SuperAPI AI Directory

idea-refine

General & other

View repository

Updated 10-16 06:43

About this skill

Idea Refine

Overview

Turn vague product ideas or technical concepts into concrete, evaluable proposals through structured divergent and convergent thinking. This skill runs before spec-driven-development when the problem space is still ambiguous.

When to Use

  • A feature request is vague or open-ended
  • You have competing approaches and need to evaluate tradeoffs
  • A stakeholder describes a problem without a clear solution
  • You need to explore the design space before committing to a direction

Process

Step 1: Problem Framing

State the problem clearly:

  • Who is affected
  • What outcome they need
  • What constraints exist
  • What success looks like

Step 2: Divergent Exploration

Generate multiple approaches without judgment:

  • List at least three substantially different solutions
  • Include at least one unconventional approach
  • Note assumptions each approach depends on
  • Identify risks unique to each approach

Step 3: Convergent Evaluation

Score each approach against:

  • User impact: how well it solves the stated problem
  • Implementation cost: time, complexity, and dependencies
  • Risk: what could go wrong and how recoverable it is
  • Extensibility: how well it supports future related work

Step 4: Proposal

Write a concise proposal covering:

  • Recommended approach and why
  • Key tradeoffs accepted
  • Assumptions that need validation
  • Open questions for the team

Common Rationalizations

Rationalization Reality
"The first idea is usually fine" First ideas tend to be the most obvious, not the most effective.
"We can refine later" Later refinement is more expensive than early exploration.
"This is too small for structured thinking" Even small features benefit from a minute of structured evaluation.

Verification

  • Problem is stated clearly with affected users
  • At least three approaches were considered
  • Each approach was scored against consistent criteria
  • A clear recommendation exists with accepted tradeoffs documented

Anti-Rationalization Table

Excuse Counter
"The first idea is usually fine" First ideas tend to be the most obvious, not the most effective.
"We can refine later" Later refinement is more expensive than early exploration.
"This is too small for structured thinking" Even small features benefit from a minute of structured evaluation.
"Three approaches is too many for this" Two approaches is the minimum for comparison. Three reveals non-obvious options.
"The stakeholder already knows what they want" Stakeholders describe problems, not solutions. Explore the solution space.