test(ai-vault): add first-user-prompt UTF-16 safety tests

Ensure truncation at safety limits doesn't split UTF-16 surrogate pairs,
preventing corruption of astral characters in captured prompts.

src/main/ai-vault/session-scanner-first-user-prompt.test.ts

@@ -0,0 +1,39 @@
+import { describe, expect, it } from 'vitest'
+import { normalizeFullFirstUserPromptText } from './session-scanner-first-user-prompt'
+
+// Mirrors FULL_FIRST_USER_PROMPT_SAFETY_LIMIT in the module under test.
+const SAFETY_LIMIT = 256 * 1024
+
+describe('AI Vault full first-user-prompt normalization', () => {
+  it('drops an astral char straddling the safety limit instead of splitting it', () => {
+    const result = normalizeFullFirstUserPromptText(`${'a'.repeat(SAFETY_LIMIT - 1)}😀tail`)
+
+    expect(result).toHaveLength(SAFETY_LIMIT - 1)
+    expect(result?.endsWith('a')).toBe(true)
+    expect(hasUnpairedSurrogate(result ?? '')).toBe(false)
+  })
+
+  it('keeps a prompt shorter than the safety limit intact', () => {
+    expect(normalizeFullFirstUserPromptText('ship it 😀')).toBe('ship it 😀')
+  })
+})
+
+function hasUnpairedSurrogate(value: string): boolean {
+  for (let index = 0; index < value.length; index += 1) {
+    const code = value.charCodeAt(index)
+    const isHigh = code >= 0xd800 && code <= 0xdbff
+    const isLow = code >= 0xdc00 && code <= 0xdfff
+    if (isHigh) {
+      const next = value.charCodeAt(index + 1)
+      if (!(next >= 0xdc00 && next <= 0xdfff)) {
+        return true
+      }
+      index += 1
+      continue
+    }
+    if (isLow) {
+      return true
+    }
+  }
+  return false
+}

src/main/ai-vault/session-scanner-grok-parser.test.ts

@@ -41,6 +41,16 @@ describe('AI Vault Grok session parser', () => {
     expect(result?.endsWith('...')).toBe(true)
   })
 
+  it('drops an astral char straddling the preview scan cap instead of splitting it', () => {
+    // Hidden context is skipped by the fold, so the 4096-code-unit scan cap lands
+    // mid-emoji while the visible text stays well under the 220-char preview cap.
+    const hidden = `<system-reminder>${'x'.repeat(4057)}</system-reminder>`
+    const result = extractGrokContentText(`${hidden}ask😀tail`)
+
+    expect(hidden).toHaveLength(4092)
+    expect(result).toBe('ask')
+  })
+
   it('stores the unwrapped user_query as firstUserPrompt under full capture', async () => {
     const root = await mkdtemp(join(tmpdir(), 'orca-ai-vault-grok-first-'))
     tempRoots.push(root)