feat: add content sanitizer to strip hidden text before AI processing

[gentlemandev]

Summary

• New utility utils/ai/content-sanitizer.ts that strips invisible/hidden content from emails before passing to AI
• Prevents attackers from embedding invisible instructions in emails that the AI would process but users can't see

What it strips

• Zero-width Unicode: \u200B, \u200C, \u200D, \u2060, \uFEFF
• RTL/LTR overrides: \u202A-\u202E, \u2066-\u2069
• Hidden HTML elements: display:none, visibility:hidden, font-size:0, opacity:0
• White-on-white text: color:#fff/#ffffff/white/rgb(255,255,255)
• Offscreen positioning: position:absolute + left:-9999px
• Zero-dimension elements: width:0/height:0 + overflow:hidden
• HTML comments: <!-- ... -->

API

stripHiddenText(text: string): string      // for plain text
stripHiddenHtml(html: string): string      // for HTML emails
sanitizeForAI({ textPlain?, textHtml? })   // convenience wrapper

Not yet wired in

This PR adds the utility and tests. A follow-up PR will call sanitizeForAI() before aiChooseRule(), aiGenerateArgs(), and aiDraftReply().

Test plan

• 32 tests covering all hidden content types
• Edge cases: nested hidden elements, mixed attacks, legitimate HTML preservation
• Undefined/empty input handling

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
inbox-zero	Ignored		Apr 4, 2026 0:41am

[cubic-dev-ai]

2 issues found across 2 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="apps/web/utils/ai/content-sanitizer.ts">

<violation number="1" location="apps/web/utils/ai/content-sanitizer.ts:14">
P1: The `color` regex also matches `background-color`, causing false positives that remove visible content.</violation>

<violation number="2" location="apps/web/utils/ai/content-sanitizer.ts:46">
P2: `sanitizeForAI` drops empty-string inputs by using truthy checks; use an explicit `undefined` check so empty content is preserved.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P1: The color regex also matches background-color, causing false positives that remove visible content.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/web/utils/ai/content-sanitizer.ts, line 14:

<comment>The `color` regex also matches `background-color`, causing false positives that remove visible content.</comment>

<file context>
@@ -0,0 +1,87 @@
+  /visibility\s*:\s*hidden/i,
+  /font-size\s*:\s*0(?:px|em|rem|%|pt)?\s*[;"']/i,
+  /opacity\s*:\s*0\s*[;"']/i,
+  /color\s*:\s*(?:#fff(?:fff)?|white|rgb\(\s*255\s*,\s*255\s*,\s*255\s*\))\s*[;"']/i,
+];
+
</file context>

[cubic-dev-ai]

P2: sanitizeForAI drops empty-string inputs by using truthy checks; use an explicit undefined check so empty content is preserved.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At apps/web/utils/ai/content-sanitizer.ts, line 46:

<comment>`sanitizeForAI` drops empty-string inputs by using truthy checks; use an explicit `undefined` check so empty content is preserved.</comment>

<file context>
@@ -0,0 +1,87 @@
+  textHtml?: string;
+}): { textPlain?: string; textHtml?: string } {
+  return {
+    textPlain: input.textPlain ? stripHiddenText(input.textPlain) : undefined,
+    textHtml: input.textHtml ? stripHiddenHtml(input.textHtml) : undefined,
+  };
</file context>