davehallmon/reasoning-lens

Organon: thinking tools by Ockham.
A Claude Skill that runs your idea past seven philosophers' ways of reasoning, one at a time, so you see what each one notices, and where they disagree.
> Experimental pre-release (0.2.1). No eval rounds have run yet. See How It Will Be Tested.
## Why
Ask for "a few perspectives" and you usually get one answer said seven ways. reasoning-lens makes each lens commit to a stance, then lists only the disagreements that come from what each lens actually checks. When the lenses agree, it says so instead of inventing conflict.
The seven reasoning methods come from a published study: Harb et al. (2026), The ballad of LLM agents: philosophical reasoning for chemistry. The research boundary matters: the study tested the methods independently on chemistry questions; reasoning-lens extends them into a sequential sweep for general ideas.
## Research Foundation
reasoning-lens takes its starting point from Harb et al. (2026), The ballad of LLM agents: philosophical reasoning for chemistry. The study encoded seven philosophy-inspired reasoning methods as fixed system prompts and tested them independently on numerical chemistry questions.
Philosophical prompting changed model performance.
Harb et al. (2026) tested seven philosophy-inspired system prompts on 243 numerical ChemBench questions. Different reasoning methods helped different models and problems; no single philosopher was universally best.
reasoning-lens starts from those seven reasoning patterns, but asks a different question: what happens when they are orchestrated as a sequential reasoning system for general problems?
Read the paper →
Publisher / DOI →
Research boundary →
| Harb et al. (2026) | reasoning-lens |
| Chemistry benchmark | General ideas, plans, claims, and decisions |
| Seven philosophy agents tested independently | Seven reasoning lenses run sequentially |
| Numerical-answer accuracy | Disagreement calibration and resolution conditions |
| Tests whether philosophy-inspired prompting changes model performance | Tests whether orchestration adds useful thinking beyond the lens cards |
The distinction is deliberate: published findings are treated as evidence for the individual reasoning patterns, not as proof that this Skill's orchestration works. That claim is tested separately in this repository.
## What It Does
Restates your idea in one sentence, with any assumptions stated.
Runs seven lenses in order: Socrates, Plato, Aristotle, Descartes, Hume, Kant, Hegel. Each says what it notices in three sentences or fewer and commits to a stance: Supports, Supports if, Challenges, or Can't judge yet.
Lists one to three real disagreements, each with the fact or value it turns on. Or says "They mostly agree."
Gives you two ways forward:
Run with it: what to do now, which side it takes, and the signal that would prove it wrong. Usable without replying.
Investigate further: the one question or test that would settle the biggest disagreement.
Skips the sweep for rewrites, lookups, definitions, code, and one-answer problems. It answers directly and offers the sweep anyway.
## How It Works
reasoning-lens keeps the seven methods independent during the sweep, then compares their committed stances to find genuine disagreement and a useful next step.