fix(sync): normalize section separators in regenerate()

A previous run could merge the '---' separator with the next section
header (---## ✅ Befejezett), breaking section detection. Prefix/suffix
are now trimmed and rejoined with fixed separators. Regression test
added; also create Plane issues for local-only completed work
(MITHOME-28 Custom CMS, MITHOME-29 Webmail) to avoid ID collisions.

scripts/plane/plane-sync.test.js

@@ -162,3 +162,25 @@ test('buildTable sorts rows by sequence number', () => {
   assert.equal(lines[2], '| MITHOME-1 | A | ✅ |');
   assert.equal(lines[3], '| MITHOME-15 | B | ✅ |');
 });
+
+// Regression: a previous run dropping the blank line after `---` must not merge
+// the separator with the section header (`---## ✅ Befejezett`).
+test('regenerate always separates `---` from section headers', () => {
+  const planeIssues = [
+    { sequence_id: 1, name: 'Kezdőlap', state: { group: 'completed' } },
+    { sequence_id: 13, name: 'CI pipeline', state: { group: 'unstarted' } },
+    { sequence_id: 17, name: 'Winston logger + Loki integráció', state: { group: 'started' } },
+    { sequence_id: 19, name: 'Accessibility', state: { group: 'backlog' } }
+  ];
+  const { rows } = parseTodoFile(SAMPLE_TODO);
+  const buildData = buildDiscrepancies(planeIssues, rows, 'MITHOME');
+  const decisions = { moveToPlane: new Set(), keepOrphan: new Set(), addNew: new Set() };
+  const final = buildFinalSections(buildData, decisions);
+
+  const corrupted = SAMPLE_TODO.replace('---\n\n## ✅ Befejezett', '---## ✅ Befejezett');
+  const output = regenerate(corrupted, final);
+
+  assert.ok(output.includes('---\n\n## ✅ Befejezett'), 'header must be separated from ---');
+  assert.ok(!output.includes('---##'), 'no merged separator+header');
+  assert.ok(output.includes('## ✅ Befejezett'));
+});

scripts/plane/todo.js

@@ -64,7 +64,9 @@ function buildTable(sec, rows) {
   return [header, separator, ...body].join('\n');
 }
 
-// WHY: only the section region is regenerated, prefix/suffix (overview, commands) are preserved
+// WHY: only the section region is regenerated, prefix/suffix (overview, commands) are preserved.
+// Prefix/suffix are trimmed and rejoined with normalized separators so that a previous run
+// that dropped the blank line after `---` cannot corrupt the section headers (`---## ✅`).
 function regenerate(content, finalSections) {
   const lines = content.split('\n');
   const firstIdx = lines.findIndex(l => SECTIONS[0].header.test(l));
@@ -79,8 +81,8 @@ function regenerate(content, finalSections) {
     }
   }
 
-  const prefix = lines.slice(0, firstIdx).join('\n');
-  const suffix = lines.slice(endIdx + 1).join('\n');
+  const prefix = lines.slice(0, firstIdx).join('\n').trimEnd();
+  const suffix = lines.slice(endIdx + 1).join('\n').trimStart();
 
   const headerOf = key => {
     const sec = SECTIONS.find(s => s.key === key);
@@ -92,9 +94,8 @@ function regenerate(content, finalSections) {
     const rows = finalSections[sec.key] || [];
     return `## ${headerOf(sec.key)}\n\n${buildTable(sec, rows)}`;
   });
-  const middle = blocks.join('\n\n---\n\n') + '\n\n---\n';
-
-  const result = `${prefix}${middle}${suffix}`;
+  const middle = blocks.join('\n\n---\n\n');
+  const result = `${prefix}\n\n${middle}\n\n---\n\n${suffix}`;
   return result.endsWith('\n') ? result : `${result}\n`;
 }