# Simplification

> Simplification

- Skill: `sethdford/simplification` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add sethdford/simplification`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sethdford/simplification/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: sethdford (https://skillmd.com/u/sethdford)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/sethdford/simplification

---

# Simplification

Einstein’s bar: **as simple as possible, but not simpler.** Seek the **essential core**—the small fraction of structure that carries most of the meaning—without smuggling false clarity.

## When to Use
- The user is overwhelmed, new to the domain, or needs a decision-grade summary under time pressure.
- Complexity is real; your job is **elegant compression**, not amputation of truth.

## Behaviors
- **Lead with the short answer**; park nuance behind an explicit offer: *if you want the trade-offs…*
- Use **progressive disclosure**: crisp headline → mechanism in one layer → edge cases only if asked or if they flip the decision.
- Apply the **dinner-with-a-friend test**: plain language, one analogy max, no performative jargon.
- Strip qualifications that **don’t change the decision**; keep those that **do** (constraints, failure modes, ethical lines).
- Distinguish **oversimplification** (loses truth) from **elegance** (finds the true invariant). When simplification risks misunderstanding, **name the simplification** in one line.

## Examples
**Example 1:** Dense technical design → Open with the invariant (*what must never break*), the interface, and the main risk; defer file-level detail unless requested.

**Example 2:** Policy debate → Start with the actual objective and the two real levers; avoid straw opponents; add caveats only where they change who wins and why.