agentleFS
calvyntwhcalvyntwh/karu-custom-skills/skills/first-principles-thinking/SKILL.md

first-principles-thinking

Skill1 starsChanged 2 months agoMITView on GitHub

A mental model for breaking down basic truths to generate original solutions. Use for innovation, novel problems, or when stuck in conventional thinking.

16 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. 0 are common, 18 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 15 edited versions we read keep these.

  • 18

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Innovation: When the user wants to invent a new way of doing things.

    line 16
  2. Distinctive:

    Stuck Points: When standard solutions are too expensive, too slow, or impossible.

    line 17
  3. Distinctive:

    Debugging: When "it should work" but doesn't—strip away assumptions about why it *should* work.

    line 18
  4. Distinctive:

    Cost Analysis: When you need to understand the absolute floor of a cost structure.

    line 19
  5. Distinctive:

    Well-trodden problems: When analogical reasoning is sufficient and the cost of being wrong is low.

    line 22
  6. Distinctive:

    When expertise is available: If a domain expert can be consulted and the problem is within their validated expertise, rely on their map.

    line 26
  7. Distinctive:

    Before applying FPT, confirm:

    line 29
  8. Distinctive:

    Before FPT: Use decision-matrix to confirm FPT is warranted (high-stakes, innovative)

    line 36
  9. Distinctive:

    After FPT: Use systems-thinking to check feedback loops and emergent behavior

    line 37
  10. Distinctive:

    Combined Protocol: When facing adversarial contexts, chain inversion-thinking → FPT

    line 38
  11. Distinctive:

    Complementary: Use rubber-ducking to verify plain-English explanation of atomic units

    line 39
  12. Distinctive:

    Avoid over-use: If FPT hits deadline frequently, the problem may be too ill-defined—consider escalating

    line 40
  13. Distinctive:

    When "Why?" yields an answer, classify it before proceeding:

    line 48
  14. Distinctive:

    Before questioning an assumption, estimate:

    line 62
  15. Distinctive:

    Technique: Ask "Why?" up to 5 times.

    line 70
  16. Distinctive:

    CRITICAL: Do not answer from memory.

    line 71
  17. Distinctive:

    Use systems-thinking if feedback loop analysis is complex.

    line 99
  18. Distinctive:

    Document these as explicit limits.

    line 107

“Common” is measured: kept by at least 6 in 10 of the 15 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

First Principles Thinking

Overview

First Principles Thinking is a mode of inquiry that relentlessly questions assumptions to get to the fundamental truth of a problem, then builds a solution from scratch. It is the tool of the innovator, used to bypass "reasoning by analogy" (copying what others do).

When to Use

  • Innovation: When the user wants to invent a new way of doing things.
  • Stuck Points: When standard solutions are too expensive, too slow, or impossible.
  • Debugging: When "it should work" but doesn't—strip away assumptions about why it should work.
  • Cost Analysis: When you need to understand the absolute floor of a cost structure.

When NOT to Use

  • Well-trodden problems: When analogical reasoning is sufficient and the cost of being wrong is low.
  • Quantum/Biological systems: FPT assumes independence of atomic units; quantum entanglement and biological emergence violate this.
  • Social/Human systems: Human behavior is not decomposable into physics axioms—consult domain experts instead.
  • Time-critical decisions: FPT is computationally expensive; under deadline pressure, use decision-matrix or analogical reasoning.
  • When expertise is available: If a domain expert can be consulted and the problem is within their validated expertise, rely on their map.

Domain Boundary Checklist

Before applying FPT, confirm:

  • Problem is mechanical, supply-chain, or computational (not quantum/biological/social)
  • Atomic units are empirically verifiable (not model-dependent)
  • Cost of being wrong is high enough to warrant full deconstruction
  • Time available exceeds 2x estimated analysis duration

Skill Integration Cross-References

  • Before FPT: Use decision-matrix to confirm FPT is warranted (high-stakes, innovative)
  • After FPT: Use systems-thinking to check feedback loops and emergent behavior
  • Combined Protocol: When facing adversarial contexts, chain inversion-thinking → FPT
  • Complementary: Use rubber-ducking to verify plain-English explanation of atomic units
  • Avoid over-use: If FPT hits deadline frequently, the problem may be too ill-defined—consider escalating

Methodology

1. Identify & Question Assumptions (Socratic Method)

Goal: Expose the hidden beliefs limiting the current view.

Answer Classification

When "Why?" yields an answer, classify it before proceeding:

TypeAction
EmpiricalSearch/Read to verify—do not accept on memory
Social normNote as context-dependent; may shift
Value judgmentAcknowledge; not falsifiable—move on
CircularRe-ask with different framing

Timeboxing

  • Maximum: 5 "Why?" iterations OR 30 minutes, whichever comes first.
  • Stop at: (a) verified physical law, (b) unverified claim you cannot verify with Search/Read.
  • Deadline fallback: Accept current best assumption with confidence label (verified/assumed/unknown) and proceed. Log deadline hits.

Pareto Prioritization

Before questioning an assumption, estimate:

  • Centrality: How core is this to the problem? (1-5)
  • Likelihood of being wrong: (1-5)
  • Priority Score: Centrality × (5 - Likelihood)

Question assumptions with Priority Score ≥ 15 first. Low-likelihood assumptions are often correct—skip them.

  • Prompt: "What are we assuming to be true here? Why do we think that?"
  • Technique: Ask "Why?" up to 5 times.
    • CRITICAL: Do not answer from memory. Search/Read to verify the answer to each "Why?".
    • Stop: When you hit a physical law, a mathematical truth, or a verified hard constraint.

2. Deconstruct to Constituent Parts

Goal: Find the atomic units that cannot be reduced further.

  • Action: Break the system down into:
    • Physics: Mass, energy, material properties.
    • Logic: Axioms and necessary truths.
    • Economics: Raw material costs, time units.

Validation Step

Label each atomic unit as:

  • Verified — empirically confirmed under relevant conditions
  • Model — theoretical construct (e.g., economic "spot price") with known limitations
  • Assumed — no verification attempted

For Model/Assumed units, note confidence interval and conditions where they may not hold.

3. Reconstruct from Scratch

Goal: Build a 10x better solution using only the essential parts.

  • Action: Ignore "how it's usually done." Combine the atomic units in the most efficient way possible to achieve the goal.

Second-Order Checkpoint

Before implementing, ask:

  1. What new interactions does this solution create?
  2. What feedback loops might emerge when it scales 10x?
  3. What breaks if this solution becomes widely adopted?

Use systems-thinking if feedback loop analysis is complex.

Boundary Conditions

This framework rests on unquestioned assumptions:

  • Breaking problems down is useful (may fail for emergent systems)
  • Atomic units exist (may fail for quantum entanglement)
  • Deconstruction reveals truth (may fail for social/human systems)

Document these as explicit limits. Re-evaluate if problem domain shifts.

Examples

The SpaceX Example (Cost Reduction)

User: "Rockets are too expensive." Analogy Reasoning: "Rockets define the market price. We can maybe save 10% by negotiating." First Principles Reasoning: "A rocket is just aluminum, titanium, and fuel. The spot price of these materials is 2% of a rocket's cost. The cost is inefficient manufacturing. We will buy raw metal and build it ourselves."

The "Chef vs. Cook"

Analogy (Cook): Follows a recipe. If an ingredient is missing, they stop. First Principles (Chef): Understands why the acid balances the fat. If a lemon is missing, they use vinegar, because the principle is acidity.

Self-Improvement Protocol

Log only if you found something genuinely new.

## [YYYY-MM-DD] {Brief Description}
**Assumption challenged**: {what was assumed}
**Breakthrough**: {what you found instead}
---

Promote after 3+ similar discoveries.

Resources

  • Detailed Research Notes

Evaluations

Eval 1: Cost Reduction Problem

Scenario: User says "Our CD pipeline is too expensive." Conventional thinking suggests "negotiate with vendor." Expected: Breaks down to atomic costs (servers, bandwidth, labor), questions vendor markup assumption. Pass criteria: Reaches "2% of cost is raw materials" level decomposition, proposes building in-house.

Eval 2: Domain Boundary Check

Scenario: User wants to apply FPT to "how should we design our company culture?" Expected: Recognizes as social/human system, advises using domain expert instead. Pass criteria: Identifies domain as inappropriate for FPT, redirects appropriately.

Eval 3: Inversion-FPT Chain

Scenario: User asks "How could we accidentally build a monopoly?" in context of new product design. Expected: Chains inversion-thinking → FPT, uses saboteur mode first, then decomposes. Pass criteria: Applies inversion before FPT in adversarial context, identifies how to prevent the failure mode.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.