nathan13888/nice-skills/mvp

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Product design, feature planning, and technical architecture for new projects. Explores the problem space through deep requirements gathering, suggests creative features, makes architecture decisions, and produces a structured MVP plan with scope boundaries, a future roadmap, and a deliverable tracker. Uses plan mode for deliberate thinking before writing any artifacts. Use when the user says "mvp", "plan a product", "design features", "what should I build", "feature planning", "scope an MVP", or describes a product they want to plan.

Install

npx vskill@latest install nathan13888/nice-skills/mvp

SKILL.md

MVP Product Planner

Deep requirements gathering, feature ideation, scope management, technical architecture, and living documentation for new products. Produces 5 planning artifacts: OVERVIEW, FEATURES, MVP_SCOPE, ROADMAP, and DELIVERABLES.

Quick Start

Follow all steps sequentially across 3 phases. Do NOT write any files until the plan is approved in Step 6.

Workflow

Phase 1: Understand (Steps 0-3)

Step 0: Enter Plan Mode & Parse Invocation
  1. Call EnterPlanMode immediately -- no files are written until the plan is approved in Step 6.
  2. Parse the invocation:
  • /mvp (bare) -- start the full workflow from Step 1
  • /mvp a fitness tracker -- seed description provided, use it as the answer to Step 1's seed question and skip that question
  • /mvp --resume -- Glob for existing docs in ./docs/mvp/, ./plans/, ./mvp/. If found, read them and resume from the appropriate step based on what's already documented. If not found, start fresh from Step 1.
Step 1: Problem Space Exploration

Deep requirements gathering via AskUserQuestion with concrete options. Ask ALL of these questions, adapting options based on prior answers:

  1. Seed question (skip if a description was provided in the invocation):

> What are you building? Describe the problem or product in a few sentences.

  1. Target users (select one):
  • Yourself / Developers / Internal team / End consumers (B2C) / Businesses (B2B)
  1. Problem severity (select one):
  • Hair on fire / Annoying / Nice to have / New behavior
  1. Existing solutions (multiSelect):
  • Nothing / Manual processes / Competitors / Internal tools / Open source
  1. Scale expectations (select one):
  • Just me / Small group (<100) / Medium (hundreds-thousands) / Large (10k+) / Unknown

After gathering all answers, produce a Problem Brief and confirm it with the user via AskUserQuestion:

markdown
## Problem Brief
**Product:** {one-line description}
**Target users:** {audience}
**Core problem:** {the pain being solved}
**Current alternatives:** {what exists today}
**Scale:** {expected scale}
**Key insight:** {what makes this solution different/better}

Ask: "Does this Problem Brief capture your vision correctly? (Yes / Adjust -- tell me what to change)"

Step 2: Requirements Deep-Dive

Targeted questions based on Step 1 answers. Use AskUserQuestion for each:

  1. Core workflows -- Ask:

> What is the SINGLE most important thing a user does?

Then ask clarifying sub-questions based on the answer:

  • Does this require authentication?
  • Is it real-time or async?
  • What data does it involve (structured, unstructured, files)?

Repeat for up to 2 more workflows: "What's the SECOND most important workflow?" and "Any third critical workflow?"

  1. Technical constraints (multiSelect):

> Which of these apply to your product?

Options: Auth required / Real-time updates / Offline support / Mobile-first / API integrations / Data persistence / Multi-tenancy / Privacy compliance (GDPR etc.) / None of these

  1. Anti-goals (multiSelect):

> What should we explicitly EXCLUDE from the MVP? These are common features that add scope without proving the core idea.

Options: Admin dashboard / Analytics / Payments / Notifications / Mobile app / User customization / Social features / Search / Internationalization

Step 3: Technical Architecture

Architecture questions tailored to the product type. Use AskUserQuestion for each:

  1. Application type (select one):
  • Web app / CLI tool / API service / Mobile app / Desktop app / Library/SDK
  1. Framework (options based on app type):

For Web app: Next.js / SvelteKit / Remix / Plain React + Vite / Astro / Other For CLI tool: Node.js (Commander) / Python (Click/Typer) / Rust (Clap) / Go (Cobra) / Other For API service: Express / Fastify / FastAPI / Actix / Go net/http / Other For Mobile app: React Native / Flutter / Swift / Kotlin / Other For Desktop app: Electron / Tauri / Other For Library/SDK: TypeScript / Python / Rust / Go / Other

Mark the best fit as "(Recommended)" based on requirements gathered in Steps 1-2. Explain why in one sentence.

  1. Database (select one):
  • PostgreSQL / SQLite / MongoDB / Redis / None (stateless) / Other
  • Recommend based on scale expectations and data model needs from Step 2.
  1. Deployment (select one):
  • Cloudflare Workers / Vercel / AWS / Self-hosted / Local only / Other
  1. Architecture pattern (select one):
  • Monolith (Recommended for MVP) / Microservices / Serverless functions / Hybrid

After gathering, add a Tech Stack section and confirm with the user:

markdown
## Tech Stack
**Type:** {app type}
**Framework:** {framework}
**Database:** {database}
**Deployment:** {platform}
**Architecture:** {pattern}
**Key rationale:** {why these choices fit the requirements}

Ask: "Does this tech stack look right? (Yes / Adjust -- tell me what to change)"


Phase 2: Design (Steps 4-6)

Step 4: Feature Ideation (Agent-Powered)

Deploy 2 Task agents in parallel (both with Plan subagent_type) with the full context from Steps 1-3.

Construct the prompts from the templates in AGENT_PROMPTS.md (located in this skill's directory). Read that file to get the prompt templates.

Agent 1: Product Architect

  • Proposes a structured feature set organized by user workflow
  • For each feature: name, description, complexity (S/M/L), dependencies, MVP-essential vs enhancement
  • Considers how the tech stack choices enable or constrain features

Agent 2: Domain Explorer

  • Thinks from the user's perspective
  • Suggests features the user hasn't considered: onboarding UX, error states as features, data lifecycle (import/export/backup), sharing/collaboration, accessibility, admin needs
  • Outputs "likely needed" vs "nice surprise" categories

After both agents return, synthesize their outputs into a unified, deduplicated feature list. Merge overlapping suggestions, preserve the best description, and track the source of each feature.

Step 5: Feature Presentation & Prioritization

Present the unified feature table to the user:

markdown
## Proposed Features
| # | Feature | Description | Complexity | Source | Suggested Tier |
| --- | ---------- | -------------------- | ---------- | ----------------- | -------------- |
| 1 | User auth | Email/password login | M | User | MVP |
| 2 | Dashboard | Main overview screen | M | Product Architect | MVP |
| 3 | CSV export | Download data | S | Domain Explorer | Future |

Source column values:

  • User -- explicitly requested by the user
  • Inferred -- derived from requirements (e.g., auth implied by multi-user)
  • Suggested -- proposed by an agent

Then use AskUserQuestion twice:

  1. (multiSelect): "Which features MUST be in your MVP? (select all that apply)"
  • List all features as options
  1. "Any features to ADD that are missing, or REMOVE entirely? (Add / Remove / Looks good)"
  • If Add: ask what to add, then incorporate and re-present
  • If Remove: ask which to remove
Step 6: Scope Definition & Approval

Organize features into two sections:

MVP Scope -- ordered by implementation sequence respecting dependencies:

markdown
### MVP Scope (v1)
| Order | Feature | Depends On | Complexity | Notes |
| ----- | ----------- | ----------- | ---------- | ------------------------- |
| 1 | Data models | -- | S | Foundation |
| 2 | User auth | Data models | M | Required by most features |

Future Roadmap -- grouped into release waves:

markdown
### Future Roadmap
#### v1.1 -- Quick Wins
| Feature | Complexity | Why deferred |
#### v1.2 -- Enhanced Experience
| Feature | Complexity | Why deferred |
#### Backlog -- Investigate Later
| Feature | Complexity | Why deferred |

Present both sections to the user, then use AskUserQuestion as a hard approval gate:

Ready to finalize this plan? Pick an option:
1. Approve -- proceed to documentation (Recommended)
2. Adjust scope -- move features between MVP and future
3. Adjust order -- resequence the implementation plan
4. Deep-dive -- expand a specific feature before approving

Handle each option:

  • Approve: Call ExitPlanMode to unlock file writes, proceed to Step 7.
  • Adjust scope: Ask which features to move and in which direction, update the tables, re-present, and ask for approval again.
  • Adjust order: Ask what to resequence, update, re-present, and ask again.
  • Deep-dive: Expand the specified feature with more detail (user stories, edge cases, technical considerations), then ask for approval again.

Loop until the user approves.

On approval, call ExitPlanMode to unlock file writes.


Phase 3: Document (Steps 7-8)

Step 7: Documentation Output

First, ask where to save using AskUserQuestion:

Where should I save the MVP planning documents?
1. ./docs/mvp/ (Recommended)
2. ./plans/
3. ./mvp/
4. Custom path

Create the chosen directory if it doesn't exist.

Produce 5 artifacts:

1. OVERVIEW.md -- Product Vision & Tech Stack
markdown
# {Product Name}
## Vision
{2-3 sentence product vision}
## Problem Brief
**Product:** {one-line description}
**Target users:** {audience}
**Core problem:** {the pain being solved}
**Current alternatives:** {what exists today}
**Scale:** {expected scale}
**Key insight:** {what makes this solution different/better}
## Tech Stack
**Type:** {app type}
**Framework:** {framework}
**Database:** {database}
**Deployment:** {platform}
**Architecture:** {pattern}
**Key rationale:** {why these choices fit}
## Product Principles
1.{principle 1}
2.{principle 2}
3.{principle 3}
## Competitive Landscape
| Alternative | Strengths | Weaknesses | Our Differentiator |
| ----------- | --------- | ---------- | ------------------ |
2. FEATURES.md -- Complete Feature Catalog

Every feature (MVP + future) with user stories, acceptance criteria, complexity, dependencies, and tier.

markdown
# Feature Catalog
## MVP Features
### {Feature Name}
-**Tier:** MVP
-**Complexity:** {S/M/L}
-**Dependencies:** {list or none}
-**Source:** {User/Inferred/Suggested}
**User Story:** As a {user type}, I want to {action} so that {benefit}.
**Acceptance Criteria:**
-[ ] {criterion 1}
-[ ] {criterion 2}
-[ ] {criterion 3}
---
## Future Features
### {Feature Name}
...
3. MVP_SCOPE.md -- Ordered Implementation Plan
markdown
# MVP Implementation Plan
## Implementation Sequence
| Order | Feature | Depends On | Complexity | Notes |
| ----- | ------- | ---------- | ---------- | ----- |
| 1 | ... | -- | S | ... |
## Architecture Decisions
-{Decision 1 and rationale}
-{Decision 2 and rationale}
## Explicitly Out of Scope
These items are intentionally excluded from the MVP:
-{item 1} -- {reason}
-{item 2} -- {reason}
## Mini-Specs
### {Feature 1}
{Brief technical approach -- 3-5 bullets on how to implement}
### {Feature 2}
...
4. ROADMAP.md -- Future Release Waves
markdown
# Product Roadmap
## v1.1 -- Quick Wins
| Feature | Complexity | Why Deferred | Trigger to Reconsider |
| ------- | ---------- | ------------ | --------------------- |
## v1.2 -- Enhanced Experience
| Feature | Complexity | Why Deferred | Trigger to Reconsider |
| ------- | ---------- | ------------ | --------------------- |
## Backlog -- Investigate Later
| Feature | Complexity | Why Deferred | Trigger to Reconsider |
| ------- | ---------- | ------------ | --------------------- |
## Open Questions
-{question 1}
-{question 2}

This is the start of the file. Read the full SKILL.md on GitHub.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Refreshed 17 Mar 2026, 7 months ago.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/nathan13888/nice-skills/mvp/badge)](https://verified-skill.com/skills/nathan13888/nice-skills/mvp)

Something wrong with this skill or its page? Tell us.