euparliamentmonitor 0.9.20 → 0.9.22

package/README.md

@@ -136,7 +136,7 @@ The published site is the audience-facing companion to this npm/TypeScript packa
 
 **MCP Server Integration**: The project uses the
 [European-Parliament-MCP-Server](https://github.com/Hack23/European-Parliament-MCP-Server)
-v1.3.9 for accessing real EU Parliament data via the Model Context Protocol.
+v1.3.10 for accessing real EU Parliament data via the Model Context Protocol.
 
 - **MCP Server Status**: ✅ Fully operational — 60+ EP data tools available
   (feeds, direct lookups, analytical tools, intelligence correlation)
@@ -432,7 +432,7 @@ import type { ArticleCategory, LanguageCode } from 'euparliamentmonitor/types';
 
 ## 🔌 Data Sources
 
-**Primary — European Parliament MCP Server** ([Hack23/European-Parliament-MCP-Server](https://github.com/Hack23/European-Parliament-MCP-Server) v1.3.9+, fully operational):
+**Primary — European Parliament MCP Server** ([Hack23/European-Parliament-MCP-Server](https://github.com/Hack23/European-Parliament-MCP-Server) v1.3.10+, fully operational):
 
 - 🗳️ Plenary sessions, voting records, roll-call votes
 - 📜 Adopted texts, motions, resolutions, urgency files

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "euparliamentmonitor",
-  "version": "0.9.20",
+  "version": "0.9.22",
   "type": "module",
   "description": "European Parliament Intelligence Platform - Monitor political activity with systematic transparency",
   "main": "scripts/index.js",
@@ -71,6 +71,7 @@
     "prior-run-diff": "node scripts/aggregator/prior-run-diff.js",
     "generate-article": "node scripts/aggregator/article-generator.js",
     "generate-article:all": "node scripts/aggregator/article-generator.js --all",
+    "dump:article-seo": "node scripts/dump-article-seo.js",
     "generate-news-indexes": "node scripts/generators/news-indexes.js",
     "generate-sitemap": "node scripts/generators/sitemap.js",
     "image:generate": "node scripts/generate-responsive-images.js",
@@ -164,6 +165,7 @@
     "chartjs-plugin-annotation": "3.1.0",
     "clean-css": "^5.3.3",
     "d3": "7.9.0",
+    "esbuild": "0.28.0",
     "eslint": "10.4.0",
     "eslint-config-prettier": "10.1.8",
     "eslint-plugin-jsdoc": "63.0.0",
@@ -194,7 +196,7 @@
     "node": ">=26"
   },
   "dependencies": {
-    "european-parliament-mcp-server": "1.3.9",
+    "european-parliament-mcp-server": "1.3.10",
     "markdown-it": "^14.1.1",
     "markdown-it-anchor": "^9.2.0",
     "markdown-it-attrs": "^4.3.1",
@@ -208,6 +210,7 @@
     "flatted": ">=3.4.2",
     "path-to-regexp": ">=8.4.0",
     "ip-address": ">=10.1.1",
-    "uuid": ">=11.1.1"
+    "uuid": ">=11.1.1",
+    "qs": "6.15.2"
   }
 }

package/scripts/aggregator/editorial-brief-resolver.d.ts

@@ -73,4 +73,42 @@ export declare function resolveLocalizedBriefHighlight(runDir: string, lang: Lan
  * brief candidate file exists
  */
 export declare function discoverLocalizedBriefs(runDir: string, languages: readonly LanguageCode[]): readonly LanguageCode[];
+/**
+ * Localized brief body suitable for HTML rendering.
+ *
+ * Unlike {@link resolveLocalizedBriefHighlight} — which extracts a few
+ * short SEO/metadata fields (headline, summary) for `<meta>` tags and
+ * JSON-LD — this helper returns the **full body** of the translated
+ * executive brief, with the SPDX preamble stripped, so the caller can
+ * render it through {@link renderMarkdown} and splice the resulting
+ * HTML into the per-language article variant.
+ *
+ * Used by the article-generator HTML pipeline (`render-one.ts`) to
+ * upgrade non-English variants from the English aggregated body to a
+ * truly localized one whenever a translated `executive-brief_<lang>.md`
+ * exists in the run directory.
+ */
+export interface LocalizedBriefBody {
+    /** Markdown body of the localized brief (post-SPDX strip). */
+    readonly markdown: string;
+    /** Run-relative path of the file that produced {@link markdown}. */
+    readonly sourceFile: string;
+}
+/**
+ * Read the **full markdown body** of a translated executive brief for
+ * `lang` from `runDir`, searching the standard candidate paths
+ * (`executive-brief_<lang>.md` → `extended/executive-brief_<lang>.md`).
+ * SPDX HTML-comment preambles are stripped using the same logic as the
+ * SEO-metadata path, so the returned markdown starts at the first real
+ * content line (`# Headline` or similar).
+ *
+ * Returns `null` when `runDir` is missing, the language is English, or
+ * no candidate file exists. The caller is expected to fall back to the
+ * English aggregated body in that case — see `render-one.ts`.
+ *
+ * @param runDir - Absolute run directory
+ * @param lang - Target language code (omitted when `lang === 'en'`)
+ * @returns Localized brief body + source file, or `null` when absent
+ */
+export declare function readLocalizedBriefBody(runDir: string, lang: LanguageCode): LocalizedBriefBody | null;
 //# sourceMappingURL=editorial-brief-resolver.d.ts.map

package/scripts/aggregator/editorial-brief-resolver.js

@@ -217,4 +217,36 @@ export function discoverLocalizedBriefs(runDir, languages) {
     }
     return out;
 }
+/**
+ * Read the **full markdown body** of a translated executive brief for
+ * `lang` from `runDir`, searching the standard candidate paths
+ * (`executive-brief_<lang>.md` → `extended/executive-brief_<lang>.md`).
+ * SPDX HTML-comment preambles are stripped using the same logic as the
+ * SEO-metadata path, so the returned markdown starts at the first real
+ * content line (`# Headline` or similar).
+ *
+ * Returns `null` when `runDir` is missing, the language is English, or
+ * no candidate file exists. The caller is expected to fall back to the
+ * English aggregated body in that case — see `render-one.ts`.
+ *
+ * @param runDir - Absolute run directory
+ * @param lang - Target language code (omitted when `lang === 'en'`)
+ * @returns Localized brief body + source file, or `null` when absent
+ */
+export function readLocalizedBriefBody(runDir, lang) {
+    if (!runDir || lang === 'en')
+        return null;
+    if (!fs.existsSync(runDir))
+        return null;
+    for (const rel of localizedBriefCandidates(lang)) {
+        const abs = path.join(runDir, rel);
+        if (!fs.existsSync(abs))
+            continue;
+        const body = readArtefactBody(abs);
+        if (body.trim().length === 0)
+            continue;
+        return { markdown: body, sourceFile: rel };
+    }
+    return null;
+}
 //# sourceMappingURL=editorial-brief-resolver.js.map

package/scripts/aggregator/generator/render-one.js

@@ -16,6 +16,11 @@ import { resolveArticleMetadata, extractStrongProseLine, } from '../article-meta
 import { buildArticleMeta, serializeArticleMeta } from '../article-meta.js';
 import { renderMarkdown } from '../markdown-renderer.js';
 import { wrapArticleHtml, getArticleFilename, localizeArticleBody, enhanceTradecraftCards, enhanceAnalysisIndexCards, } from '../article-html.js';
+import { replaceExecutiveBriefSection } from '../html/localize-body.js';
+import { readLocalizedBriefBody } from '../editorial-brief-resolver.js';
+import { extractRunMentions } from '../seo-entity-extractor.js';
+import { SECTION_TITLE_LABELS } from '../../constants/ui/related-analysis.js';
+import { getLocalizedString } from '../../constants/language-core.js';
 import { buildReaderIntelligenceGuideHtml, stripInlineReaderGuide, } from '../reader-intelligence-guide.js';
 import { ALL_LANGUAGES } from '../../constants/language-core.js';
 import { blobUrl } from '../infra/github-urls.js';
@@ -84,6 +89,10 @@ function buildJekyllArticleMarkdown(aggregated, metadata, slug, sourceFolder) {
  *        canonical English Markdown source written by the same run
  * @param chromeOptions.articleCount - Total article count surfaced in the
  *        site footer's `<p class="footer-stats">…</p>` line
+ * @param chromeOptions.mentions - SEO `mentions` list (organization names
+ *        extracted from `intelligence/stakeholder-map.md` and
+ *        `extended/media-framing-analysis.md`) emitted into JSON-LD on
+ *        every language variant
  * @param opts - CLI options (needed for `outDir`)
  * @returns Relative filename of the HTML file written
  */
@@ -96,6 +105,30 @@ function writeLanguageVariant(lang, slug, aggregated, englishHtml, chromeOptions
         metaSource = fs.readFileSync(langMdAbs, 'utf8');
         bodyHtml = renderMarkdown(metaSource).html;
     }
+    else if (lang !== 'en') {
+        // No full per-language source markdown — but the run may still
+        // ship a translated `executive-brief_<lang>.md`. When present,
+        // splice its rendered HTML into the `#section-executive-brief`
+        // block so non-English readers see localized BLUF + key findings
+        // instead of English fallback prose. SEO metadata (`<title>`,
+        // `<meta description>`, JSON-LD `headline`) is already localized
+        // via `resolveLocalizedBriefHighlight` upstream, so this hook
+        // exclusively touches the rendered article body.
+        const localized = opts.runDir !== null ? readLocalizedBriefBody(opts.runDir, lang) : null;
+        if (localized) {
+            const localizedRendered = renderMarkdown(localized.markdown).html;
+            // Strip the first H1 from the translated brief —
+            // `replaceExecutiveBriefSection` re-emits the canonical
+            // `<h2 id="section-executive-brief">…</h2>` heading itself,
+            // and the brief's own `# Headline` is duplicate chrome.
+            const briefBodyHtml = localizedRendered.replace(/<h1[^>]*>[\s\S]*?<\/h1>\s*/, '');
+            const briefHeadingMap = SECTION_TITLE_LABELS['executive-brief'];
+            const localizedHeading = briefHeadingMap
+                ? getLocalizedString(briefHeadingMap, lang)
+                : 'Executive Brief';
+            bodyHtml = replaceExecutiveBriefSection(bodyHtml, localizedHeading, briefBodyHtml);
+        }
+    }
     bodyHtml = stripInlineReaderGuide(bodyHtml);
     bodyHtml = bodyHtml.replace(/<h1[^>]*>[\s\S]*?<\/h1>\s*/, '');
     const guideHtml = buildReaderIntelligenceGuideHtml(lang, aggregated.sectionToc, aggregated.includedArtifacts);
@@ -123,6 +156,7 @@ function writeLanguageVariant(lang, slug, aggregated, englishHtml, chromeOptions
         toc: aggregated.sectionToc,
         articleCount: chromeOptions.articleCount,
         isBasedOn: aggregated.includedArtifacts.map((a) => blobUrl(a.repoRelPath)),
+        mentions: chromeOptions.mentions,
     });
     const filename = getArticleFilename(slug, lang);
     fs.writeFileSync(path.join(opts.outDir, filename), html, 'utf8');
@@ -236,6 +270,7 @@ export function generateArticle(opts, runSuffix, articleCountOverride) {
             metadata: effectiveMetadata,
             sourceMarkdownRelPath: runArticleMdRelPath,
             articleCount: articleCountOverride ?? countPublishedArticles(opts.repoRoot),
+            mentions: opts.runDir ? extractRunMentions(opts.runDir) : [],
         };
         for (const lang of opts.langs) {
             const filename = writeLanguageVariant(lang, slug, aggregated, rendered.html, chromeOptions, opts);

package/scripts/aggregator/html/localize-body.d.ts

@@ -22,6 +22,38 @@ export declare function localizeArticleBody(bodyHtml: string, lang: LanguageCode
  * @returns Modified string, or `haystack` unchanged when `needle` is absent
  */
 export declare function replaceFirstStringIn(haystack: string, needle: string, replacement: string): string;
+/**
+ * Replace the **inner body** of the Executive Brief section (the
+ * `<h2 id="section-executive-brief">…</h2>` heading and everything that
+ * follows it up to — but not including — the next `<h2 id="section-…">`
+ * sibling) with the supplied replacement HTML. The Executive Brief
+ * heading itself is preserved by emitting it inline ahead of the
+ * replacement, so the in-page anchor (`#section-executive-brief`) and
+ * the table-of-contents link continue to work.
+ *
+ * Used by the article-generator HTML pipeline to inject the rendered
+ * markdown of a translated `executive-brief_<lang>.md` into the
+ * non-English language variants without forking the whole aggregated
+ * article into 14 source-language copies — see
+ * `editorial-brief-resolver.readLocalizedBriefBody` and
+ * `render-one.writeLanguageVariant`.
+ *
+ * Implementation uses `indexOf`/slice exclusively to stay within
+ * CodeQL's safe-regex envelope. Returns `html` unchanged when the
+ * Executive Brief heading is absent or malformed.
+ *
+ * @param html - Full article body HTML
+ * @param localizedHeading - Localized text for the Executive Brief H2
+ *                           (e.g. `"Sammanfattning"` for `sv`). Must be
+ *                           plain text — caller is responsible for any
+ *                           escaping (it's passed through `escapeHTML`).
+ * @param replacementBodyHtml - HTML to splice in **after** the heading.
+ *                              Should not contain its own `<h2>` for
+ *                              the Executive Brief — the heading is
+ *                              re-emitted by this helper.
+ * @returns Updated HTML with the localized brief body in place.
+ */
+export declare function replaceExecutiveBriefSection(html: string, localizedHeading: string, replacementBodyHtml: string): string;
 /**
  * Replace an H2 heading's text content by locating it via its `id` attribute.
  * Uses indexOf-based search to avoid polynomial regex backtracking (CodeQL).

package/scripts/aggregator/html/localize-body.js

@@ -102,6 +102,75 @@ export function replaceFirstStringIn(haystack, needle, replacement) {
         return haystack;
     return haystack.slice(0, idx) + replacement + haystack.slice(idx + needle.length);
 }
+/**
+ * Replace the **inner body** of the Executive Brief section (the
+ * `<h2 id="section-executive-brief">…</h2>` heading and everything that
+ * follows it up to — but not including — the next `<h2 id="section-…">`
+ * sibling) with the supplied replacement HTML. The Executive Brief
+ * heading itself is preserved by emitting it inline ahead of the
+ * replacement, so the in-page anchor (`#section-executive-brief`) and
+ * the table-of-contents link continue to work.
+ *
+ * Used by the article-generator HTML pipeline to inject the rendered
+ * markdown of a translated `executive-brief_<lang>.md` into the
+ * non-English language variants without forking the whole aggregated
+ * article into 14 source-language copies — see
+ * `editorial-brief-resolver.readLocalizedBriefBody` and
+ * `render-one.writeLanguageVariant`.
+ *
+ * Implementation uses `indexOf`/slice exclusively to stay within
+ * CodeQL's safe-regex envelope. Returns `html` unchanged when the
+ * Executive Brief heading is absent or malformed.
+ *
+ * @param html - Full article body HTML
+ * @param localizedHeading - Localized text for the Executive Brief H2
+ *                           (e.g. `"Sammanfattning"` for `sv`). Must be
+ *                           plain text — caller is responsible for any
+ *                           escaping (it's passed through `escapeHTML`).
+ * @param replacementBodyHtml - HTML to splice in **after** the heading.
+ *                              Should not contain its own `<h2>` for
+ *                              the Executive Brief — the heading is
+ *                              re-emitted by this helper.
+ * @returns Updated HTML with the localized brief body in place.
+ */
+export function replaceExecutiveBriefSection(html, localizedHeading, replacementBodyHtml) {
+    const idMarker = 'id="section-executive-brief"';
+    const idIdx = html.indexOf(idMarker);
+    if (idIdx === -1)
+        return html;
+    // Walk back to the opening `<h2` of the Executive Brief heading.
+    const h2Open = html.lastIndexOf('<h2', idIdx);
+    if (h2Open === -1)
+        return html;
+    // Find the end of the heading element.
+    const h2CloseTagIdx = html.indexOf('</h2>', idIdx);
+    if (h2CloseTagIdx === -1)
+        return html;
+    const afterHeading = h2CloseTagIdx + '</h2>'.length;
+    // Find the next `<h2 id="section-...">` boundary — the start of the
+    // following article section. If there is no further section heading
+    // we conservatively bail out (replacing through end-of-body would
+    // also drop appendix content like Reader Guide / Key Takeaways).
+    const nextSectionId = html.indexOf('id="section-', afterHeading);
+    if (nextSectionId === -1)
+        return html;
+    const nextH2 = html.lastIndexOf('<h2', nextSectionId);
+    if (nextH2 === -1 || nextH2 <= afterHeading)
+        return html;
+    // Find the start of the line containing the next `<h2` so we don't
+    // strip leading whitespace from the next section. We look at most
+    // one newline back.
+    let cutEnd = nextH2;
+    const prevNewline = html.lastIndexOf('\n', nextH2 - 1);
+    if (prevNewline !== -1 && prevNewline >= afterHeading) {
+        cutEnd = prevNewline + 1;
+    }
+    const newHeading = `<h2 id="section-executive-brief">${escapeHTML(localizedHeading)}</h2>\n`;
+    const trimmedReplacement = replacementBodyHtml.endsWith('\n')
+        ? replacementBodyHtml
+        : `${replacementBodyHtml}\n`;
+    return html.slice(0, h2Open) + newHeading + trimmedReplacement + html.slice(cutEnd);
+}
 /**
  * Replace an H2 heading's text content by locating it via its `id` attribute.
  * Uses indexOf-based search to avoid polynomial regex backtracking (CodeQL).

package/scripts/aggregator/html/shell.d.ts

@@ -59,6 +59,16 @@ export interface WrapArticleOptions {
      * Emitted as `isBasedOn` in the JSON-LD `NewsArticle` schema for provenance.
      */
     readonly isBasedOn?: readonly string[];
+    /**
+     * Optional: real-world organizations (political groups, media outlets,
+     * institutions) named in the article's intelligence and media-framing
+     * artifacts. Emitted as JSON-LD `mentions` Organization entries to give
+     * search engines and AI overviews high-precision entity grounding.
+     * Currently only extractable from the English intelligence corpus; the
+     * same list is reused across every language variant because the entities
+     * are language-independent proper nouns.
+     */
+    readonly mentions?: readonly string[];
 }
 /**
  * Render the full article HTML document with the shared chrome.

package/scripts/aggregator/html/shell.js

@@ -23,6 +23,7 @@ import { getSitemapFilename } from '../../generators/sitemap/index.js';
 import { truncateHeadline, getTitleSeparator, getLocalizedArticleType, getLocalizedArticleTypePlain, } from './headline.js';
 import { getArticleFilename, buildArticleHreflangLinks, buildLanguageSwitcher, } from './hreflang.js';
 import { buildArticleToc } from './toc.js';
+import { blobUrl } from '../infra/github-urls.js';
 /** Publisher organization name used in JSON-LD, meta tags. */
 export const PUBLISHER_NAME = 'Hack23 AB';
 /** Site name used across meta tags and structured data. */
@@ -50,8 +51,9 @@ export function wrapArticleHtml(options) {
     const sitemapLabel = getLocalizedString(FOOTER_SITEMAP_LABELS, safeLang);
     const politicalIntelligenceHref = `../${getPoliticalIntelligenceFilename(safeLang)}`;
     const sitemapHref = `../${getSitemapFilename(safeLang)}`;
+    const sourceMdHref = options.sourceMarkdownRelPath ? blobUrl(options.sourceMarkdownRelPath) : '';
     const sourceMdLink = options.sourceMarkdownRelPath
-        ? `<p class="article-source-md"><a href="${BASE_URL}/${options.sourceMarkdownRelPath}" rel="alternate" type="text/markdown"><svg class="icon icon-inline" width="16" height="16" viewBox="0 0 24 24" role="img" aria-hidden="true" focusable="false"><path d="M9 5H7a2 2 0 0 0-2 2v10a2 2 0 0 0 2 2h10a2 2 0 0 0 2-2v-2M12 3h6a2 2 0 0 1 2 2v6M10 14 20 4" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round"/></svg> ${escapeHTML(sourceMdLabel)}</a></p>`
+        ? `<p class="article-source-md"><a href="${escapeHTML(sourceMdHref)}" rel="alternate" type="text/markdown"><svg class="icon icon-inline" width="16" height="16" viewBox="0 0 24 24" role="img" aria-hidden="true" focusable="false"><path d="M9 5H7a2 2 0 0 0-2 2v10a2 2 0 0 0 2 2h10a2 2 0 0 0 2-2v-2M12 3h6a2 2 0 0 1 2 2v6M10 14 20 4" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round"/></svg> ${escapeHTML(sourceMdLabel)}</a></p>`
         : '';
     const tocHtml = buildArticleToc(options.toc ?? [], safeLang);
     const articleMainClass = tocHtml.length > 0 ? 'article-main--with-toc' : 'article-main--no-toc';
@@ -128,6 +130,14 @@ export function wrapArticleHtml(options) {
                 isBasedOn: options.isBasedOn.map((url) => ({ '@type': 'CreativeWork', url })),
             }
             : {}),
+        ...(options.mentions && options.mentions.length > 0
+            ? {
+                mentions: options.mentions.map((name) => ({
+                    '@type': 'Organization',
+                    name,
+                })),
+            }
+            : {}),
     };
     const breadcrumbLd = {
         '@context': 'https://schema.org',

package/scripts/aggregator/markdown-renderer.d.ts

@@ -67,6 +67,26 @@ export declare function stripMarkdownFrontMatter(markdown: string): string;
  * @returns Slug of up to 80 ASCII-ish characters, with dashes as separators
  */
 export declare function slugify(text: string): string;
+/**
+ * Decode the small set of HTML entities that Markdown authors (and
+ * upstream generators) occasionally pre-encode inside fenced mermaid
+ * blocks — typically & for & in political-group labels like
+ * S&D or Greens/EFA. Without this decode step, the subsequent
+ * escapeHtml pass would re-escape & to & and emit
+ * S&D into the rendered HTML, which the Mermaid client
+ * library then renders verbatim instead of as S&D.
+ *
+ * Uses indexOf/split/join exclusively (no RegExp) to stay
+ * within CodeQL's safe-regex envelope. Only the canonical entity
+ * forms are decoded — anything more exotic (e.g. &) is left
+ * alone so we never accidentally swallow a literal that the author
+ * intended to keep encoded.
+ *
+ * @param content - Raw fenced-block content (post-sanitizeMermaidQuadrantChart)
+ * @returns Content with pre-encoded HTML entities normalised back to
+ *          their literal characters, ready for a single escapeHtml.
+ */
+export declare function decodeMermaidPreEncodedEntities(content: string): string;
 /**
  * Auto-quote unquoted `quadrantChart` labels so the Mermaid v11 lexer
  * accepts them. The Mermaid `quadrantChart` grammar treats unquoted
@@ -74,30 +94,9 @@ export declare function slugify(text: string): string;
  * en-dashes (`–`, U+2013), ellipsis (`…`), parentheses, colons, and
  * non-ASCII currency symbols (`€`) all trigger
  * `Lexical error … Unrecognized text` and prevent the diagram from
- * rendering, leaving the raw `<pre>` source visible on the page.
- *
- * The style guide already instructs authors to wrap every quadrant /
- * axis / data-point label in double quotes (see
- * `analysis/methodologies/political-style-guide.md` § Standard
- * `quadrantChart` init block), but AI-generated `article.md` files
- * occasionally drop the quoting. Rather than reject the article at
- * Stage C we sanitize at the renderer boundary so every published
- * HTML page renders, regardless of upstream authoring discipline.
- *
- * Sanitization is deliberately scoped to `quadrantChart` blocks —
- * `flowchart`, `sequenceDiagram`, `mindmap`, `pie`, `gantt`, and
- * `xychart-beta` accept the same Unicode characters in their unquoted
- * labels and are passed through unchanged.
- *
- * Lines normalised:
- *   - `x-axis Left --> Right`     → `x-axis "Left" --> "Right"`
- *   - `y-axis Low --> High`       → `y-axis "Low" --> "High"`
- *   - `quadrant-N Label text`     → `quadrant-N "Label text"`
- *   - `Data Label: [x, y]`        → `"Data Label": [x, y]`
- *
- * Already-quoted operands are preserved byte-for-byte. The `title`
- * line, the `%%{init:…}%%` directive, and any line not matching one
- * of the recognised shapes are also left untouched.
+ * rendering. Sanitization is scoped to `quadrantChart` blocks only;
+ * other diagram types accept those characters in unquoted labels and
+ * are passed through unchanged.
  *
  * @param content - Raw mermaid fence body
  * @returns The same content with `quadrantChart` labels auto-quoted;

package/scripts/aggregator/markdown-renderer.js

@@ -180,6 +180,40 @@ function rewriteQuadrantChartLine(line) {
     }
     return line;
 }
+/**
+ * Decode the small set of HTML entities that Markdown authors (and
+ * upstream generators) occasionally pre-encode inside fenced mermaid
+ * blocks — typically & for & in political-group labels like
+ * S&D or Greens/EFA. Without this decode step, the subsequent
+ * escapeHtml pass would re-escape & to & and emit
+ * S&D into the rendered HTML, which the Mermaid client
+ * library then renders verbatim instead of as S&D.
+ *
+ * Uses indexOf/split/join exclusively (no RegExp) to stay
+ * within CodeQL's safe-regex envelope. Only the canonical entity
+ * forms are decoded — anything more exotic (e.g. &) is left
+ * alone so we never accidentally swallow a literal that the author
+ * intended to keep encoded.
+ *
+ * @param content - Raw fenced-block content (post-sanitizeMermaidQuadrantChart)
+ * @returns Content with pre-encoded HTML entities normalised back to
+ *          their literal characters, ready for a single escapeHtml.
+ */
+export function decodeMermaidPreEncodedEntities(content) {
+    // Order matters: decode the named entities first (which all contain
+    // `&` followed by ASCII letters), then finally `&` itself so we
+    // don't double-decode `&amp;lt;` -> `<`.
+    // Each replacement is a plain string `split(needle).join(replacement)`
+    // which is linear and trivially CodeQL-safe.
+    let out = content;
+    out = out.split('&lt;').join('<');
+    out = out.split('&gt;').join('>');
+    out = out.split('&quot;').join('"');
+    out = out.split('&#39;').join("'");
+    out = out.split('&apos;').join("'");
+    out = out.split('&amp;').join('&');
+    return out;
+}
 /**
  * Auto-quote unquoted `quadrantChart` labels so the Mermaid v11 lexer
  * accepts them. The Mermaid `quadrantChart` grammar treats unquoted
@@ -187,30 +221,9 @@ function rewriteQuadrantChartLine(line) {
  * en-dashes (`–`, U+2013), ellipsis (`…`), parentheses, colons, and
  * non-ASCII currency symbols (`€`) all trigger
  * `Lexical error … Unrecognized text` and prevent the diagram from
- * rendering, leaving the raw `<pre>` source visible on the page.
- *
- * The style guide already instructs authors to wrap every quadrant /
- * axis / data-point label in double quotes (see
- * `analysis/methodologies/political-style-guide.md` § Standard
- * `quadrantChart` init block), but AI-generated `article.md` files
- * occasionally drop the quoting. Rather than reject the article at
- * Stage C we sanitize at the renderer boundary so every published
- * HTML page renders, regardless of upstream authoring discipline.
- *
- * Sanitization is deliberately scoped to `quadrantChart` blocks —
- * `flowchart`, `sequenceDiagram`, `mindmap`, `pie`, `gantt`, and
- * `xychart-beta` accept the same Unicode characters in their unquoted
- * labels and are passed through unchanged.
- *
- * Lines normalised:
- *   - `x-axis Left --> Right`     → `x-axis "Left" --> "Right"`
- *   - `y-axis Low --> High`       → `y-axis "Low" --> "High"`
- *   - `quadrant-N Label text`     → `quadrant-N "Label text"`
- *   - `Data Label: [x, y]`        → `"Data Label": [x, y]`
- *
- * Already-quoted operands are preserved byte-for-byte. The `title`
- * line, the `%%{init:…}%%` directive, and any line not matching one
- * of the recognised shapes are also left untouched.
+ * rendering. Sanitization is scoped to `quadrantChart` blocks only;
+ * other diagram types accept those characters in unquoted labels and
+ * are passed through unchanged.
  *
  * @param content - Raw mermaid fence body
  * @returns The same content with `quadrantChart` labels auto-quoted;
@@ -267,7 +280,8 @@ function installMermaidFence(md) {
             const labelFn = env2.mermaidLabel ?? ((n) => `Mermaid diagram ${n + 1}`);
             const label = md.utils.escapeHtml(labelFn(currentIndex, token.content));
             const sanitized = sanitizeMermaidQuadrantChart(token.content);
-            const body = md.utils.escapeHtml(sanitized);
+            const decoded = decodeMermaidPreEncodedEntities(sanitized);
+            const body = md.utils.escapeHtml(decoded);
             return `<figure class="mermaid-figure" role="img" aria-label="${label}">\n<pre class="mermaid">${body}</pre>\n</figure>\n`;
         }
         return defaultFence(tokens, idx, opts, env, self);

package/scripts/aggregator/metadata/artifact-walker.js

@@ -17,7 +17,7 @@ import fs from 'fs';
 import path from 'path';
 import { extractFirstH1 } from './h1-extractor.js';
 import { extractLedeAfterHeading, extractStrongProseLine } from './lede-extractor.js';
-import { isGenericHeading, stripArtifactCategoryAffix } from './heading-rules.js';
+import { isGenericHeading, isArtifactCategoryHeading, stripArtifactCategoryAffix, } from './heading-rules.js';
 import { truncateTitle } from './text-utils.js';
 import { extractPriorityFindingHighlight } from './priority-finding-highlight.js';
 /** Ordered list of artefact filenames that typically carry the editorial H1. */
@@ -132,7 +132,7 @@ function probeCandidateForHighlight(runDir, rel, articleType, date) {
     // distinctive editorial headline ("Digital Markets Act Enforcement",
     // "Ukraine War Accountability") instead of a stripped category noun.
     const priority = extractPriorityFindingHighlight(body);
-    if (priority?.headline) {
+    if (priority?.headline && !isArtifactCategoryHeading(priority.headline)) {
         return {
             cleanHighlight: {
                 headline: truncateTitle(priority.headline),

package/scripts/aggregator/metadata/heading-rules.js

@@ -69,6 +69,7 @@ export const ARTIFACT_CATEGORY_PREFIXES = [
     'commission wp alignment',
     'committee activity report',
     'cross run continuity',
+    'data availability assessment',
     'deep analysis',
     'economic context',
     'executive brief',

package/scripts/aggregator/metadata/resolve-helpers.js

@@ -123,13 +123,19 @@ export function composeContextualTitle(fallbackTitle, editorialHeadline, runId)
  */
 export function composeContextualDescription(lang, baseDescription, editorial, date, _runId) {
     const labels = getLocalizedString(SEO_CONTEXT_LABELS, lang);
-    const parts = [baseDescription.trim()];
-    parts.push(`${labels.date} ${date}.`);
+    const base = baseDescription.trim();
+    const parts = [base];
+    const datePart = `${labels.date} ${date}.`;
+    if (!containsNormalized(base, `${labels.date} ${date}`)) {
+        parts.push(datePart);
+    }
     const context = pickFirstNonEmpty([editorial.summary, editorial.headline]);
     if (context && !containsNormalized(parts[0] ?? '', context)) {
         parts.push(`${labels.context}: ${context}`);
     }
-    parts.push(labels.reader);
+    if (!containsNormalized(parts.join(' '), labels.reader)) {
+        parts.push(labels.reader);
+    }
     return truncateDescription(parts.join(' '));
 }
 /**

package/scripts/aggregator/reader-guide/builder.js

@@ -3,7 +3,7 @@
 import { getLocalizedString, getTextDirection } from '../../constants/language-core.js';
 import { escapeHTML } from '../../utils/file-utils.js';
 import { READER_GUIDE_SECTION_ID } from '../reader-guide-constants.js';
-import { READER_GUIDE_TITLE_LABELS, READER_GUIDE_INTRO_LABELS, READER_GUIDE_COL_NEED_LABELS, READER_GUIDE_COL_VALUE_LABELS, } from './labels.js';
+import { READER_GUIDE_TITLE_LABELS, READER_GUIDE_INTRO_LABELS, READER_GUIDE_TIP_LABELS, READER_GUIDE_COL_NEED_LABELS, READER_GUIDE_COL_VALUE_LABELS, } from './labels.js';
 import { READER_GUIDE_ROWS } from './rows.js';
 import { getReaderGuideSectionIcon } from './icons.js';
 /**
@@ -40,11 +40,13 @@ export function buildReaderIntelligenceGuideHtml(lang, sections, _included = [])
         return '';
     const title = getLocalizedString(READER_GUIDE_TITLE_LABELS, lang);
     const intro = getLocalizedString(READER_GUIDE_INTRO_LABELS, lang);
+    const tip = getLocalizedString(READER_GUIDE_TIP_LABELS, lang);
     const colNeed = getLocalizedString(READER_GUIDE_COL_NEED_LABELS, lang);
     const colValue = getLocalizedString(READER_GUIDE_COL_VALUE_LABELS, lang);
     return `<section id="${READER_GUIDE_SECTION_ID}" data-component="reader-intelligence-guide" aria-label="${escapeHTML(title)}"${dir === 'rtl' ? ' dir="rtl"' : ''}>
 <h2 id="${READER_GUIDE_SECTION_ID}-heading"><span class="guide-icon" aria-hidden="true">🧭</span> ${escapeHTML(title)}</h2>
 <p class="reader-guide-intro">${escapeHTML(intro)}</p>
+<p class="reader-guide-tip"><span class="guide-icon" aria-hidden="true">💡</span> ${escapeHTML(tip)}</p>
 <div class="table-scroll" role="region" tabindex="0" aria-labelledby="${READER_GUIDE_SECTION_ID}-heading">
 <table class="reader-guide-table">
 <caption class="sr-only">${escapeHTML(title)}</caption>

package/scripts/aggregator/reader-guide/labels.d.ts

@@ -10,6 +10,13 @@ import type { LanguageMap } from '../../types/index.js';
 export declare const READER_GUIDE_TITLE_LABELS: LanguageMap;
 /** Introduction text for the Reader Intelligence Guide */
 export declare const READER_GUIDE_INTRO_LABELS: LanguageMap;
+/**
+ * Practical "how to read this article" tip rendered immediately under the
+ * intro. Distinct from the intro so existing snapshot tests continue to
+ * match the intro string verbatim, and so styles can target the two
+ * paragraphs independently.
+ */
+export declare const READER_GUIDE_TIP_LABELS: LanguageMap;
 /** Table header: "Reader need" */
 export declare const READER_GUIDE_COL_NEED_LABELS: LanguageMap;
 /** Table header: "What you'll get" */