fix: harden frontmatter fence sanitization

Author: nashsu

src/lib/ingest-sanitize.test.ts

@@ -35,6 +35,16 @@ describe("sanitizeIngestedFileContent", () => {
     expect(sanitizeIngestedFileContent(input)).toBe("---\ntype: x\n---\n\n# Body")
   })
 
+  it("strips a frontmatter fence after leading blank lines with a case-insensitive label", () => {
+    const input = "\n  \n```YAML\n---\ntype: x\n---\n```\n# Body"
+    expect(sanitizeIngestedFileContent(input)).toBe("---\ntype: x\n---\n# Body")
+  })
+
+  it("strips a CRLF fence around empty frontmatter", () => {
+    const input = "```yaml\r\n---\r\n---\r\n```\r\n\r\n# Body"
+    expect(sanitizeIngestedFileContent(input)).toBe("---\r\n---\r\n\r\n# Body")
+  })
+
   it("does NOT strip a non-fence-wrapped document containing a fenced code block in the body", () => {
     const input =
       "---\ntype: x\n---\n\n# Heading\n\n```js\nconsole.log('hi')\n```\n\nmore body"

src/lib/ingest-sanitize.ts

@@ -90,7 +90,9 @@ export function sanitizeIngestedFileContent(content: string): string {
 
 /** Top-level fence wrapper. Removes the open + matching close fence lines. */
 function stripOuterCodeFence(content: string): string {
-  const open = content.match(/^[ \t]*```(?:yaml|md|markdown)?[ \t]*\r?\n/)
+  const open = content.match(
+    /^(?:\uFEFF)?(?:[ \t]*\r?\n)*[ \t]*```(?:yaml|md|markdown)?[ \t]*\r?\n/i,
+  )
   if (!open) return content
   const afterOpen = content.slice(open[0].length)
 
@@ -103,7 +105,7 @@ function stripOuterCodeFence(content: string): string {
   // continue with an unfenced Markdown body. Only strip this shape when
   // the fenced section is exactly a complete `---` frontmatter block.
   const frontmatterOnly = afterOpen.match(
-    /^(---[ \t]*\r?\n[\s\S]*?\r?\n---[ \t]*\r?\n)[ \t]*```[ \t]*(?:\r?\n|$)/,
+    /^(---[ \t]*\r?\n[\s\S]*?^---[ \t]*\r?\n)[ \t]*```[ \t]*(?:\r?\n|$)/m,
   )
   if (!frontmatterOnly) return content
   return frontmatterOnly[1] + afterOpen.slice(frontmatterOnly[0].length)