feat(news-harvester): single-article refetch by URL

- refetch.js: fetch one article via its source recipe on the site-shared
  browser instance; refresh 原文.md in place for already-archived articles
  (registry ownership-checked, index rebuilt from registry) and re-push
  via push.js --url; new URLs archive to today's dir pending summaries.
  Busy site locks are respected; only self-launched instances are recycled.
- push.js: --url <articleUrl> re-pushes a single article (query/hash
  normalized match, ledger ignored for it); recordResult now appends a
  ledger entry when the article postdates the ledger instead of silently
  dropping a successful push result (refetch-into-finalized-day case).
- harvest.js: export downloadImage/normalizeUrl for refetch.

skills/news-harvester/scripts/harvest.js

@@ -791,6 +791,19 @@ Examples:
         console.error(`  ✗ Fetch failed: ${e.message}`);
         continue;
       }
+      if (articleData && articleData.isArticle === false) {
+        // Lazy-rendered pages often miss og:type/body on first paint — wait 5s
+        // and reload once before judging it a non-article page.
+        console.error('  ⏳ 首载未过文章校验 — 等 5 秒重载一次再判');
+        await sleep(5000);
+        try {
+          const reloaded = await getPage(url);
+          if (reloaded) articleData = reloaded;
+        } catch (e) {
+          if (e instanceof HumanInterventionError) throw e;
+          console.error(`  ⚠️ 重载失败（保留首载结果再判）: ${e.message}`);
+        }
+      }
       const articleInfo = await processArticle(url, articleData);
       if (articleInfo) {
         processed.push(articleInfo);
@@ -881,5 +894,7 @@ module.exports = {
   slugFromUrl,
   sentinelFor,
   originalFilename,
-  dateShort
+  dateShort,
+  downloadImage,
+  normalizeUrl
 };

skills/news-harvester/scripts/push.js

@@ -69,6 +69,13 @@ function loadJson(p, label) {
 
 function sleep(ms) { return new Promise(r => setTimeout(r, ms)); }
 
+// Canonical URL key for matching a registry article regardless of tracking
+// params: strip query string and hash (same normalization harvest.js applies).
+function normKey(u) {
+  try { const p = new URL(u); p.search = ''; p.hash = ''; return p.toString(); }
+  catch (e) { return u; }
+}
+
 function uuid() {
   return crypto.randomUUID();
 }
@@ -418,8 +425,15 @@ function newSyncState(registry) {
 
 // Mark a url's result in the ledger.
 function recordResult(state, url, result) {
-  const entry = state.articles.find(a => a.url === url);
-  if (!entry) return;
+  let entry = state.articles.find(a => a.url === url);
+  if (!entry) {
+    // Articles appended to the registry AFTER the ledger was created (e.g.
+    // refetch.js adding one to an already-synced day dir) have no entry yet.
+    // Create it here instead of dropping the result — otherwise a successful
+    // push is never recorded and the article re-pushes on every later run.
+    entry = { url, slug: null, status: 'pending', articleId: null, version: null, error: null };
+    state.articles.push(entry);
+  }
   if (result.status === 'error') {
     entry.status = 'error';
     entry.error = result.error || null;
@@ -478,10 +492,26 @@ async function pushDayDir(dayDir, cfg, opts) {
     state = newSyncState(registry);
     console.error('   --force: ledger reset, re-pushing all articles.');
   }
-  const pending = articles.filter(a => {
+  let pending;
+  if (opts.url) {
+    // --url: single-article re-push (refetch flow). Ignores the ledger status
+    // of that one article (it is being re-pushed deliberately); the rest of
+    // the day's ledger stays untouched. Comparison strips query/hash — the
+    // same normalization harvest.js applies to registry keys.
+    const key = normKey(opts.url);
+    pending = articles.filter(a => normKey(a.url) === key);
+    if (pending.length === 0) {
+      console.error(`✗ --url: no registry article matches ${opts.url}`);
+      console.error('  (registry URLs are compared with query/hash stripped)');
+      process.exit(1);
+    }
+    console.error(`   --url: re-pushing ${pending.length} matching article (ledger ignored for it)`);
+  } else {
+    pending = articles.filter(a => {
       const e = state.articles.find(x => x.url === a.url);
       return !e || e.status !== 'synced';
     });
+  }
   if (pending.length === 0) {
     console.error('   ✅ All articles already synced — nothing to do.');
     console.error('   (use --force to re-push)');
@@ -697,6 +727,10 @@ function printHelp() {
 Usage:
   node scripts/push.js <dayDir>              Push a day directory (incremental)
   node scripts/push.js <dayDir> --force      Ignore ledger, re-push all articles
+  node scripts/push.js <dayDir> --url <articleUrl>
+                          Re-push a single article (matched with query/hash
+                          stripped); its ledger status is ignored, the rest of
+                          the day's ledger is untouched (refetch flow)
   node scripts/push.js <dayDir> --bump-processed-at
                           Override processedAt to now() so the backend's LWW
                           accepts the update (use when re-pushing corrected
@@ -740,14 +774,23 @@ function main() {
     process.exit(0);
   }
 
-  const positional = args.filter(a => !a.startsWith('-'));
-  const flags = args.filter(a => a.startsWith('-'));
+  // --url <url>: its value is a positional-looking string (starts with http),
+  // so pull the pair out BEFORE splitting positional/flags.
+  const urlIdx = args.indexOf('--url');
+  const urlFilter = urlIdx >= 0 ? (args[urlIdx + 1] || null) : null;
+  const skipIdx = urlIdx >= 0 ? urlIdx + 1 : -1;
+  const positional = args.filter((a, i) => !a.startsWith('-') && i !== skipIdx);
+  const flags = args.filter((a, i) => a.startsWith('-') && i !== skipIdx);
 
   if (positional.length === 0) {
-    console.error('Usage: node scripts/push.js <dayDir> [--force] [--bump-processed-at] [--status]');
+    console.error('Usage: node scripts/push.js <dayDir> [--force] [--bump-processed-at] [--status] [--url <articleUrl>]');
     console.error('Run "node scripts/push.js --help" for details.');
     process.exit(1);
   }
+  if (urlIdx >= 0 && !urlFilter) {
+    console.error('✗ --url requires a value: --url https://...');
+    process.exit(1);
+  }
 
   const dayDir = positional[0];
   if (!fs.existsSync(dayDir) || !fs.statSync(dayDir).isDirectory()) {
@@ -776,6 +819,7 @@ function main() {
   pushDayDir(dayDir, cfg, {
     force: flags.includes('--force'),
     bumpProcessedAt: flags.includes('--bump-processed-at'),
+    url: urlFilter,
   }).catch(e => {
     console.error(`Fatal: ${e.message}`);
     process.exit(1);