Tweak Idea: score a startup idea across 14 dimensions

What it is

Tweak Idea is an open-source skillset for Claude Code. You give it a startup concept, and it evaluates that concept across 14 weighted dimensions — pain, willingness to pay, market size, defensibility and more — then synthesizes the results into a scored report you can act on. The goal is a fast, repeatable read on whether an idea is worth pursuing, without spending days on a manual deep dive.

The problem it solves

Judging an idea usually means answering hard questions one at a time. Is there a real market? Will customers actually pay? Do incumbents already own the space? Each one is its own research project, so most ideas get judged on intuition instead. Tweak Idea runs those questions in parallel and grounds the answer in evidence rather than gut feel.

How it works

The pipeline moves an idea through a few bounded stages:

1. 1

   Capture

   Records the problem and proposed solution, and confirms founder–market fit so the evaluation is anchored to who's building it.

2. 2

   Extract hypotheses

   Pulls the testable claims out of your pitch and asks you to confirm which ones you can back up. Unsupported claims don't get to inflate the score.

3. 3

   Research

   Searches the market — competitors, demand signals, comparable products — to build shared evidence for the scoring stage.

4. 4

   Score

   Fourteen isolated agents each evaluate a single dimension, drawing on the shared research or running their own targeted searches. Isolation stops one strong dimension from coloring the others.

5. 5

   Synthesize

   A final agent combines the dimension scores into a weighted scorecard — actual score, potential score, strengths, weaknesses, and next steps ranked by impact.

Most of the repetitive work runs as deterministic Python scripts with fixed JSON schemas, and the LLM is reserved for the judgment calls. That split keeps runs fast and the output consistent from one idea to the next — a full evaluation lands in roughly ten minutes.

The 14 dimensions

Every idea is scored on the same set of weighted dimensions:

• Pain intensity
• Willingness to pay
• Solution gap
• Founder–market fit
• Urgency
• Frequency
• Market size
• Defensibility
• Market growth
• Scalability
• Target-customer clarity
• Behavior-change difficulty
• Mandatory nature
• Incumbent indifference

Using it

You submit an idea with a command and get back a scored report with ranked validation priorities. A handful of supporting commands round out the workflow:

• list see every idea you've evaluated
• show reopen a past report without re-running it
• compare diff two iterations side by side to see what moved
• improve get concrete tweaks for the weakest dimensions
• browse-hn surface fresh idea candidates from technology shifts

Why it's built this way

The design bet is reliability: give the model small, bounded jobs and let plain code handle everything repeatable. That keeps results reproducible, makes runs fast, and means the scorecard you get is consistent enough to trust when you compare one idea against another.

Try it

Tweak Idea is open source. See a real run on the skill page, or read the code and install it from GitHub.