fix: add fallback parser for prose-style roadmaps without ## Slices section (#807) (#835)

When the LLM writes freeform prose roadmaps with `## Slice S01: Title`
headers instead of the machine-readable `## Slices` checklist,
parseRoadmapSlices() returned zero slices, causing deriveState() to
permanently block with 'No slice eligible'.

Add a fallback parser that detects prose-style `## Slice SXX:` headers
(and variants like `## S01:`, `## S01 —`) and extracts slice IDs,
titles, and dependencies from the prose. Also parses `Depends on:`
text patterns. All fallback slices default to risk:medium and done:false.

src/resources/extensions/gsd/roadmap-slices.ts

@@ -53,7 +53,12 @@ function extractSlicesSection(content: string): string {
 export function parseRoadmapSlices(content: string): RoadmapSliceEntry[] {
   const slicesSection = extractSlicesSection(content);
   const slices: RoadmapSliceEntry[] = [];
-  if (!slicesSection) return slices;
+  if (!slicesSection) {
+    // Fallback: detect prose-style slice headers (## Slice S01: Title)
+    // when the LLM writes freeform prose instead of the ## Slices checklist.
+    // This prevents a permanent "No slice eligible" block (#807).
+    return parseProseSliceHeaders(content);
+  }
 
   const checkboxItems = slicesSection.split("\n");
   let currentSlice: RoadmapSliceEntry | null = null;
@@ -88,3 +93,42 @@ export function parseRoadmapSlices(content: string): RoadmapSliceEntry[] {
   if (currentSlice) slices.push(currentSlice);
   return slices;
 }
+
+/**
+ * Fallback parser for prose-style roadmaps where the LLM wrote
+ * `## Slice S01: Title` headers instead of the machine-readable
+ * `## Slices` checklist. Extracts slice IDs and titles so auto-mode
+ * can at least identify slices and plan them.
+ *
+ * Also handles `## S01: Title` and `## S01 — Title` variants.
+ */
+function parseProseSliceHeaders(content: string): RoadmapSliceEntry[] {
+  const slices: RoadmapSliceEntry[] = [];
+  const headerPattern = /^##\s+(?:Slice\s+)?(S\d+)[:\s—–-]+\s*(.+)/gm;
+  let match: RegExpExecArray | null;
+
+  while ((match = headerPattern.exec(content)) !== null) {
+    const id = match[1]!;
+    const title = match[2]!.trim();
+
+    // Try to extract depends from prose: "Depends on: S01" or "**Depends on:** S01, S02"
+    const afterHeader = content.slice(match.index + match[0].length);
+    const nextHeader = afterHeader.search(/^##\s/m);
+    const section = nextHeader !== -1 ? afterHeader.slice(0, nextHeader) : afterHeader.slice(0, 500);
+
+    const depsMatch = section.match(/\*{0,2}Depends\s+on:?\*{0,2}\s*(.+)/i);
+    let depends: string[] = [];
+    if (depsMatch) {
+      const rawDeps = depsMatch[1]!.replace(/none/i, "").trim();
+      if (rawDeps) {
+        depends = expandDependencies(
+          rawDeps.split(/[,;]/).map(s => s.trim().replace(/[^A-Za-z0-9]/g, "")).filter(Boolean)
+        );
+      }
+    }
+
+    slices.push({ id, title, risk: "medium" as RiskLevel, depends, done: false, demo: "" });
+  }
+
+  return slices;
+}