Fix/im markdown emphasis spacing

shortcuts/im/helpers.go

@@ -18,6 +18,8 @@ import (
 	"regexp"
 	"strconv"
 	"strings"
+	"unicode"
+	"unicode/utf8"
 
 	"github.com/larksuite/cli/extension/fileio"
 	"github.com/larksuite/cli/internal/output"
@@ -788,8 +790,9 @@ var (
 )
 
 const (
-	markdownCodeBlockPlaceholder = "___CB_"
-	postBlankLinePlaceholder     = "\u200B"
+	markdownCodeBlockPlaceholder  = "___CB_"
+	markdownInlineCodePlaceholder = "___IC_"
+	postBlankLinePlaceholder      = "\u200B"
 )
 
 type markdownPart struct {
@@ -816,8 +819,40 @@ func restoreMarkdownCodeBlocks(text string, codeBlocks []string) string {
 	return restored
 }
 
+func protectMarkdownInlineCode(text string) (string, []string) {
+	var inlineCodes []string
+	lines := strings.Split(text, "\n")
+	for i, line := range lines {
+		spans := scanInlineCodeSpans(line)
+		if len(spans) == 0 {
+			continue
+		}
+		var sb strings.Builder
+		pos := 0
+		for _, span := range spans {
+			sb.WriteString(line[pos:span[0]])
+			idx := len(inlineCodes)
+			inlineCodes = append(inlineCodes, line[span[0]:span[1]])
+			sb.WriteString(fmt.Sprintf("%s%d___", markdownInlineCodePlaceholder, idx))
+			pos = span[1]
+		}
+		sb.WriteString(line[pos:])
+		lines[i] = sb.String()
+	}
+	return strings.Join(lines, "\n"), inlineCodes
+}
+
+func restoreMarkdownInlineCode(text string, inlineCodes []string) string {
+	restored := text
+	for i, code := range inlineCodes {
+		restored = strings.Replace(restored, fmt.Sprintf("%s%d___", markdownInlineCodePlaceholder, i), code, 1)
+	}
+	return restored
+}
+
 func optimizeMarkdownStyle(text string) string {
 	r, codeBlocks := protectMarkdownCodeBlocks(text)
+	r = normalizeMarkdownEmphasisSpacing(r)
 
 	// Only downgrade when original text has H1~H3; order matters (H2~H6 first).
 	if reHasH1toH3.MatchString(text) {
@@ -849,6 +884,165 @@ func optimizeMarkdownStyle(text string) string {
 	return r
 }
 
+// normalizeMarkdownEmphasisSpacing trims whitespace immediately inside simple
+// *...*, **...**, and ***...*** spans while preserving fenced code blocks and
+// inline code. This hardens AI-generated markdown such as "** bold **" into
+// "**bold**" so Feishu's md renderer can recognize emphasis instead of
+// leaking literal '*'.
+func normalizeMarkdownEmphasisSpacing(markdown string) string {
+	protected, codeBlocks := protectMarkdownCodeBlocks(markdown)
+	protected, inlineCodes := protectMarkdownInlineCode(protected)
+
+	lines := strings.Split(protected, "\n")
+	for i, line := range lines {
+		lines[i] = normalizeEmphasisSpacingSegment(line)
+	}
+
+	normalized := strings.Join(lines, "\n")
+	normalized = restoreMarkdownInlineCode(normalized, inlineCodes)
+	return restoreMarkdownCodeBlocks(normalized, codeBlocks)
+}
+
+// scanInlineCodeSpans returns byte ranges [start, end) for inline code spans
+// using matching backtick runs, so emphasis normalization skips literal code.
+func scanInlineCodeSpans(line string) [][2]int {
+	var spans [][2]int
+	i := 0
+	for i < len(line) {
+		if line[i] != '`' {
+			i++
+			continue
+		}
+		start := i
+		for i < len(line) && line[i] == '`' {
+			i++
+		}
+		delim := line[start:i]
+		j := i
+		for j <= len(line)-len(delim) {
+			if line[j] == '`' {
+				k := j
+				for k < len(line) && line[k] == '`' {
+					k++
+				}
+				if k-j == len(delim) {
+					spans = append(spans, [2]int{start, k})
+					i = k
+					break
+				}
+				j = k
+			} else {
+				j++
+			}
+		}
+	}
+	return spans
+}
+
+func normalizeEmphasisSpacingSegment(seg string) string {
+	if !strings.Contains(seg, "*") {
+		return seg
+	}
+
+	var sb strings.Builder
+	pos := 0
+	for pos < len(seg) {
+		openStart, openEnd, ok := nextAsteriskRun(seg, pos)
+		if !ok {
+			sb.WriteString(seg[pos:])
+			break
+		}
+
+		sb.WriteString(seg[pos:openStart])
+
+		markerLen := openEnd - openStart
+		if markerLen != 1 && markerLen != 2 && markerLen != 3 {
+			sb.WriteString(seg[openStart:openEnd])
+			pos = openEnd
+			continue
+		}
+
+		closeStart, closeEnd, ok := nextAsteriskRun(seg, openEnd)
+		if !ok || closeEnd-closeStart != markerLen {
+			sb.WriteString(seg[openStart:openEnd])
+			pos = openEnd
+			continue
+		}
+		if !hasSimpleEmphasisBoundaries(seg, openStart, closeEnd) {
+			sb.WriteString(seg[openStart:closeEnd])
+			pos = closeEnd
+			continue
+		}
+
+		payload := seg[openEnd:closeStart]
+		normalized, shouldNormalize := normalizeEmphasisPayload(payload)
+		if !shouldNormalize {
+			sb.WriteString(seg[openStart:closeEnd])
+			pos = closeEnd
+			continue
+		}
+
+		marker := seg[openStart:openEnd]
+		sb.WriteString(marker)
+		sb.WriteString(normalized)
+		sb.WriteString(marker)
+		pos = closeEnd
+	}
+	return sb.String()
+}
+
+func nextAsteriskRun(s string, start int) (runStart, runEnd int, ok bool) {
+	for i := start; i < len(s); i++ {
+		if s[i] != '*' {
+			continue
+		}
+		j := i
+		for j < len(s) && s[j] == '*' {
+			j++
+		}
+		return i, j, true
+	}
+	return 0, 0, false
+}
+
+func hasSimpleEmphasisBoundaries(s string, openStart, closeEnd int) bool {
+	if openStart > 0 {
+		prev, _ := utf8.DecodeLastRuneInString(s[:openStart])
+		if isWordLikeRune(prev) {
+			return false
+		}
+	}
+	if closeEnd < len(s) {
+		next, _ := utf8.DecodeRuneInString(s[closeEnd:])
+		if isWordLikeRune(next) {
+			return false
+		}
+	}
+	return true
+}
+
+func isWordLikeRune(r rune) bool {
+	return unicode.IsLetter(r) || unicode.IsDigit(r) || unicode.Is(unicode.Han, r)
+}
+
+func normalizeEmphasisPayload(payload string) (string, bool) {
+	trimmedLeft := strings.TrimLeftFunc(payload, unicode.IsSpace)
+	trimmed := strings.TrimRightFunc(trimmedLeft, unicode.IsSpace)
+	if trimmed == "" {
+		return payload, false
+	}
+
+	hasLeadingSpace := len(trimmedLeft) != len(payload)
+	hasTrailingSpace := len(trimmed) != len(trimmedLeft)
+	if !hasLeadingSpace && !hasTrailingSpace {
+		return payload, false
+	}
+	if strings.Contains(trimmed, "*") {
+		return payload, false
+	}
+	return trimmed, true
+}
+
 func shouldUseSegmentedPost(markdown string) bool {
 	protected, _ := protectMarkdownCodeBlocks(markdown)
 	return reBlankLineSeparator.MatchString(protected)

shortcuts/im/helpers_test.go

@@ -519,6 +519,50 @@ func TestWrapMarkdownAsPost(t *testing.T) {
 		}
 	})
 
+	t.Run("normalizes malformed bold spacing", func(t *testing.T) {
+		got := wrapMarkdownAsPost("hello ** world **")
+		node := decodePostParagraphForTest(t, got, 0)
+		if node["text"] != "hello **world**" {
+			t.Fatalf("wrapMarkdownAsPost() text = %#v, want %q", node["text"], "hello **world**")
+		}
+	})
+
+	t.Run("normalizes malformed italic and bold italic spacing", func(t *testing.T) {
+		got := wrapMarkdownAsPost("* italic * and *** both ***")
+		node := decodePostParagraphForTest(t, got, 0)
+		if node["text"] != "*italic* and ***both***" {
+			t.Fatalf("wrapMarkdownAsPost() text = %#v, want %q", node["text"], "*italic* and ***both***")
+		}
+	})
+
+	t.Run("preserves inline code and fenced code spacing", func(t *testing.T) {
+		input := "code `** keep **` and prose ** fix **\n```md\n** keep fenced **\n```"
+		got := wrapMarkdownAsPost(input)
+		node := decodePostParagraphForTest(t, got, 0)
+		text, _ := node["text"].(string)
+		if text != "code `** keep **` and prose **fix**\n```md\n** keep fenced **\n```" {
+			t.Fatalf("wrapMarkdownAsPost() text = %#v", node["text"])
+		}
+	})
+
+	t.Run("preserves multiple inline code spans", func(t *testing.T) {
+		input := "`** keep **` prose ** fix ** ``* keep *`` and * okay *"
+		got := wrapMarkdownAsPost(input)
+		node := decodePostParagraphForTest(t, got, 0)
+		text, _ := node["text"].(string)
+		if text != "`** keep **` prose **fix** ``* keep *`` and *okay*" {
+			t.Fatalf("wrapMarkdownAsPost() text = %#v", node["text"])
+		}
+	})
+
+	t.Run("preserves embedded non emphatic spaced asterisks", func(t *testing.T) {
+		got := wrapMarkdownAsPost("hello** world **there")
+		node := decodePostParagraphForTest(t, got, 0)
+		if node["text"] != "hello** world **there" {
+			t.Fatalf("wrapMarkdownAsPost() text = %#v", node["text"])
+		}
+	})
+
 	t.Run("bare URL becomes a tag", func(t *testing.T) {
 		got := wrapMarkdownAsPost("see https://example.com/flow_id=abc_def done")
 		if !strings.Contains(got, `"tag":"a"`) {

skills/lark-im/references/lark-im-messages-reply.md

@@ -64,6 +64,7 @@ So `--markdown` is a convenience mode, not a full Markdown compatibility layer.
 - Block spacing and line breaks may be normalized during conversion.
 - Code blocks are preserved as code blocks.
 - Excess blank lines are compressed.
+- Emphasis with `*`, `**`, or `***` should not contain inner edge spaces. Use `**bold**` instead of `** bold **`, `** bold**`, or `**bold **`; use `*italic*` instead of `* italic *`, `* italic*`, or `*italic *`.
 - Only remote `http://...`, `https://...`, or already-uploaded `img_xxx` Markdown images are kept reliably.
 - Local paths in Markdown image syntax like `![x](./a.png)` are **not** auto-uploaded by `--markdown`.
 - If remote Markdown image handling fails, that image is removed with a warning.

skills/lark-im/references/lark-im-messages-send.md

@@ -64,6 +64,7 @@ This means `--markdown` is convenient, but it is not a full-fidelity Markdown tr
 - Block spacing and line breaks may be normalized during conversion.
 - Code blocks are preserved as code blocks.
 - Excess blank lines are compressed.
+- Emphasis with `*`, `**`, or `***` should not contain inner edge spaces. Use `**bold**` instead of `** bold **`, `** bold**`, or `**bold **`; use `*italic*` instead of `* italic *`, `* italic*`, or `*italic *`.
 - Only `http://...`, `https://...`, or already-uploaded `img_xxx` Markdown images are kept reliably.
 - Local paths in Markdown image syntax like `![x](./a.png)` are **not** auto-uploaded by `--markdown`; they may be stripped during optimization.
 - If remote Markdown image download/upload fails, that image is removed with a warning.