protovibe 1.1.9 → 1.1.10

package/README.md

@@ -38,19 +38,19 @@ ProtoVibe is a CLI tool and **must be installed globally** to work:
 npm install -g protovibe
 ```
 
-> **Note:** The npm page shows `npm i protovibe` but that is a local install — the `protovibe` command will not be found. Always use the `-g` flag.
+> **Note:** The npm page shows `npm i protovibe` — that is a local install and the `protovibe` command will not be found. Always use the `-g` flag.
 
 ProtoVibe checks for newer versions on every run and updates itself automatically. You only ever need to install it once.
 
 ### No install? Use npx
 
-If you don't want a global install, you can run ProtoVibe directly with npx from any directory:
+Run ProtoVibe directly from any directory without a permanent install:
 
 ```bash
 npx protovibe
 ```
 
-This always runs the latest version without installing anything permanently.
+This always fetches and runs the latest version on the fly.
 
 ---
 
@@ -145,28 +145,43 @@ Choose **1** to begin:
 Claude asks: *"Describe the idea for your project."*
 Speak naturally. No format required.
 
-### Stage 2 — Deep Requirements Probing
-Claude asks grouped questions covering:
-- Who are the target users?
-- What does each feature do exactly?
-- Step-by-step user flow
-- Platform (web, mobile, desktop, CLI)?
-- Authentication — required? What type?
-- Data — what's created, stored, deleted? Real-time?
-- Third-party integrations?
-- Constraints — performance, offline, accessibility, timeline?
+### Stage 2 — Interactive Requirements Probing
 
-Claude keeps probing until everything is clear. Then it presents a full structured summary and asks you to confirm before moving on.
+Claude first silently analyses your idea, then asks requirements questions as **interactive cards** — you click an option or type your own. No long-form typing required.
+
+Questions are asked in 3 batches:
+
+**Batch 1 — Core context**
+- Who are the primary users?
+- What platform is this for? *(Web, Mobile, Desktop, CLI)*
+- Is user authentication required?
+- How would you describe the scale at launch?
+
+**Batch 2 — Features & data**
+- What data does this app create or store? *(multi-select)*
+- Any third-party integrations needed? *(multi-select)*
+- Does the app need real-time features?
+
+**Batch 3 — Constraints**
+- Do you have a preferred tech stack?
+- What is your timeline for an MVP?
+- Any specific requirements? *(multi-select)*
+
+Options are generated from your idea — not generic defaults. Each question has a recommended answer and an "Other" option to type anything.
+
+If anything is still unclear after your answers, Claude asks targeted follow-up questions the same way. Then it presents a full structured summary and asks you to confirm before moving on.
 
 ### Stage 3 — Architecture Options
+
 Claude proposes **3 distinct architectural approaches**, each with:
 - Full tech stack (frontend, backend, database, auth, hosting)
 - How the system works end-to-end
 - Tradeoffs — what it's best for and what it sacrifices
-- One option marked **Recommended** with a specific reason
+
+Then presents them as **clickable cards** — pick one, or choose "Custom approach" to describe your own. The recommended option is labeled clearly.
 
 ### Stage 4 — Option Confirmation
-Pick one of the 3 options, or describe your own. If you go custom, Claude evaluates it against your requirements and flags any risks before proceeding.
+If you pick a custom approach, Claude evaluates it against your requirements and flags any risks before proceeding.
 
 ### Stage 5 — Specs File
 Claude asks for confirmation, then writes `specs.md` in your project root containing:
@@ -194,7 +209,7 @@ These are available at any time inside Claude Code during or after the workflow:
 
 | Command | What it does |
 |---|---|
-| `/takeover` | Exit the ProtoVibe workflow. Claude summarises everything captured so far (idea, requirements, architecture, files created) and then works normally with no constraints. |
+| `/takeover` | Exit the ProtoVibe workflow. Claude summarises everything captured so far and works normally with no constraints. |
 | `/summarise` | Plain-language summary of progress — what's been decided, what's been done, and what comes next. Capped at 10 bullets. |
 | `/protovibe` | Restart the workflow from Stage 0. |
 

package/dist/index.mjs

@@ -223,21 +223,29 @@ Your goal is to achieve complete, end-to-end clarity on this product before movi
 
 **How to probe:**
 
-Ask your questions in **grouped batches** \u2014 not one at a time. Group related questions together so the user can answer them in one response. After receiving answers, analyse them carefully. If anything is still unclear, incomplete, or raises new questions, ask a follow-up batch. Repeat until you have full clarity.
+First, silently analyse the user's idea thoroughly. Identify the most likely platform, user type, auth needs, data requirements, and integrations based on what they described. Use this analysis to generate smart, contextual options for each question \u2014 options should reflect what makes sense for their specific idea, not generic defaults.
 
-You must cover all of the following \u2014 do not move on until every point is resolved:
+Then ask your questions using the **AskUserQuestion tool** in 3 batches. Do not ask as plain text lists \u2014 always use the tool so the user can click options or type their own answer.
 
-- **Who** are the target users? (be specific \u2014 demographics, roles, technical level)
-- **What** does each core feature actually do? (exact behavior, not just feature names)
-- **User Flow** \u2014 how does a user move through the product from first open to completing each core action? Walk through every key path.
-- **Platform** \u2014 web, mobile, desktop, CLI, or multiple? Any specific device or OS constraints?
-- **Auth** \u2014 is login required? What type? (email, social, SSO?) Any roles or permission levels?
-- **Data** \u2014 what data is created, stored, read, or deleted? Any real-time or sync requirements?
-- **Integrations** \u2014 any third-party services, APIs, payment systems, or external tools?
-- **Constraints** \u2014 performance targets, offline support, accessibility needs, team size, timeline, existing tech choices the user wants to keep?
-- **Edge cases** \u2014 what happens when things go wrong? Any critical failure states to handle?
+**Batch 1 \u2014 Core context** (call AskUserQuestion with 4 questions):
+1. "Who are the primary users?" \u2014 generate 3 specific, idea-relevant options (e.g. "Freelance designers", "Small business owners", "Developer teams") \u2014 user can type their own via Other
+2. "What platform is this for?" \u2014 options: Web app, Mobile app (iOS/Android), Desktop app, CLI tool \u2014 mark the most likely one as recommended based on the idea
+3. "Is user authentication required?" \u2014 options: Yes \u2013 email & password, Yes \u2013 social login (Google/GitHub/etc.), No login needed, Multiple auth methods
+4. "How would you describe the scale at launch?" \u2014 options: Personal / side project, Small team (under 20 users), Public product (hundreds of users), Enterprise / large scale
 
-**There is no question limit.** If the user's answers raise new questions, ask them. Keep going until you have the full picture.
+**Batch 2 \u2014 Features & data** (call AskUserQuestion with 3 questions):
+1. "What data does this app create or store?" \u2014 generate 3\u20134 idea-relevant options (e.g. "User profiles & preferences", "Files or media uploads", "Transactions or records", "Posts or content") \u2014 use multiSelect: true
+2. "Any third-party integrations needed?" \u2014 generate idea-relevant options (e.g. "Stripe \u2013 payments", "SendGrid \u2013 email", "OpenAI \u2013 AI features", "None for now") \u2014 use multiSelect: true
+3. "Does the app need real-time features?" \u2014 options: Yes \u2013 live updates or notifications, Yes \u2013 real-time collaboration, No \u2013 standard request/response is fine, Not sure yet
+
+**Batch 3 \u2014 Constraints** (call AskUserQuestion with 3 questions):
+1. "Do you have a preferred tech stack?" \u2014 generate 3 idea-appropriate stack options (e.g. "Next.js full-stack", "React + Node.js", "Python + FastAPI") + "No preference \u2013 recommend something" marked as recommended
+2. "What is your timeline for an MVP?" \u2014 options: ASAP / this week, 2\u20134 weeks, 1\u20133 months, No hard deadline
+3. "Any specific requirements?" \u2014 options: Offline support needed, Accessibility (WCAG) required, Mobile-first design, None of these \u2014 use multiSelect: true
+
+After collecting all answers, analyse them carefully. If any critical area is still unclear (feature behaviour, user flow details, edge cases), call AskUserQuestion again with targeted follow-up questions. Keep going until you have the full picture \u2014 there is no question limit.
+
+**There is no question limit.** If the user's answers raise new questions, use AskUserQuestion again. Keep going until you have full clarity.
 
 When you are confident everything is fully clear, present a complete structured summary:
 
@@ -274,9 +282,13 @@ Analyze the confirmed requirements carefully. Propose exactly **3 distinct archi
 
 After presenting all 3, clearly mark one as **Recommended** and explain specifically why it fits this user's requirements better than the others.
 
-Then ask:
+Then call **AskUserQuestion** with a single question:
 
-> "Which option would you like to go with? You can pick one of the above, or describe your own approach."
+- Question: "Which architecture would you like to go with?"
+- Option 1: label = name of Option 1, description = one-line stack summary
+- Option 2: label = name of Option 2, description = one-line stack summary
+- Option 3: label = name of Option 3 + " (Recommended)", description = one-line stack summary + the specific reason it's recommended
+- Option 4: label = "Custom approach", description = "Describe your own stack \u2014 I'll evaluate it against your requirements"
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "protovibe",
-  "version": "1.1.9",
+  "version": "1.1.10",
   "description": "A branded CLI platform layer on top of Claude Code",
   "author": "razorgojo",
   "license": "MIT",