cross-lens

Cross-Lens

Cross-discipline thinking engine for builders. Scans your actual codebase and maps principles from behavioral design, complex systems, biology, game theory, and design patterns directly onto your code — producing insights no generic prompt could generate.

What It Does

You're building a product. You're deep in the code. Cross-Lens pulls you out of the technical bubble by connecting your specific codebase to principles from disciplines you've never studied:

• Your 5-step onboarding? Biology says immune systems verify identity in 3 stages, not 5.
• Your error messages? 11 of 14 say "Something went wrong" — Design says each error is your highest-stakes UX moment.
• Your pricing page? Game theory says a page with no recommended tier forces a coordination problem onto every visitor.

Every insight references your actual files, your actual code patterns, your actual metrics.

Installation

# Via Claude Code marketplace
/install cross-lens

# Via GitHub
claude --plugin-dir ./cross-lens

Commands

/cross-lens:spark — Full Analysis (~5 min)

Scans your codebase across 10 dimensions, scores 24 principles from 5 disciplines, detects cross-discipline contradictions, generates a dangerous idea, and proposes 3 falsifiable experiments.

Output: A CROSS-LENS-REPORT.md with:

• Summary card (30-second overview)
• Dangerous idea (the one insight that challenges your assumptions)
• 4-6 principle-to-code mappings (each grounded in specific files)
• Contradiction detector (where disciplines disagree about your code)
• 3 experiments with minimum viability measurements

/cross-lens:quick — Quick Insight (~2 min)

One discipline, one insight, one experiment. For when you need a fast perspective shift.

The 5 Disciplines

Discipline	Sources	What It Sees
Behavioral Design	BJ Fogg, Robert Cialdini	Friction, motivation, social proof, decision architecture
Complex Systems	Donella Meadows	Feedback loops, leverage points, delays, accumulations
Biology / Resilience	Nassim Taleb, Stuart Kauffman	Antifragility, redundancy, adaptation, immune patterns
Game Theory	Thomas Schelling	Defaults, commitments, incentive alignment, information asymmetry
Design Patterns	Don Norman, Christopher Alexander	Affordances, error communication, mistake-proofing, pattern language

What Makes This Different From a Prompt

You could ask any LLM "apply biology to my product." Here's what you can't get from a prompt:

1. Codebase scanning. Cross-Lens reads your auth flow, counts your onboarding steps, inventories your error messages, maps your dependencies. A prompt has no files to read.
2. Structural mapping. Not "consider antifragility" but "your retry logic in services/api.ts fails after 3 attempts — antifragile systems get stronger from failure, here's a circuit-breaker pattern."
3. Contradiction detection. It finds where two disciplines give opposite advice for the same code — that's where the real insight lives.
4. Quality gates. Every insight passes an anti-horoscope check: if it could be generated without reading your codebase, it gets dropped and replaced.

Anti-Horoscope Guarantee

Every Cross-Lens output must:

• Reference a specific file, count, or structural finding
• Be impossible to generate without codebase access
• Include directional predictions, not fake percentages
• Provide a minimum viability measurement (works without analytics tools)
• Include a null-result interpretation

If an insight fails any of these, it doesn't make the report.

Who This Is For

• Solo founders building products end-to-end who need perspectives beyond code
• Engineers who ship features and want to think about what they're building, not just how
• Indie hackers who code AND market AND design AND support — all at once

Post-Report Interaction

The report is the start of a conversation, not a static document. After delivery, you can:

• "dig into #3" — deeper analysis of a specific insight
• "show me the code change for experiment 1" — actual file edits
• "what else about game theory?" — explore principles that didn't make the cut

License

MIT