founder-skills

Adversarial idea interrogation followed by structured business modeling — in one chained pipeline.

Every existing startup skills repo is generative: you bring an idea and it helps you fill it out. This repo starts by trying to kill the idea first, then builds out the survivors.

Existing repos:   Interrogate OR Model
This repo:        Interrogate THEN Model

The Key Differentiator

/verdict — a synthesis skill that reads all prior outputs and returns an explicit BUILD / KILL / PIVOT judgment with structured reasoning. No other skills repo has this.

Pipeline

[/problem-discovery] ──> [/problem-statement] ──> /pressure-test ──> /customer-archetype ──┬──> /founder-fit
   (optional, if no                                    ↑                                    ├──> /scope-mode ──> /startup-canvas ──> /unit-economics
    specific idea yet)                          ⚠ run in fresh                              │                          │
                                                 conversation                               ├──> /market-size ─────────┤
                                                 to avoid bias                              │                          ├──> /regulatory-risk (regulated industries)
                                                                                            │                          ├──> /battle-cards
                                                                                            │                          └──> /assumption-map ──> /pretotype
                                                                                            └──────────────────────────────────────────────────────┘
                                                                                                                                                  ↓
All outputs ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────> /verdict
/verdict (BUILD) ──┬───────────────────────────────> /build-brief → coding agent
                   ├───────────────────────────────> /pitch-deck, /investor-lens, /elevator-pitch, /objection-map
                   └───────────────────────────────> /gtm-plan, /oneliner, /cold-outreach, /content-strategy
/verdict (KILL/PIVOT) ─────────────────────────────> /pivot-paths

⚠ Bias warning: /problem-discovery and /problem-statement help shape the problem — they are not adversarial. After running them, switch to a fresh Claude Code conversation before running /pressure-test. Otherwise the agent that just helped you frame the problem will be biased toward validating it, undercutting the adversarial review. This is a fundamental limit of LLM self-review, not a Claude-specific issue.

Every skill writes a structured output file to ./founder-outputs/. Downstream skills read those files. /verdict reads all of them.

Quick Start

1. Install the plugin (see below)
2. Open a project in Claude Code
3. Say "I have an idea" or run /pressure-test
4. Follow the pipeline — each skill recommends the next step

Installation

As a Claude Code plugin (recommended)

claude plugin add /path/to/founder-skills

Multi-platform install

The setup script auto-detects installed coding agents and installs the appropriate format for each: