(_ string, inner string, attrs map[string]string, imCtx imMarkdownContext) string {
+	prefix := "-"
+	if seq := strings.TrimSpace(attrs["seq"]); seq != "" && seq != "auto" {
+		prefix = strings.TrimSuffix(seq, ".") + "."
+	}
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if body == "" {
+		return ""
+	}
+	return prefix + " " + indentIMMarkdownListContinuation(body) + "\n"
+}
+
+func handleIMMarkdownCallout(_ string, inner string, attrs map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	emoji := strings.TrimSpace(attrs["emoji"])
+	if emoji != "" {
+		if body == "" {
+			body = emoji
+		} else {
+			body = emoji + " " + body
+		}
+	}
+	if body == "" {
+		return "---\n---"
+	}
+	return fmt.Sprintf("---\n%s\n---", body)
+}
+
+func handleIMMarkdownBlockquote(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx.withBlockquote()))
+	if body == "" {
+		return ""
+	}
+	lines := strings.Split(body, "\n")
+	for i, line := range lines {
+		if strings.TrimSpace(line) == "" {
+			lines[i] = ">"
+			continue
+		}
+		lines[i] = "> " + line
+	}
+	return strings.Join(lines, "\n")
+}
+
+func handleIMMarkdownPassthroughContainer(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	return strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+}
+
+func handleIMMarkdownColumn(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if body == "" {
+		return ""
+	}
+	return body + "\n"
+}
+
+func handleIMMarkdownDiscard(_ string, _ string, _ map[string]string, _ imMarkdownContext) string {
+	return ""
+}
+
+func handleIMMarkdownInlineCode(segment string, _ string, _ map[string]string, _ imMarkdownContext) string {
+	return imMarkdownInlineCode(segment)
+}
+
+func handleIMMarkdownPre(_ string, inner string, attrs map[string]string, _ imMarkdownContext) string {
+	lang := strings.TrimSpace(attrs["lang"])
+	code := strings.TrimSpace(inner)
+	if match := imMarkdownCodeBlockRE.FindStringSubmatch(code); match != nil {
+		code = match[1]
+	}
+	return imMarkdownFencedCode(html.UnescapeString(code), lang)
+}
+
+func handleIMMarkdownCode(_ string, inner string, _ map[string]string, _ imMarkdownContext) string {
+	return imMarkdownInlineCode(markdownPlainText(inner))
+}
+
+func handleIMMarkdownLatex(_ string, inner string, _ map[string]string, _ imMarkdownContext) string {
+	expr := strings.TrimSpace(markdownPlainText(inner))
+	if expr == "" {
+		return ""
+	}
+	return "$" + strings.ReplaceAll(expr, "$", `\$`) + "$"
+}
+
+func handleIMMarkdownHorizontalRule(_ string, _ string, _ map[string]string, _ imMarkdownContext) string {
+	return "---"
+}
+
+func handleIMMarkdownImage(_ string, _ string, attrs map[string]string, _ imMarkdownContext) string {
+	href := firstNonEmpty(attrs["href"], attrs["src"], attrs["url"])
+	if href == "" {
+		return ""
+	}
+	alt := firstNonEmpty(attrs["alt"], attrs["name"], attrs["title"])
+	return fmt.Sprintf("![%s](%s)", escapeMarkdownLinkText(alt), escapeMarkdownLinkDestination(href))
+}
+
+func handleIMMarkdownSource(_ string, _ string, attrs map[string]string, _ imMarkdownContext) string {
+	name := strings.TrimSpace(attrs["name"])
+	if name == "" {
+		return ""
+	}
+	return imMarkdownInlineCode(name)
+}
+
+func handleIMMarkdownResourceLabel(label string) imMarkdownHandleFunc {
+	return func(_ string, _ string, _ map[string]string, _ imMarkdownContext) string {
+		return imMarkdownInlineCode(label)
+	}
+}
+
+func handleIMMarkdownConditionalResourceLabel(label string, attrNames ...string) imMarkdownHandleFunc {
+	return func(_ string, _ string, attrs map[string]string, _ imMarkdownContext) string {
+		for _, attrName := range attrNames {
+			if strings.TrimSpace(attrs[attrName]) != "" {
+				return imMarkdownInlineCode(label)
+			}
+		}
+		return ""
+	}
+}
+
+func handleIMMarkdownSheet(segment string, _ string, attrs map[string]string, imCtx imMarkdownContext) string {
+	token := strings.TrimSpace(attrs["token"])
+	if token == "" {
+		return imMarkdownInlineCode(segment)
+	}
+	label := "sheet"
+	if sheetID := strings.TrimSpace(attrs["sheet-id"]); sheetID != "" {
+		label = "sheet " + sheetID
+	}
+	return markdownLink(label, strings.TrimRight(imCtx.baseURL, "/")+"/sheets/"+token)
+}
+
+func handleIMMarkdownBookmark(segment string, inner string, attrs map[string]string, imCtx imMarkdownContext) string {
+	href := strings.TrimSpace(attrs["href"])
+	name := firstNonEmpty(attrs["name"], attrs["title"], markdownLinkLabelText(convertToIMMarkdown(inner, imCtx)), href)
+	if href == "" {
+		return name
+	}
+	return markdownLink(name, href)
+}
+
+func handleIMMarkdownStrong(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if body == "" {
+		return ""
+	}
+	return "**" + body + "**"
+}
+
+func handleIMMarkdownEmphasis(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if body == "" {
+		return ""
+	}
+	return "*" + body + "*"
+}
+
+func handleIMMarkdownDelete(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	body := strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+	if body == "" {
+		return ""
+	}
+	return "~~" + body + "~~"
+}
+
+func handleIMMarkdownPlainInline(_ string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	return strings.TrimSpace(convertToIMMarkdown(inner, imCtx))
+}
+
+func handleIMMarkdownAnchor(_ string, inner string, attrs map[string]string, imCtx imMarkdownContext) string {
+	href := strings.TrimSpace(attrs["href"])
+	text := firstNonEmpty(markdownLinkLabelText(convertToIMMarkdown(inner, imCtx)), attrs["name"], attrs["title"], href)
+	if href == "" {
+		return text
+	}
+	return markdownLink(text, href)
+}
+
+func handleIMMarkdownCite(segment string, inner string, attrs map[string]string, imCtx imMarkdownContext) string {
+	switch strings.ToLower(strings.TrimSpace(attrs["type"])) {
+	case "user":
+		userID := firstNonEmpty(attrs["user-id"], attrs["open-id"], attrs["id"])
+		name := firstNonEmpty(attrs["user-name"], attrs["name"], markdownPlainText(inner), userID)
+		if userID == "" {
+			return name
+		}
+		return fmt.Sprintf(`<at user_id="%s">%s</at>`, html.EscapeString(userID), html.EscapeString(name))
+	case "doc":
+		title := firstNonEmpty(attrs["title"], attrs["name"], attrs["doc-id"], "document")
+		if href := firstNonEmpty(attrs["href"], attrs["url"]); href != "" {
+			return markdownLink(title, href)
+		}
+		docID := firstNonEmpty(attrs["doc-id"], attrs["token"])
+		if docID == "" {
+			return imMarkdownInlineCode(segment)
+		}
+		fileType := strings.Trim(strings.ToLower(firstNonEmpty(attrs["file-type"], "docx")), "/")
+		return markdownLink(title, strings.TrimRight(imCtx.baseURL, "/")+"/"+fileType+"/"+docID)
+	case "citation":
+		if text, href, ok := extractIMMarkdownInnerLink(inner); ok {
+			return markdownLink(text, href)
+		}
+		if href := firstNonEmpty(attrs["href"], attrs["url"]); href != "" {
+			return markdownLink(firstNonEmpty(attrs["title"], attrs["name"], href), href)
+		}
+		return markdownPlainText(convertToIMMarkdown(inner, imCtx))
+	default:
+		return imMarkdownInlineCode(segment)
+	}
+}
+
+func handleIMMarkdownTable(segment string, inner string, _ map[string]string, imCtx imMarkdownContext) string {
+	// Rows and cells are matched with tag-depth tracking instead of non-greedy
+	// regex captures. A table nested inside a cell can contain its own </tr> and
+	// </td>; treating those as the outer row/cell boundary corrupts the table.
+	rowBodies := extractIMMarkdownElementBodies(inner, imMarkdownRowTagRE)
+	if len(rowBodies) == 0 {
+		return imMarkdownInlineCode(segment)
+	}
+
+	rows := make([][]string, 0, len(rowBodies))
+	for _, rowBody := range rowBodies {
+		cellBodies := extractIMMarkdownElementBodies(rowBody, imMarkdownCellTagRE)
+		if len(cellBodies) == 0 {
+			continue
+		}
+		row := make([]string, 0, len(cellBodies))
+		for _, cellBody := range cellBodies {
+			row = append(row, normalizeIMMarkdownTableCell(convertToIMMarkdown(cellBody, imCtx)))
+		}
+		rows = append(rows, row)
+	}
+	if len(rows) == 0 {
+		return imMarkdownInlineCode(segment)
+	}
+
+	cols := 0
+	for _, row := range rows {
+		if len(row) > cols {
+			cols = len(row)
+		}
+	}
+	var out strings.Builder
+	writeIMMarkdownTableRow(&out, padIMMarkdownTableRow(rows[0], cols))
+	separator := make([]string, cols)
+	for i := range separator {
+		separator[i] = "-"
+	}
+	writeIMMarkdownTableRow(&out, separator)
+	for _, row := range rows[1:] {
+		writeIMMarkdownTableRow(&out, padIMMarkdownTableRow(row, cols))
+	}
+	return strings.TrimRight(out.String(), "\n")
+}
+
+// extractIMMarkdownElementBodies returns the inner content of each top-level
+// element matched by tagRE. tagRE must expose the optional closing slash as its
+// first capture group, matching the row/cell regexes above.
+func extractIMMarkdownElementBodies(content string, tagRE *regexp.Regexp) []string {
+	var bodies []string
+	for offset := 0; offset < len(content); {
+		loc := tagRE.FindStringSubmatchIndex(content[offset:])
+		if loc == nil {
+			break
+		}
+		openStart := offset + loc[0]
+		openEnd := offset + loc[1]
+		opening := content[openStart:openEnd]
+		if loc[2] >= 0 && content[offset+loc[2]:offset+loc[3]] == "/" {
+			offset = openEnd
+			continue
+		}
+		if isSelfClosingIMMarkdownTag(opening) {
+			bodies = append(bodies, "")
+			offset = openEnd
+			continue
+		}
+		closeStart, closeEnd, found := findIMMarkdownElementClosingTag(content, openEnd, tagRE)
+		if !found {
+			break
+		}
+		bodies = append(bodies, content[openEnd:closeStart])
+		offset = closeEnd
+	}
+	return bodies
+}
+
+func findIMMarkdownElementClosingTag(content string, from int, tagRE *regexp.Regexp) (int, int, bool) {
+	depth := 1
+	for _, loc := range tagRE.FindAllStringSubmatchIndex(content[from:], -1) {
+		start := from + loc[0]
+		end := from + loc[1]
+		token := content[start:end]
+		if loc[2] >= 0 && content[from+loc[2]:from+loc[3]] == "/" {
+			depth--
+			if depth == 0 {
+				return start, end, true
+			}
+			continue
+		}
+		if !isSelfClosingIMMarkdownTag(token) {
+			depth++
+		}
+	}
+	return 0, 0, false
+}
+
+func normalizeIMMarkdownTableCell(cell string) string {
+	const brPlaceholder = "\x00BR\x00"
+	cell = imMarkdownCellBreakRE.ReplaceAllString(cell, brPlaceholder)
+	cell = imMarkdownAnyTagRE.ReplaceAllStringFunc(cell, func(tag string) string {
+		name := strings.ToLower(strings.TrimPrefix(imMarkdownAnyTagRE.FindStringSubmatch(tag)[1], "/"))
+		if name == "at" {
+			return tag
+		}
+		return ""
+	})
+	cell = html.UnescapeString(cell)
+	cell = strings.ReplaceAll(cell, brPlaceholder, "<br>")
+	cell = strings.ReplaceAll(cell, "  \n", "<br>")
+	cell = strings.ReplaceAll(cell, "\n", "<br>")
+	cell = strings.ReplaceAll(cell, "|", `\|`)
+	lines := strings.Fields(cell)
+	if len(lines) == 0 {
+		return ""
+	}
+	return strings.Join(lines, " ")
+}
+
+func writeIMMarkdownTableRow(out *strings.Builder, row []string) {
+	out.WriteString("| ")
+	out.WriteString(strings.Join(row, " | "))
+	out.WriteString(" |\n")
+}
+
+func padIMMarkdownTableRow(row []string, cols int) []string {
+	if len(row) >= cols {
+		return row
+	}
+	padded := make([]string, cols)
+	copy(padded, row)
+	return padded
+}
+
+func convertIMMarkdownListItems(inner string, ordered bool, imCtx imMarkdownContext) string {
+	var out strings.Builder
+	for offset, index := 0, 1; offset < len(inner); {
+		loc := imMarkdownLiOpenRE.FindStringIndex(inner[offset:])
+		if loc == nil {
+			break
+		}
+		openStart := offset + loc[0]
+		openEnd := offset + loc[1]
+		opening := inner[openStart:openEnd]
+		closeStart, closeEnd, found := findIMMarkdownListItemClosingTag(inner, openEnd)
+		if !found {
+			break
+		}
+		body := strings.TrimSpace(convertToIMMarkdown(inner[openEnd:closeStart], imCtx))
+		if body != "" {
+			prefix := "-"
+			if ordered {
+				attrs := parseIMMarkdownAttrs(opening)
+				if seq := strings.TrimSpace(attrs["seq"]); seq != "" && seq != "auto" {
+					prefix = strings.TrimSuffix(seq, ".") + "."
+				} else {
+					prefix = fmt.Sprintf("%d.", index)
+				}
+				index++
+			}
+			out.WriteString(prefix)
+			out.WriteString(" ")
+			out.WriteString(indentIMMarkdownListContinuation(body))
+			out.WriteString("\n")
+		}
+		offset = closeEnd
+	}
+	return strings.TrimRight(out.String(), "\n")
+}
+
+func findIMMarkdownListItemClosingTag(content string, from int) (int, int, bool) {
+	depth := 1
+	for _, loc := range imMarkdownLiCloseRE.FindAllStringSubmatchIndex(content[from:], -1) {
+		start := from + loc[0]
+		end := from + loc[1]
+		token := content[start:end]
+		if loc[2] >= 0 && content[from+loc[2]:from+loc[3]] == "/" {
+			depth--
+			if depth == 0 {
+				return start, end, true
+			}
+			continue
+		}
+		if !isSelfClosingIMMarkdownTag(token) {
+			depth++
+		}
+	}
+	return 0, 0, false
+}
+
+func indentIMMarkdownListContinuation(body string) string {
+	return strings.ReplaceAll(body, "\n", "\n  ")
+}
+
+func extractIMMarkdownInnerLink(inner string) (string, string, bool) {
+	match := imMarkdownLinkRE.FindStringSubmatch(inner)
+	if match == nil {
+		return "", "", false
+	}
+	href := match[1]
+	if href == "" {
+		href = match[2]
+	}
+	text := strings.TrimSpace(markdownPlainText(match[3]))
+	if text == "" {
+		text = href
+	}
+	return text, html.UnescapeString(href), true
+}
+
+func markdownPlainText(s string) string {
+	s = imMarkdownCellBreakRE.ReplaceAllString(s, "\n")
+	s = imMarkdownAnyTagRE.ReplaceAllString(s, "")
+	return strings.TrimSpace(html.UnescapeString(s))
+}
+
+func markdownLinkLabelText(s string) string {
+	text := markdownPlainText(s)
+	if !strings.Contains(text, "---") {
+		return text
+	}
+	lines := strings.Split(text, "\n")
+	kept := lines[:0]
+	for _, line := range lines {
+		if strings.TrimSpace(line) == "---" {
+			continue
+		}
+		kept = append(kept, line)
+	}
+	return strings.TrimSpace(strings.Join(kept, "\n"))
+}
+
+func markdownLink(text, href string) string {
+	cleanHref := strings.TrimSpace(href)
+	return fmt.Sprintf("[%s](%s)", escapeMarkdownLinkText(firstNonEmpty(text, cleanHref)), escapeMarkdownLinkDestination(cleanHref))
+}
+
+func escapeMarkdownLinkText(text string) string {
+	text = strings.ReplaceAll(text, `\`, `\\`)
+	text = strings.ReplaceAll(text, `[`, `\[`)
+	text = strings.ReplaceAll(text, `]`, `\]`)
+	return text
+}
+
+func escapeMarkdownLinkDestination(href string) string {
+	// Lark/Feishu IM Markdown does not reliably parse raw spaces or parentheses
+	// inside (...). Keep URL delimiters like :/?#&= intact, but percent-encode
+	// characters that can terminate or split the Markdown link destination.
+	var out strings.Builder
+	out.Grow(len(href))
+	for i := 0; i < len(href); {
+		if href[i] == '%' {
+			if i+2 < len(href) && isHexDigit(href[i+1]) && isHexDigit(href[i+2]) {
+				out.WriteString(href[i : i+3])
+				i += 3
+			} else {
+				writePercentEncodedByte(&out, href[i])
+				i++
+			}
+			continue
+		}
+		if href[i] < utf8.RuneSelf {
+			if shouldPercentEncodeIMMarkdownURLByte(href[i]) {
+				writePercentEncodedByte(&out, href[i])
+			} else {
+				out.WriteByte(href[i])
+			}
+			i++
+			continue
+		}
+		r, size := utf8.DecodeRuneInString(href[i:])
+		if r == utf8.RuneError && size == 1 {
+			writePercentEncodedByte(&out, href[i])
+			i++
+			continue
+		}
+		for _, b := range []byte(href[i : i+size]) {
+			writePercentEncodedByte(&out, b)
+		}
+		i += size
+	}
+	return out.String()
+}
+
+func shouldPercentEncodeIMMarkdownURLByte(b byte) bool {
+	if b <= ' ' || b >= 0x7f {
+		return true
+	}
+	switch b {
+	case '(', ')', '<', '>', '"', '\\', '^', '`', '{', '|', '}':
+		return true
+	default:
+		return false
+	}
+}
+
+func writePercentEncodedByte(out *strings.Builder, b byte) {
+	const hex = "0123456789ABCDEF"
+	out.WriteByte('%')
+	out.WriteByte(hex[b>>4])
+	out.WriteByte(hex[b&0x0f])
+}
+
+func isHexDigit(b byte) bool {
+	return ('0' <= b && b <= '9') || ('a' <= b && b <= 'f') || ('A' <= b && b <= 'F')
+}
+
+func imMarkdownInlineCode(s string) string {
+	maxRun := 0
+	run := 0
+	for _, r := range s {
+		if r == '`' {
+			run++
+			if run > maxRun {
+				maxRun = run
+			}
+			continue
+		}
+		run = 0
+	}
+	fence := strings.Repeat("`", maxRun+1)
+	if strings.HasPrefix(s, "`") || strings.HasSuffix(s, "`") {
+		return fence + " " + s + " " + fence
+	}
+	return fence + s + fence
+}
+
+func imMarkdownFencedCode(code, lang string) string {
+	maxRun := 0
+	for _, line := range strings.Split(code, "\n") {
+		if run := leadingBacktickRun(line); run > maxRun {
+			maxRun = run
+		}
+	}
+	fenceLen := maxRun + 1
+	if fenceLen < 3 {
+		fenceLen = 3
+	}
+	fence := strings.Repeat("`", fenceLen)
+	return fence + strings.TrimSpace(lang) + "\n" + strings.Trim(code, "\n") + "\n" + fence
+}
+
+func leadingBacktickRun(s string) int {
+	run := 0
+	for _, r := range s {
+		if r != '`' {
+			break
+		}
+		run++
+	}
+	return run
+}
+
+func firstNonEmpty(values ...string) string {
+	for _, value := range values {
+		if strings.TrimSpace(value) != "" {
+			return strings.TrimSpace(value)
+		}
+	}
+	return ""
+}
--- a/shortcuts/doc/docs_fetch_im_markdown_test.go
+++ b/shortcuts/doc/docs_fetch_im_markdown_test.go
--- a/shortcuts/doc/docs_fetch_v2.go
+++ b/shortcuts/doc/docs_fetch_v2.go
@@ -17,7 +17,7 @@ import (
 // v2FetchFlags returns the flag definitions for the v2 (OpenAPI) fetch path.
 func v2FetchFlags() []common.Flag {
 	return []common.Flag{
-		{Name: "doc-format", Desc: "output content format; xml keeps DocxXML structure and optional block ids, markdown is plain export", Default: "xml", Enum: []string{"xml", "markdown"}},
+		{Name: "doc-format", Desc: "output content format; xml keeps DocxXML structure and optional block ids, markdown is plain export, im-markdown downgrades residual DocxXML fragments for IM messages", Default: "xml", Enum: []string{"xml", "markdown", "im-markdown"}},
 		{Name: "detail", Desc: "detail level; simple for reading, with-ids for block references, full for styles and edit metadata", Default: "simple", Enum: []string{"simple", "with-ids", "full"}},
 		{Name: "lang", Desc: "user cite display language, e.g. en-US, zh-CN, ja-JP"},
 		{Name: "revision-id", Desc: "document revision id; -1 means latest", Type: "int", Default: "-1"},
@@ -72,6 +72,9 @@ func executeFetchV2(_ context.Context, runtime *common.RuntimeContext) error {
 	if warning := addFetchDetailDowngradeWarning(runtime, data); warning != "" && runtime.Format == "pretty" {
 		fmt.Fprintf(runtime.IO().ErrOut, "warning: %s\n", warning)
 	}
+	if isIMMarkdownFetch(runtime) {
+		applyFetchIMMarkdown(data, runtime.Str("doc"))
+	}

 	runtime.OutFormatRaw(data, nil, func(w io.Writer) {
 		if doc, ok := data["document"].(map[string]interface{}); ok {
@@ -85,7 +88,7 @@ func executeFetchV2(_ context.Context, runtime *common.RuntimeContext) error {

 func buildFetchBody(runtime *common.RuntimeContext) map[string]interface{} {
 	body := map[string]interface{}{
-		"format": runtime.Str("doc-format"),
+		"format": effectiveFetchFormat(runtime),
 	}
 	if v := runtime.Int("revision-id"); v > 0 {
 		body["revision_id"] = v
@@ -122,6 +125,14 @@ func buildFetchBody(runtime *common.RuntimeContext) map[string]interface{} {
 	return body
 }

+func effectiveFetchFormat(runtime *common.RuntimeContext) string {
+	format := strings.TrimSpace(runtime.Str("doc-format"))
+	if format == "im-markdown" {
+		return "markdown"
+	}
+	return format
+}
+
 func resolveFetchLang(runtime *common.RuntimeContext) string {
 	if runtime.Changed("lang") {
 		return strings.TrimSpace(runtime.Str("lang"))
--- a/shortcuts/doc/docs_fetch_v2_test.go
+++ b/shortcuts/doc/docs_fetch_v2_test.go
@@ -6,9 +6,12 @@ package doc
 import (
 	"context"
 	"encoding/json"
+	"errors"
+	"reflect"
 	"strings"
 	"testing"

+	"github.com/larksuite/cli/errs"
 	"github.com/larksuite/cli/internal/cmdutil"
 	"github.com/larksuite/cli/internal/core"
 	"github.com/larksuite/cli/internal/httpmock"
@@ -104,6 +107,369 @@ func TestBuildFetchBodyExplicitBlankLangOmitsLang(t *testing.T) {
 	}
 }

+func TestBuildFetchBodyIncludesRevisionAndFullDetail(t *testing.T) {
+	t.Parallel()
+
+	runtime := newFetchBodyTestRuntime(context.Background())
+	mustSetFetchFlag(t, runtime, "revision-id", "42")
+	mustSetFetchFlag(t, runtime, "detail", "full")
+
+	body := buildFetchBody(runtime)
+	if got := body["revision_id"]; got != 42 {
+		t.Fatalf("revision_id = %#v, want 42", got)
+	}
+	exportOption, _ := body["export_option"].(map[string]interface{})
+	want := map[string]interface{}{
+		"export_block_id":        true,
+		"export_style_attrs":     true,
+		"export_cite_extra_data": true,
+	}
+	if !reflect.DeepEqual(exportOption, want) {
+		t.Fatalf("export_option = %#v, want %#v", exportOption, want)
+	}
+}
+
+func TestBuildFetchBodyIncludesWithIDsDetail(t *testing.T) {
+	t.Parallel()
+
+	runtime := newFetchBodyTestRuntime(context.Background())
+	mustSetFetchFlag(t, runtime, "detail", "with-ids")
+
+	body := buildFetchBody(runtime)
+	exportOption, _ := body["export_option"].(map[string]interface{})
+	want := map[string]interface{}{
+		"export_block_id": true,
+	}
+	if !reflect.DeepEqual(exportOption, want) {
+		t.Fatalf("export_option = %#v, want %#v", exportOption, want)
+	}
+}
+
+func TestBuildFetchBodyIncludesReadOption(t *testing.T) {
+	t.Parallel()
+
+	runtime := newFetchBodyTestRuntime(context.Background())
+	mustSetFetchFlag(t, runtime, "scope", "section")
+	mustSetFetchFlag(t, runtime, "start-block-id", "blk_heading")
+
+	body := buildFetchBody(runtime)
+	want := map[string]interface{}{
+		"read_mode":      "section",
+		"start_block_id": "blk_heading",
+	}
+	if got := body["read_option"]; !reflect.DeepEqual(got, want) {
+		t.Fatalf("read_option = %#v, want %#v", got, want)
+	}
+}
+
+func TestBuildReadOptionModes(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		setFlags map[string]string
+		want     map[string]interface{}
+	}{
+		{
+			name: "full omits read option",
+			setFlags: map[string]string{
+				"scope": "full",
+			},
+			want: nil,
+		},
+		{
+			name: "outline with max depth",
+			setFlags: map[string]string{
+				"scope":     "outline",
+				"max-depth": "3",
+			},
+			want: map[string]interface{}{
+				"read_mode": "outline",
+				"max_depth": "3",
+			},
+		},
+		{
+			name: "range with block ids and context",
+			setFlags: map[string]string{
+				"scope":          "range",
+				"start-block-id": "blk_start",
+				"end-block-id":   "blk_end",
+				"context-before": "2",
+				"context-after":  "1",
+				"max-depth":      "0",
+			},
+			want: map[string]interface{}{
+				"read_mode":      "range",
+				"start_block_id": "blk_start",
+				"end_block_id":   "blk_end",
+				"context_before": "2",
+				"context_after":  "1",
+				"max_depth":      "0",
+			},
+		},
+		{
+			name: "keyword with query",
+			setFlags: map[string]string{
+				"scope":          "keyword",
+				"keyword":        "foo|bar",
+				"context-before": "1",
+			},
+			want: map[string]interface{}{
+				"read_mode":      "keyword",
+				"keyword":        "foo|bar",
+				"context_before": "1",
+			},
+		},
+		{
+			name: "section keeps unlimited depth omitted",
+			setFlags: map[string]string{
+				"scope":          "section",
+				"start-block-id": "blk_heading",
+				"max-depth":      "-1",
+			},
+			want: map[string]interface{}{
+				"read_mode":      "section",
+				"start_block_id": "blk_heading",
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			runtime := newFetchBodyTestRuntime(context.Background())
+			for name, value := range tt.setFlags {
+				mustSetFetchFlag(t, runtime, name, value)
+			}
+
+			if got := buildReadOption(runtime); !reflect.DeepEqual(got, tt.want) {
+				t.Fatalf("buildReadOption() = %#v, want %#v", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestValidateReadModeFlagsRejectsInvalidScopeOptions(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name       string
+		setFlags   map[string]string
+		wantParam  string
+		wantParams []string
+	}{
+		{
+			name: "negative context before",
+			setFlags: map[string]string{
+				"scope":          "range",
+				"start-block-id": "blk_start",
+				"context-before": "-1",
+			},
+			wantParam: "--context-before",
+		},
+		{
+			name: "negative context after",
+			setFlags: map[string]string{
+				"scope":          "range",
+				"start-block-id": "blk_start",
+				"context-after":  "-1",
+			},
+			wantParam: "--context-after",
+		},
+		{
+			name: "max depth below unlimited sentinel",
+			setFlags: map[string]string{
+				"scope":          "range",
+				"start-block-id": "blk_start",
+				"max-depth":      "-2",
+			},
+			wantParam: "--max-depth",
+		},
+		{
+			name: "range needs boundary",
+			setFlags: map[string]string{
+				"scope": "range",
+			},
+			wantParams: []string{
+				"--start-block-id",
+				"--end-block-id",
+			},
+		},
+		{
+			name: "keyword needs keyword",
+			setFlags: map[string]string{
+				"scope": "keyword",
+			},
+			wantParam: "--keyword",
+		},
+		{
+			name: "section needs start block",
+			setFlags: map[string]string{
+				"scope": "section",
+			},
+			wantParam: "--start-block-id",
+		},
+		{
+			name: "unknown scope",
+			setFlags: map[string]string{
+				"scope": "bad",
+			},
+			wantParam: "--scope",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			runtime := newFetchBodyTestRuntime(context.Background())
+			for name, value := range tt.setFlags {
+				mustSetFetchFlag(t, runtime, name, value)
+			}
+
+			err := validateReadModeFlags(runtime)
+			if err == nil {
+				t.Fatal("validateReadModeFlags() succeeded, want error")
+			}
+			assertValidationContract(t, err, errs.SubtypeInvalidArgument, tt.wantParam, tt.wantParams...)
+		})
+	}
+}
+
+func TestValidateReadModeFlagsAcceptsValidScopeOptions(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		setFlags map[string]string
+	}{
+		{
+			name: "outline",
+			setFlags: map[string]string{
+				"scope": "outline",
+			},
+		},
+		{
+			name: "range with end block",
+			setFlags: map[string]string{
+				"scope":        "range",
+				"end-block-id": "blk_end",
+			},
+		},
+		{
+			name: "keyword with keyword",
+			setFlags: map[string]string{
+				"scope":   "keyword",
+				"keyword": "bug|缺陷",
+			},
+		},
+		{
+			name: "section with start block",
+			setFlags: map[string]string{
+				"scope":          "section",
+				"start-block-id": "blk_heading",
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			runtime := newFetchBodyTestRuntime(context.Background())
+			for name, value := range tt.setFlags {
+				mustSetFetchFlag(t, runtime, name, value)
+			}
+
+			if err := validateReadModeFlags(runtime); err != nil {
+				t.Fatalf("validateReadModeFlags() error = %v", err)
+			}
+		})
+	}
+}
+
+func TestValidateFetchV2RejectsInvalidDocAndScope(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name      string
+		setFlags  map[string]string
+		wantParam string
+	}{
+		{
+			name: "invalid doc",
+			setFlags: map[string]string{
+				"doc": "https://example.com/sheets/sht_token",
+			},
+			wantParam: "--doc",
+		},
+		{
+			name: "invalid scope",
+			setFlags: map[string]string{
+				"scope": "bad",
+			},
+			wantParam: "--scope",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			runtime := newFetchShortcutTestRuntime(t, "", tt.setFlags)
+			err := validateFetchV2(context.Background(), runtime)
+			if err == nil {
+				t.Fatal("validateFetchV2() succeeded, want error")
+			}
+			assertValidationContract(t, err, errs.SubtypeInvalidArgument, tt.wantParam)
+		})
+	}
+}
+
+func TestAddFetchDetailDowngradeWarningNoops(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		setFlags map[string]string
+	}{
+		{
+			name: "xml format",
+			setFlags: map[string]string{
+				"doc-format": "xml",
+				"detail":     "full",
+			},
+		},
+		{
+			name: "markdown simple detail",
+			setFlags: map[string]string{
+				"doc-format": "markdown",
+				"detail":     "simple",
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			runtime := newFetchBodyTestRuntime(context.Background())
+			for name, value := range tt.setFlags {
+				mustSetFetchFlag(t, runtime, name, value)
+			}
+
+			data := map[string]interface{}{}
+			if got := addFetchDetailDowngradeWarning(runtime, data); got != "" {
+				t.Fatalf("warning = %q, want empty", got)
+			}
+			if _, ok := data["warnings"]; ok {
+				t.Fatalf("unexpected warnings: %#v", data["warnings"])
+			}
+		})
+	}
+}
+
 func TestDocsFetchDryRunDefaultsToV2Endpoint(t *testing.T) {
 	t.Parallel()

@@ -141,36 +507,54 @@ func TestDocsFetchAPIVersionV1StillUsesV2Endpoint(t *testing.T) {
 	}
 }

+func TestDocsFetchIMMarkdownRequestsMarkdownFromAPI(t *testing.T) {
+	t.Parallel()
+
+	runtime := newFetchShortcutTestRuntime(t, "", map[string]string{
+		"doc-format": "im-markdown",
+	})
+	if err := validateFetchV2(context.Background(), runtime); err != nil {
+		t.Fatalf("validateFetchV2() error = %v", err)
+	}
+
+	dry := decodeDocDryRun(t, DocsFetch.DryRun(context.Background(), runtime))
+	if got, want := dry.API[0].Body["format"], "markdown"; got != want {
+		t.Fatalf("dry-run format = %#v, want %q", got, want)
+	}
+}
+
 func TestDocsFetchMarkdownDetailDowngradesToSimple(t *testing.T) {
 	t.Parallel()

-	for _, detail := range []string{"with-ids", "full"} {
-		t.Run(detail, func(t *testing.T) {
-			t.Parallel()
+	for _, format := range []string{"markdown", "im-markdown"} {
+		for _, detail := range []string{"with-ids", "full"} {
+			t.Run(format+"/"+detail, func(t *testing.T) {
+				t.Parallel()

-			runtime := newFetchShortcutTestRuntime(t, "", map[string]string{
-				"doc-format": "markdown",
-				"detail":     detail,
+				runtime := newFetchShortcutTestRuntime(t, "", map[string]string{
+					"doc-format": format,
+					"detail":     detail,
+				})
+				if err := validateFetchV2(context.Background(), runtime); err != nil {
+					t.Fatalf("validateFetchV2() error = %v", err)
+				}
+
+				dry := decodeDocDryRun(t, DocsFetch.DryRun(context.Background(), runtime))
+				exportOption, _ := dry.API[0].Body["export_option"].(map[string]interface{})
+				if exportOption == nil {
+					t.Fatalf("missing export_option: %#v", dry.API[0].Body)
+				}
+				if got := exportOption["export_block_id"]; got != false {
+					t.Fatalf("export_block_id = %#v, want false after markdown detail downgrade", got)
+				}
+				if got := exportOption["export_style_attrs"]; got != false {
+					t.Fatalf("export_style_attrs = %#v, want false after markdown detail downgrade", got)
+				}
+				if got := exportOption["export_cite_extra_data"]; got != false {
+					t.Fatalf("export_cite_extra_data = %#v, want false after markdown detail downgrade", got)
+				}
 			})
-			if err := validateFetchV2(context.Background(), runtime); err != nil {
-				t.Fatalf("validateFetchV2() error = %v", err)
-			}
-
-			dry := decodeDocDryRun(t, DocsFetch.DryRun(context.Background(), runtime))
-			exportOption, _ := dry.API[0].Body["export_option"].(map[string]interface{})
-			if exportOption == nil {
-				t.Fatalf("missing export_option: %#v", dry.API[0].Body)
-			}
-			if got := exportOption["export_block_id"]; got != false {
-				t.Fatalf("export_block_id = %#v, want false after markdown detail downgrade", got)
-			}
-			if got := exportOption["export_style_attrs"]; got != false {
-				t.Fatalf("export_style_attrs = %#v, want false after markdown detail downgrade", got)
-			}
-			if got := exportOption["export_cite_extra_data"]; got != false {
-				t.Fatalf("export_cite_extra_data = %#v, want false after markdown detail downgrade", got)
-			}
-		})
+		}
 	}
 }

@@ -261,6 +645,107 @@ func TestDocsFetchMarkdownDetailDowngradeWarnsInPrettyOutput(t *testing.T) {
 	}
 }

+func TestDocsFetchV2ReturnsAPIError(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, docsTestConfigWithAppID("docs-fetch-api-error"))
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/docs_ai/v1/documents/doxcnFetchAPIError/fetch",
+		Body: map[string]interface{}{
+			"code": 999999,
+			"msg":  "fetch failed",
+		},
+	})
+
+	err := mountAndRunDocs(t, DocsFetch, []string{
+		"+fetch",
+		"--doc", "doxcnFetchAPIError",
+		"--as", "bot",
+	}, f, stdout)
+	if err == nil {
+		t.Fatal("mountAndRunDocs() succeeded, want API error")
+	}
+	var apiErr *errs.APIError
+	if !errors.As(err, &apiErr) {
+		t.Fatalf("error type = %T, want *errs.APIError (%v)", err, err)
+	}
+	p, ok := errs.ProblemOf(err)
+	if !ok {
+		t.Fatalf("ProblemOf() ok = false for %T (%v)", err, err)
+	}
+	if p.Category != errs.CategoryAPI {
+		t.Errorf("category = %q, want %q", p.Category, errs.CategoryAPI)
+	}
+	if p.Subtype != errs.SubtypeUnknown {
+		t.Errorf("subtype = %q, want %q", p.Subtype, errs.SubtypeUnknown)
+	}
+	if p.Code != 999999 {
+		t.Errorf("code = %d, want 999999", p.Code)
+	}
+	if p.Message != "fetch failed" {
+		t.Errorf("message = %q, want %q", p.Message, "fetch failed")
+	}
+	if cause := errors.Unwrap(err); cause != nil {
+		t.Fatalf("unexpected wrapped cause for API response error: %T %v", cause, cause)
+	}
+}
+
+func TestDocsFetchIMMarkdownConvertsContentInJSONOutput(t *testing.T) {
+	t.Setenv("LARKSUITE_CLI_CONFIG_DIR", t.TempDir())
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, docsTestConfigWithAppID("docs-fetch-im-markdown"))
+	reg.Register(&httpmock.Stub{
+		Method: "POST",
+		URL:    "/open-apis/docs_ai/v1/documents/doxcnFetchIMMarkdown/fetch",
+		Body: map[string]interface{}{
+			"code": 0,
+			"msg":  "ok",
+			"data": map[string]interface{}{
+				"document": map[string]interface{}{
+					"document_id": "doxcnFetchIMMarkdown",
+					"revision_id": float64(1),
+					"content": strings.Join([]string{
+						`<title>Doc Title</title>`,
+						`<callout emoji="💡">Read **this**.</callout>`,
+						`<bookmark name="Example" href="https://example.com"></bookmark>`,
+					}, "\n\n"),
+				},
+			},
+		},
+	})
+
+	err := mountAndRunDocs(t, DocsFetch, []string{
+		"+fetch",
+		"--doc", "doxcnFetchIMMarkdown",
+		"--doc-format", "im-markdown",
+		"--as", "bot",
+	}, f, stdout)
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	var envelope map[string]interface{}
+	if err := json.Unmarshal(stdout.Bytes(), &envelope); err != nil {
+		t.Fatalf("decode output: %v\nraw=%s", err, stdout.String())
+	}
+	data, _ := envelope["data"].(map[string]interface{})
+	doc, _ := data["document"].(map[string]interface{})
+	content, _ := doc["content"].(string)
+	for _, want := range []string{
+		"# Doc Title",
+		"---\n💡 Read **this**.\n---",
+		"[Example](https://example.com)",
+	} {
+		if !strings.Contains(content, want) {
+			t.Fatalf("converted content missing %q:\n%s", want, content)
+		}
+	}
+	if strings.Contains(content, "<title>") || strings.Contains(content, "<callout") || strings.Contains(content, "<bookmark") {
+		t.Fatalf("converted content still contains downgraded XML tags:\n%s", content)
+	}
+}
+
 func TestDocsFetchRejectsLegacyFlags(t *testing.T) {
 	tests := []struct {
 		name     string
@@ -291,6 +776,7 @@ func TestDocsFetchRejectsLegacyFlags(t *testing.T) {
 			if err == nil {
 				t.Fatal("expected v2-only validation error")
 			}
+			assertValidationContract(t, err, errs.SubtypeInvalidArgument, "--offset")
 			for _, want := range tt.want {
 				if !strings.Contains(err.Error(), want) {
 					t.Fatalf("error missing %q: %v", want, err)
@@ -316,6 +802,14 @@ func newFetchBodyTestRuntime(ctx context.Context) *common.RuntimeContext {
 	return common.TestNewRuntimeContextWithCtx(ctx, cmd, nil)
 }

+func mustSetFetchFlag(t *testing.T, runtime *common.RuntimeContext, name, value string) {
+	t.Helper()
+
+	if err := runtime.Cmd.Flags().Set(name, value); err != nil {
+		t.Fatalf("set %s: %v", name, err)
+	}
+}
+
 func newFetchShortcutTestRuntime(t *testing.T, apiVersion string, setFlags map[string]string) *common.RuntimeContext {
 	t.Helper()

--- a/skills/lark-doc/SKILL.md
+++ b/skills/lark-doc/SKILL.md
@@ -67,7 +67,7 @@ Shortcut 是对常用操作的高级封装（`lark-cli docs +<verb> [flags]`）
 | Shortcut | 说明 |
 |----------|------|
 | [`+create`](references/lark-doc-create.md) | Create a Lark document (XML / Markdown) |
-| [`+fetch`](references/lark-doc-fetch.md) | Fetch Lark document content (XML / Markdown) |
+| [`+fetch`](references/lark-doc-fetch.md) | Fetch Lark document content (XML / Markdown / im-markdown; `im-markdown` only after fetch for `lark-im`) |
 | [`+update`](references/lark-doc-update.md) | Update a Lark document (str_replace / block_insert_after / block_replace / ...) |
 | [`+media-insert`](references/lark-doc-media-insert.md) | Insert a local image or file at the end of a Lark document (4-step orchestration + auto-rollback). Prefer `--from-clipboard` when the image is already on the system clipboard (screenshots, copy from Feishu/browser); use `--file` only for on-disk sources. |
 | [`+media-download`](references/lark-doc-media-download.md) | Download document media or whiteboard thumbnail (auto-detects extension) |
--- a/skills/lark-doc/references/lark-doc-fetch.md
+++ b/skills/lark-doc/references/lark-doc-fetch.md
@@ -91,7 +91,7 @@ lark-cli docs +fetch --api-version v2 --doc Z1Fj...tnAc \
 }
 ```

-`content` 的格式由 `--doc-format` 决定。设置 `--scope` 时会被 `<fragment>` 包裹，详见上文"局部读取的输出结构"。
+`content` 的格式由 `--doc-format` 决定；`im-markdown` 仅用于获取内容后在 `lark-im` 场景下使用。设置 `--scope` 时会被 `<fragment>` 包裹，详见上文"局部读取的输出结构"。

 ## 参数

@@ -99,7 +99,7 @@ lark-cli docs +fetch --api-version v2 --doc Z1Fj...tnAc \
 |------|------|------|
 | `--api-version` | 是 | 固定传 `v2` |
 | `--doc` | 是 | 文档 URL 或 token（支持 `/docx/` 和 `/wiki/`） |
-| `--doc-format` | 否 | `xml`（默认）\| `markdown` \| `text` |
+| `--doc-format` | 否 | `xml`（默认）\| `markdown` \| `text` \| `im-markdown`（仅用于获取内容后在 `lark-im` 场景下使用） |
 | `--detail` | 否 | `simple`（默认）\| `with-ids` \| `full` |
 | `--revision-id` | 否 | 文档版本号，`-1` = 最新（默认） |
 | `--scope` | 否 | `outline` \| `range` \| `keyword` \| `section`（省略 = 读整篇） |
--- a/skills/lark-doc/references/lark-doc-md.md
+++ b/skills/lark-doc/references/lark-doc-md.md
@@ -1,6 +1,6 @@
 # Markdown 格式参考

-`docs +fetch --api-version v2` / `docs +create --api-version v2` / `docs +update --api-version v2` 使用 `--doc-format markdown` 时适用。
+`docs +fetch --api-version v2` / `docs +create --api-version v2` / `docs +update --api-version v2` 使用 `--doc-format markdown` 时适用；fetch 的 `--doc-format im-markdown` 仅用于获取内容后在 `lark-im` 场景下使用，不作为 create/update 写入格式。

 ## 创建文档标题

--- a/skills/lark-im/SKILL.md
+++ b/skills/lark-im/SKILL.md
@@ -67,6 +67,10 @@ Card messages (`interactive` type) are not yet supported for compact conversion

 `--audio` sends a voice message and supports only Opus audio files, for example `.opus` files or Ogg Opus (`.ogg`) files. For `mp3`, `wav`, or other non-Opus audio, either convert to `.opus` first and keep using `--audio`, or send the original file as an attachment with `--file`.

+### Sending Doc Content as a Message
+
+When sending content fetched from a Lark doc as a message, fetch the doc with --doc-format im-markdown, then send it as a message using the --markdown format. The fetched content is already in markdown; in any content-forwarding scenario, keep the fetched original text and send it in the --markdown format. Note: if the doc contains a cite tag with type="user", keep it as-is and do not strip the tag.
+
 ### Flag Types

 Flags support two layers: