whywhy

Drills repeated why-questions into a goal or problem to expose root causes and better directions.

4|Updated Jun 9, 2026
One-click install
npx skills add https://github.com/laicluse/agent-fieldkit --skill whywhy-laicluse
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: whywhy
Source: https://github.com/laicluse/agent-fieldkit/tree/main/.agents/plugins/generated/whywhy/skills/whywhy
Command: npx skills add https://github.com/laicluse/agent-fieldkit --skill whywhy-laicluse

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Decisions often feel unclear, goals stay vague, and problems get patched at the surface instead of fixed at the root. This Skill applies the Toyota 5 Whys method, extended to 10 layers by default, to force a structured chain of self-questioning that uncovers the real motivation or root cause behind a statement. ## Core Features & Use Cases - Layered Why-Chain: Asks a sharp, specific "why?" at each layer, answers it, and builds the next question on the previous answer, with a configurable depth of 3 to 20 layers. - Pattern Analysis: Detects convergence, breakpoints, circles, and deepening across the chain to identify the core theme or hidden assumption. - Reframing and Direction: Produces a reframed goal from the deepest layer plus a concrete next-direction proposal. - Use Case: A developer asks "/whywhy 5 why is this PR so large?" and the chain reveals the team lacks a shared definition of done, reframing the fix from splitting the PR to changing the review workflow. ## Quick Start Ask the agent to run /whywhy followed by your question or goal, for example "/whywhy why does our deploy keep failing on Fridays?"

Frequently Asked Questions about whywhy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a 5 Whys root cause analysis with an AI agent?▼

Invoke /whywhy followed by your question or statement, optionally prefixed with a layer count. The agent asks a specific why-question at each layer, answers it, and builds the next question on the previous answer, then analyzes the full chain for patterns.

How many why layers should I use for root cause analysis?▼

The default is 10 layers, extending Toyota's classic 5 Whys. You can set any count from 3 to 20 by passing an integer first, such as /whywhy 5; values outside that range are clamped to the boundary.

Can the why-chain use my codebase instead of just model reasoning?▼

Yes. The instructions require answers to come from verifiable sources where relevant, so the agent uses code search, file reading, and web tools rather than relying purely on model weights for questions like why a deploy runs slowly.

What patterns does the analysis look for in a why-chain?▼

It checks for convergence where layers point to one theme, breakpoints where an answer changes direction revealing an assumption, circles where an answer repeats an earlier layer, and deepening where each answer goes further than the last.

When should I not use repeated why-questioning?▼

Avoid it when you need a quick factual answer or an immediate fix rather than reflection. The method is designed for unclear decisions, vague goals, and root-cause exploration, not for tasks with a known deterministic solution.