# First Principles

> First-principles thinking framework. Triggers when users need to analyze the essence of problems, make important decisions, evaluate options, question assumptions, or understand complex systems. Can be combined with any other skill as an underlying analytical method.

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

---


# First-Principles Thinking Skill

## Definition

Break down problems to their most basic facts, then re-derive solutions from there — rather than relying on analogies, conventions, or "that's how everyone does it."

> Core question: "What is the fundamental truth about this? Starting from that truth, what can I derive?"

## When to Activate

**Activation conditions (any one triggers):**
- Existing approaches may be wrong, but no one has questioned them
- The problem is new with no good precedents
- High-stakes decision worth deep analysis
- User is stuck between contradictory advice
- Need to understand "why it works" not just "does it work"

**Do not activate when:**
- Problem has a clear standard answer (just look it up)
- Routine operations with proven processes
- Extreme time pressure requiring immediate action

Knowing when NOT to use it is part of first-principles thinking.

## Analysis Process

### Step 1: Identify the Purpose

Ask: **"What is this ultimately trying to achieve?"**

Not "what am I being asked to do" but "what outcome is this action meant to produce."
If the purpose itself is flawed, all subsequent analysis is meaningless.

### Step 2: Break Down to Basic Facts

Ask: **"What do I know for certain to be true about this?"**

Rules:
- Accept only verifiable facts, not assumptions, conventions, or "usually"
- Distinguish "facts" from "interpretations of facts"
- Distinguish "causation" from "correlation"

### Step 3: Identify Hidden Assumptions

Ask: **"What unexamined assumptions am I accepting?"**

Common hidden assumptions:
- "This has always been done this way" (convention ≠ optimal)
- "Everyone says so" (consensus ≠ fact)
- "It worked last time so it will work again" (history ≠ future)
- "This is too complex to change" (complex ≠ immutable)

### Step 4: Re-derive from Basic Facts

Ask: **"Given only these facts, uninfluenced by existing solutions, what would I do?"**

Build a solution from scratch, then compare with existing approaches. The differences are where deeper examination is needed.

### Step 5: Validate the Derivation

Ask: **"Which step in my reasoning is weakest? If that step is wrong, what happens to my conclusion?"**

- Find the weakest link in the reasoning chain
- Test: if that link reverses, does the conclusion still hold?
- If not, that's where more information or careful judgment is needed

## Common Pitfalls

| Pitfall | Symptom | Solution |
|---------|---------|----------|
| Fake first principles | Treating "industry wisdom" as basic facts | Ask one more layer: "Where did this come from?" |
| Analysis paralysis | Endless decomposition, never deciding | Set a depth limit; decide with best available info |
| Ignoring practicality | Deriving theoretical optima that can't be executed | Include "feasibility" as a basic fact |
| Dogmatism | Using only your own reasoning, ignoring others' experience | First principles doesn't reject experience; it rejects *unexamined* experience |

## Combining with Other Skills

This skill is a foundational methodology. Combine with other skills as follows:

1. **First, use first principles to clarify the problem's essence** (What is this really trying to solve?)
2. **Then, use domain skills for specialized frameworks** (What proven methods exist in that domain?)
3. **Finally, use first principles to validate conclusions** (Does every step of this solution hold up?)

## Response Principles

- Don't pretend things are simple — if genuinely complex, explain where the complexity lies
- Distinguish known from unknown — clearly mark "this is fact" vs "this is my inference"
- Show the reasoning process, not just conclusions — let users verify each step
- Challenge users' hidden assumptions, but through questions rather than assertions