fix: exclude review summary card from copy/export output

Polish review caught a cross-feature bug: the agent review-doc summary card
(#24) renders as an <aside class="review-card"> inside .markdown-body, which
both the rich-text copy and the HTML/PDF/DOCX export capture wholesale — so a
copied/exported review doc embedded the synthesized card (badges, chips) that
was never in the source. Add cloneWithoutInjectedChrome() to strip injected UI
chrome from both capture paths; +2 tests. (A second reported issue — duplicate
React keys in the card — was a false positive: the parser emits at most one
finding per line, so f.line is unique.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: masonwyatt23

src/components/viewer/ReviewSummaryCard.tsx

@@ -39,6 +39,8 @@ interface ReviewSummaryCardProps {
 function useFindingAnchors(summary: ReviewSummary): Map<number, string> {
   return useMemo(() => {
     const slugger = new GithubSlugger();
+    // Keyed by source line, which is unique per finding (the parser emits at
+    // most one finding per line).
     const map = new Map<number, string>();
     for (const f of summary.findings) {
       if (!f.title) continue;

src/lib/copyRichText.test.ts

@@ -1,5 +1,27 @@
 import { describe, expect, it } from "vitest";
 import { wrapRichTextBody } from "./copyRichText";
+import { cloneWithoutInjectedChrome } from "./export";
+
+describe("cloneWithoutInjectedChrome", () => {
+  it("strips the injected review card but keeps document content", () => {
+    const body = document.createElement("div");
+    body.className = "markdown-body";
+    body.innerHTML =
+      '<aside class="review-card">SUMMARY</aside><h1>Real Title</h1><p>Body text</p>';
+    const cleaned = cloneWithoutInjectedChrome(body);
+    expect(cleaned.querySelector(".review-card")).toBeNull();
+    expect(cleaned.innerHTML).toContain("<h1>Real Title</h1>");
+    expect(cleaned.innerHTML).toContain("Body text");
+    // The original element is untouched (we clone, not mutate in place).
+    expect(body.querySelector(".review-card")).not.toBeNull();
+  });
+
+  it("is a no-op for a body with no injected chrome", () => {
+    const body = document.createElement("div");
+    body.innerHTML = "<h1>Doc</h1><p>x</p>";
+    expect(cloneWithoutInjectedChrome(body).innerHTML).toBe("<h1>Doc</h1><p>x</p>");
+  });
+});
 
 describe("wrapRichTextBody", () => {
   it("wraps body markup in an inline-styled div", () => {

src/lib/copyRichText.ts

@@ -22,7 +22,7 @@
 import DOMPurify from "dompurify";
 import { useDocumentStore } from "../store/documentStore";
 import { toast } from "../store/toastStore";
-import { buildStandaloneHtml } from "./export";
+import { buildStandaloneHtml, cloneWithoutInjectedChrome } from "./export";
 import { waitForElement } from "./waitForElement";
 
 /**
@@ -73,7 +73,10 @@ export function buildRichTextHtml(bodyInnerHtml: string): string {
  */
 function captureRenderedBodyHtml(): string {
   const el = document.querySelector(".markdown-body");
-  if (el) return el.innerHTML;
+  // Strip injected chrome (the review card) so the clipboard gets the document,
+  // not Ashlr's rendering overlay. The fallback below routes through
+  // buildStandaloneHtml → captureMarkdownBody, which already strips it.
+  if (el) return cloneWithoutInjectedChrome(el).innerHTML;
 
   // Fallback: reuse the export pipeline, then extract just the rendered body.
   try {

src/lib/export.ts

@@ -39,6 +39,18 @@ function currentLayoutVars(): string {
   return `--content-width:${width};--content-font-size:${fontSize};`;
 }
 
+/**
+ * Clone an element and strip Ashlr-injected UI chrome that lives inside
+ * `.markdown-body` but is NOT part of the document — currently the review
+ * summary card. Exports and rich-text copies should contain the document, not
+ * the app's rendering overlay.
+ */
+export function cloneWithoutInjectedChrome(el: Element): Element {
+  const clone = el.cloneNode(true) as Element;
+  for (const node of clone.querySelectorAll(".review-card")) node.remove();
+  return clone;
+}
+
 /**
  * Reads the live `.markdown-body` element and returns its `outerHTML`.
  * Throws a descriptive string (shown in the dialog) when the element is
@@ -49,7 +61,7 @@ function captureMarkdownBody(): string {
   if (!el) {
     throw "Switch to Read view before exporting.";
   }
-  return el.outerHTML;
+  return cloneWithoutInjectedChrome(el).outerHTML;
 }
 
 /**