# Prototype Plan

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-stanislavnianko-product-discovery-claude-skills-prototype-plan`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [stanislavnianko](https://agentstack.voostack.com/s/stanislavnianko)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [stanislavnianko](https://github.com/stanislavnianko)
- **Source:** https://github.com/stanislavnianko/product-discovery-claude-skills/tree/main/plugins/discovery-phase/skills/prototype-plan

## Install

```sh
agentstack add skill-stanislavnianko-product-discovery-claude-skills-prototype-plan
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Prototype Plan

> Part of the **discovery-phase** skill pack · `validation` group · reads `discovery-context.md` (run `profile-builder` first if missing).

A prototype proves user/product assumptions in a realistic environment. Built minimally, demoed deliberately, torn down by contract.

## Step 1 — Read context

Read `discovery-context.md` (section **3. Engagement** — mode + budget drive prototype ambition), `scope-doc.md` (prototype is a slice of MVP scope), `risk-assumption-map.md` (top-3 assumptions), and optionally `tech-spike-report.md`.

If `discovery-context.md` is missing, ask the BA inline: "who's running the prototype build — solo BA / BA + engineer / engineering lead?" — tag `[ASSUMED]`. If `scope-doc.md` is missing, proceed tagged `[NO-SCOPE-DOC]`. If `risk-assumption-map.md` is missing, ask for 1–3 assumptions inline or proceed tagged `[NO-RISK-FRAME]`. Never block; recommend `profile-builder` / `feature-scoping` / `risk-assumption-mapping` for high-stakes work.

**Reality checks (warn + confirm, do not halt):**
- If `tech-spike-report.md` says feasibility is **Red**, warn that the prototype will test politics, not tech. Offer: resolve the blocker first (re-spike / re-scope / pivot via `opportunity-mapping`), defer to delivery, or proceed with a `⚠ RED-FEASIBILITY-OVERRIDE` banner and aggressive tear-down plan.
- If context shows **no engineering capacity**, warn that the plan will be conceptual rather than runnable. Offer: loop in an engineer, defer to delivery, switch to `feasibility-spike` (smaller), or proceed solo with a `⚠ NO-ENGINEER-OVERRIDE` banner — Step 7 (runnable skeleton) will be skipped or stubbed.

## Step 2 — State the prototype objective

One sentence. End-to-end happy path.

> "The prototype lets a single SMB admin import a CSV, see preview, and confirm — in one session, on real data."

## Step 3 — Pre-committed success criteria

Before any code:

- Happy-path completion time: `` minutes
- `` users complete unassisted
- Specific signal `` crosses threshold ``

Without pre-commit, the team will declare success regardless.

## Step 4 — Cut scope (delta vs scope-doc)

The prototype is narrower than the MVP scope. Build a scope delta table:

| Area | MVP scope | Prototype scope | Why cut |
|---|---|---|---|

Cut MORE aggressively than feels comfortable. Polished UI, edge errors, scale, integrations — all default OUT for prototype.

## Step 5 — Architecture sketch

10-minute ASCII or mermaid. Identify:

- The prototype/production boundary (what's throwaway vs what could survive)
- External dependencies
- Happy-path data flow

## Step 6 — Generate construction plan

If the prototype is ≥3 build steps:

> "Use `blueprint` to generate a step-by-step construction plan for `` with the listed exclusions. One PR per step."

If ; success_criteria: ; build_steps: ; demo_ready: `.

## Anti-patterns

- **Building the MVP.** Prototype proves assumptions; MVP delivers value. Conflating them triples build time.
- **No tear-down plan.** Prototypes become production via inertia.
- **Polishing the UI** when assumptions are about backend / data / model.
- **Demo script written after build.** Writing demo first forces the build to serve the demo, not a feature list.
- **Skipping pre-committed success criteria.** Without them, "the prototype works" becomes whatever's most defensible post-hoc.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [stanislavnianko](https://github.com/stanislavnianko)
- **Source:** [stanislavnianko/product-discovery-claude-skills](https://github.com/stanislavnianko/product-discovery-claude-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** yes
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-stanislavnianko-product-discovery-claude-skills-prototype-plan
- Seller: https://agentstack.voostack.com/s/stanislavnianko
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.