fix(discuss): add multi-round questioning to new-project discuss phase

The discuss.md prompt (used for /gsd new project creation) only asked
"What's the vision?" once and relied on LLM judgment for follow-ups.
This led to the agent asking a single question then jumping straight
to planning without gathering enough context.

Add explicit Question Rounds section with:
- 1-3 questions per round structure
- Conditional ask_user_questions vs plain text support
- Incremental persistence (CONTEXT-DRAFT save every 2 rounds)
- Depth-matching rule (1-2 rounds for simple, 4+ for large visions)
- Round cadence that drives toward the depth enforcement checklist

Thread structuredQuestionsAvailable through buildDiscussPrompt() and
prepareAndBuildDiscussPrompt() so the template variable resolves
correctly at runtime.

Closes #3976

src/resources/extensions/gsd/guided-flow.ts

@@ -426,8 +426,9 @@ function resolveAvailableModel<T extends { id: string; provider: string }>(
  * Build the discuss-and-plan prompt for a new milestone.
  * Used by all three "new milestone" paths (first ever, no active, all complete).
  */
-function buildDiscussPrompt(nextId: string, preamble: string, _basePath: string, preparationContext?: string): string {
+function buildDiscussPrompt(nextId: string, preamble: string, _basePath: string, pi: ExtensionAPI, ctx: ExtensionCommandContext, preparationContext?: string): string {
   const milestoneRel = `.gsd/milestones/${nextId}`;
+  const structuredQuestionsAvailable = getStructuredQuestionsAvailability(pi, ctx);
   const inlinedTemplates = [
     inlineTemplate("project", "Project"),
     inlineTemplate("requirements", "Requirements"),
@@ -439,6 +440,7 @@ function buildDiscussPrompt(nextId: string, preamble: string, _basePath: string,
     milestoneId: nextId,
     preamble,
     preparationContext: preparationContext ?? "",
+    structuredQuestionsAvailable,
     contextPath: `${milestoneRel}/${nextId}-CONTEXT.md`,
     roadmapPath: `${milestoneRel}/${nextId}-ROADMAP.md`,
     inlinedTemplates,
@@ -486,6 +488,7 @@ function buildHeadlessDiscussPrompt(nextId: string, seedContext: string, _basePa
  */
 async function prepareAndBuildDiscussPrompt(
   ctx: ExtensionCommandContext,
+  pi: ExtensionAPI,
   nextId: string,
   preamble: string,
   basePath: string,
@@ -520,7 +523,7 @@ async function prepareAndBuildDiscussPrompt(
     }
   }
 
-  return buildDiscussPrompt(nextId, preamble, basePath, preparationContext);
+  return buildDiscussPrompt(nextId, preamble, basePath, pi, ctx, preparationContext);
 }
 
 /**
@@ -780,7 +783,7 @@ export async function showDiscuss(
       const uniqueMilestoneIds = !!loadEffectiveGSDPreferences()?.preferences?.unique_milestone_ids;
       const nextId = nextMilestoneIdReserved(milestoneIds, uniqueMilestoneIds);
       pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: false, createdAt: Date.now() });
-      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId, `New milestone ${nextId}.`, basePath), "gsd-run", ctx, "discuss-milestone");
+      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId, `New milestone ${nextId}.`, basePath), "gsd-run", ctx, "discuss-milestone");
     }
     return;
   }
@@ -1185,7 +1188,7 @@ async function handleMilestoneActions(
     const uniqueMilestoneIds = !!loadEffectiveGSDPreferences()?.preferences?.unique_milestone_ids;
     const nextId = nextMilestoneIdReserved(milestoneIds, uniqueMilestoneIds);
     pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: stepMode, createdAt: Date.now() });
-    await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId,
+    await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId,
       `New milestone ${nextId}.`,
       basePath
     ), "gsd-run", ctx, "discuss-milestone");
@@ -1375,7 +1378,7 @@ export async function showSmartEntry(
     if (isFirst) {
       // First ever — skip wizard, just ask directly
       pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: stepMode, createdAt: Date.now() });
-      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId,
+      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId,
         `New project, milestone ${nextId}. Do NOT read or explore .gsd/ — it's empty scaffolding.`,
         basePath
       ), "gsd-run", ctx, "discuss-milestone");
@@ -1396,7 +1399,7 @@ export async function showSmartEntry(
 
       if (choice === "new_milestone") {
         pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: stepMode, createdAt: Date.now() });
-        await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId,
+        await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId,
           `New milestone ${nextId}.`,
           basePath
         ), "gsd-run", ctx, "discuss-milestone");
@@ -1435,7 +1438,7 @@ export async function showSmartEntry(
       const nextId = nextMilestoneIdReserved(milestoneIds, uniqueMilestoneIds);
 
       pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: stepMode, createdAt: Date.now() });
-      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId,
+      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId,
         `New milestone ${nextId}.`,
         basePath
       ), "gsd-run", ctx, "discuss-milestone");
@@ -1502,7 +1505,7 @@ export async function showSmartEntry(
       const uniqueMilestoneIds = !!loadEffectiveGSDPreferences()?.preferences?.unique_milestone_ids;
       const nextId = nextMilestoneIdReserved(milestoneIds, uniqueMilestoneIds);
       pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: stepMode, createdAt: Date.now() });
-      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId,
+      await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId,
         `New milestone ${nextId}.`,
         basePath
       ), "gsd-run", ctx, "discuss-milestone");
@@ -1599,7 +1602,7 @@ export async function showSmartEntry(
         const uniqueMilestoneIds = !!loadEffectiveGSDPreferences()?.preferences?.unique_milestone_ids;
         const nextId = nextMilestoneIdReserved(milestoneIds, uniqueMilestoneIds);
         pendingAutoStartMap.set(basePath, { ctx, pi, basePath, milestoneId: nextId, step: stepMode, createdAt: Date.now() });
-        await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, nextId,
+        await dispatchWorkflow(pi, await prepareAndBuildDiscussPrompt(ctx, pi, nextId,
           `New milestone ${nextId}.`,
           basePath
         ), "gsd-run", ctx, "discuss-milestone");

src/resources/extensions/gsd/prompts/discuss.md

@@ -49,6 +49,26 @@ This happens ONCE, before the first round. The goal: your first questions should
 
 For subsequent rounds, continue investigating between rounds — check docs, search, or scout as needed to make each round's questions smarter. But the first-round investigation is mandatory and explicit. Distribute searches across turns rather than clustering them in one turn.
 
+## Question Rounds
+
+Ask **1–3 questions per round**. Keep each round tightly focused on one or two of the depth checklist dimensions — do not try to cover all six in one round.
+
+**If `{{structuredQuestionsAvailable}}` is `true`:** use `ask_user_questions` for each round. 1–3 questions per call, each as a separate question object. Keep option labels short (3–5 words). Always include a freeform "Other / let me explain" option. When the user picks that option or writes a long freeform answer, switch to plain text follow-up for that thread before resuming structured questions. **IMPORTANT: Call `ask_user_questions` exactly once per turn. Never make multiple calls with the same or overlapping questions — wait for the user's response before asking the next round.**
+
+**If `{{structuredQuestionsAvailable}}` is `false`:** ask questions in plain text. Keep each round to 1–3 focused questions. Wait for answers before asking the next round.
+
+After each answer set, investigate further if any answer opens a new unknown, then ask the next round.
+
+### Round cadence
+
+After each round of answers, decide whether you already have enough depth to write strong output.
+
+- **Incremental persistence:** After every 2 question rounds, silently save a `{{milestoneId}}-CONTEXT-DRAFT.md` using `gsd_summary_save` with `artifact_type: "CONTEXT-DRAFT"` and `milestone_id: "{{milestoneId}}"`. This protects confirmed work against session crashes. Do NOT mention this save to the user.
+- If not ready, continue to the next round immediately. Do **not** ask a meta "ready to wrap up?" question after every round.
+- **Depth-matching rule:** Simple, well-defined work needs fewer rounds — maybe 1–2. Large, ambiguous visions need more — maybe 4+. Do not pad rounds to hit a number. Stop when the Depth Enforcement checklist below is fully satisfied.
+- Do not count the reflection step as a question round. Rounds start after reflection is confirmed.
+- When you genuinely believe the depth checklist is satisfied, move to the Depth Verification step below. Do not ask a separate "ready to wrap up?" gate — the depth verification IS the gate.
+
 ## Questioning Philosophy
 
 You are a thinking partner, not an interviewer.
@@ -94,10 +114,6 @@ Do NOT offer to proceed until ALL of the following are satisfied. Track these in
 
 Before offering to proceed, demonstrate absorption: reference specific things the user emphasized, specific terminology they used, specific nuance they sharpened — and show how those shaped your understanding. Synthesize, don't recite. "Your emphasis on X led me to prioritize Y over Z" is good. "You said X, you said Y, you said Z" is not. The user should feel heard in the specifics, not just acknowledged in the abstract.
 
-**Questioning depth should match scope.** Simple, well-defined work needs fewer rounds — maybe 1-2. Large, ambiguous visions need more — maybe 4+. Don't pad rounds to hit a number. Stop when the depth checklist is satisfied and you genuinely understand the work.
-
-Do not count the reflection step as a question round. Rounds start after reflection is confirmed.
-
 ## Depth Verification
 
 Before moving to the wrap-up gate, present a structured depth summary as a checkpoint.

src/resources/extensions/gsd/tests/discuss-incremental-persistence.test.ts

@@ -27,10 +27,19 @@ describe("discuss incremental persistence (#2152)", () => {
     assert.match(content, /Incremental persistence/, "should have incremental persistence section");
   });
 
+  test("new-project discuss prompt includes CONTEXT-DRAFT save instruction", () => {
+    const content = readFileSync(join(promptsDir, "discuss.md"), "utf-8");
+    assert.match(content, /CONTEXT-DRAFT/, "should mention CONTEXT-DRAFT");
+    assert.match(content, /Incremental persistence/, "should have incremental persistence section");
+    assert.match(content, /gsd_summary_save/, "should use gsd_summary_save tool");
+  });
+
   test("drafts are saved silently without user notification", () => {
     const milestone = readFileSync(join(promptsDir, "guided-discuss-milestone.md"), "utf-8");
     const slice = readFileSync(join(promptsDir, "guided-discuss-slice.md"), "utf-8");
+    const discuss = readFileSync(join(promptsDir, "discuss.md"), "utf-8");
     assert.match(milestone, /Do NOT mention this save to the user/);
     assert.match(slice, /Do NOT mention this to the user/);
+    assert.match(discuss, /Do NOT mention this save to the user/);
   });
 });