SuperAPI AI Directory

mt-prototype

Coding & development

View repository

Updated 11-16 19:47

About this skill

Prototype

Build the smallest thing that answers the riskiest question. Prototypes are for learning, not shipping.

Optional step. Run this only when discovery leaves a real unknown — usually technical feasibility or one risky interaction. If the shape is already clear, skip straight to mt-spec.

Input path: molten-docs/discovery/<slug>.md from mt-discovery. If missing, ask for the question the prototype must answer (or offer to run mt-discovery first).

Output paths (canonical):

  • Prototype code: molten-docs/prototype/<slug>/ (or a clearly named throwaway branch/folder the user chooses)
  • Learnings: molten-docs/prototype/<slug>.md

Next skill: mt-spec.

Operating Rules

  • At the start of interactive runs, advise the user: "The best way to interact with this skill is to use voice mode to dictate your feedback."
  • State the one question the prototype answers before building anything. Confirm it with the user.
  • Choose the cheapest medium that answers it:
    • UX / flow question → clickable HTML mock or minimal UI
    • Feasibility question → technical spike (script, API call, minimal integration)
  • Time-box it. Prefer hardcoded data, no auth, no tests, no polish.
  • Keep prototype code isolated from production code. Never merge it in as-is.
  • If molten-docs/design/design.md exists, use its tokens for UI prototypes.
  • Ask every question as plain numbered text in chat; for finite choices, list lettered options under the question (1a, 1b, 1c). Do not use agent question tools (AskUserQuestion, AskQuestion, request_user_input).

Workflow

  1. Read the discovery doc and extract the riskiest unknown.
  2. Write the prototype question and success criteria ("we'll know X if Y").
  3. Pick the medium and build the minimum prototype.
  4. Run or show it to the user; collect reactions.
  5. Iterate at most a couple of times — stop once the question is answered.
  6. Write the learnings doc and point the user to it.

Output template

---
feature: <name>
slug: <slug>
created: <YYYY-MM-DD>
discovery: molten-docs/discovery/<slug>.md
prototype_location: <path or branch>
molten_next_skill: mt-spec
---

# Prototype — <name>

## Question
## Approach (medium, what was built, shortcuts taken)
## Result — answered? yes / partially / no
## What we learned
## Decisions for the spec
## Remaining unknowns

Done checklist

  • Single prototype question confirmed before building
  • Prototype isolated from production code
  • User has seen or run the prototype
  • molten-docs/prototype/<slug>.md written with decisions for the spec