Food Chain Ideation

The Problem

Founders validate ideas by asking friendly people friendly questions. The market is not friendly. What an idea needs before a line of code is written is a hostile environment — reviewers with different incentives, each hunting for the one flaw that kills it.

Food Chain Ideation builds that environment: an ecosystem of adversarial agents where your idea is the prey, and only theses that deserve to survive do.

How It Works

68 AI "animals" across three libraries attack an idea from every angle in live elimination rounds. Each attacker is isolated and role-locked, and after every patch it is forced to find a new angle — no recycling the same objection. Agents consume each other to consolidate arguments, and the idea gets patched between rounds.

Every battle ends in one of three outcomes: survived — the thesis held under attack; restructured — the original framing died but a pivot was found; or killed — the problem doesn't exist. The repo ships battle transcripts demonstrating all three.

Key Features

• 68 Adversarial Agents — Prompt-engineered animal personas across three libraries, each attacking from a distinct role and angle.
• Live Elimination Rounds — Attackers are isolated and role-locked; each patch forces them to find a fresh angle of attack.
• Agent Consumption — Agents consume each other to consolidate arguments, so the strongest objections rise up the food chain.
• Patch-Between-Rounds Loop — The idea evolves under fire; each round tests the patched thesis, not the original.
• Three Honest Outcomes — Survived, restructured, or killed — with battle transcripts included that demonstrate all three.
• One-Command Install — Published to the Claude Code plugin marketplace: claude plugin marketplace add CodedRichy/food-chain-ideation.

Tech Stack

Claude Code plugin/skill system · 68 prompt-engineered agent personas · Markdown battle transcripts