An adjacent discovery engine for cross-domain, divergent thinking. Surfaces non-obvious connections, hidden mechanism chains, and cross-field analogies that standard analysis misses.
Inspired by the AI Lateral Thinking Research Generator by jconorgrogan.
Standard analysis finds what's known. This skill finds what's connectable but not yet connected — mechanism transfers across distant fields, hidden causal chains, and structural analogies that practitioners in a single domain would never encounter.
It works on any domain: engineering, business strategy, product design, scientific research, writing, education, health, policy, and more.
- Ring 0-1 (skipped): Known solutions, standard advice, what a domain expert says first.
- Ring 2 (start here): Component decomposition — what regulates the problem that nobody is watching?
- Ring 3 (the point): Cross-domain raid — the same mechanism exists in a completely different field.
- "What are we missing?"
- "Go deeper" / "Think harder"
- "Think laterally" / "Cross-domain"
- Any situation where incremental improvements feel insufficient
Claude Code expects skills in a named subdirectory with a SKILL.md file:
mkdir -p ~/.claude/skills/lateral-thinking
cp lateral-thinking.md ~/.claude/skills/lateral-thinking/SKILL.mdOnce installed, the skill activates automatically when you use trigger phrases in Claude Code, or you can invoke it directly:
/lateral-thinking
The skill produces structured output with:
- Mechanism Skeleton — the raw dynamics of the problem, stripped of jargon
- Ring 2: Component Discoveries — non-obvious observations about the problem's primitives
- Ring 3: Cross-Domain Hypotheses — ranked by mechanistic plausibility × domain distance × testability
- Cross-Domain Pointers — specific fields and concepts to explore further
- Recommended Actions — what to try first, investigate, or discard
MIT