AgentSkillsCN

First Principles Analysis

将问题拆解为基本真理，从零开始重新构建解决方案。当您觉得常规方案行不通、成本似乎难以改变、耳边总听到“事情本就该这样办”，或者亟需突破性进展而非渐进式改进时，可使用此技能。

SKILL.md
--- frontmatter
name: "First Principles Analysis"
description: "Deconstruct problems to fundamental truths and rebuild solutions from scratch. Use when conventional solutions feel wrong, costs seem fixed, you hear 'that's how it's done,' or need breakthrough rather than incremental improvement."

First Principles Analysis

Overview

Break problems down to their fundamental truths, question every assumption, and rebuild solutions unconstrained by convention.

Work through six phases, but within each phase, have a natural dialogue. Ask one question at a time, follow interesting threads, and don't rush to the next phase until the current one feels complete.

When to Use This

  • •Existing solutions feel like cargo cult or "we've always done it this way"
  • •You hear "that's impossible" or "too expensive"
  • •Industry conventions don't make logical sense
  • •Need breakthrough rather than incremental improvement
  • •Something feels like inherited constraint vs actual constraint

The Process

Phase 1: Define the Problem

Don't skip this. Most failed analyses come from solving the wrong problem.

Start by understanding what we're actually trying to achieve. Ask questions like:

  • •What's the core problem in your own words?
  • •What would success look like concretely?
  • •Why does this problem exist?
  • •What prompted this now?

Stay here until the problem is crystal clear. If it feels fuzzy, keep asking.

Phase 2: Surface Assumptions

Now list everything we're assuming - especially the "obvious" stuff that nobody questions.

Explore together:

  • •What do we assume to be true about this problem?
  • •What's "just how it's done" in this space?
  • •What would someone with zero context find strange?
  • •What constraints do we take for granted?

Present assumptions back conversationally: "So we're assuming X, Y, and Z - does that capture it, or is there more?"

Phase 3: Question Each Assumption

This is the core of first principles. Take each assumption and interrogate it.

For each one, explore:

  • •Is this actually true? What's the evidence?
  • •Why do we believe this? Where did it come from?
  • •What if the opposite were true?
  • •Is this physics/logic, or just convention?

Have a genuine dialogue here. Some assumptions will survive scrutiny (those are fundamentals). Others will crumble (those were conventions masquerading as constraints).

Flag each as: Fundamental (keep) or Convention (question further)

Phase 4: Identify Fundamentals

After the questioning, take stock of what remains. These are your building blocks.

Reflect together:

  • •What constraints are actually real?
  • •What requirements are truly non-negotiable?
  • •What's left when we strip away the conventions?

Present: "Here's what I see as the irreducible fundamentals: [list]. Does this feel right?"

Don't proceed until there's agreement on the foundation.

Phase 5: Rebuild From Scratch

Now the creative part. Forget how it's currently done. Using ONLY the fundamentals, explore what's possible.

Guide the exploration:

  • •If we were starting fresh today, what would we build?
  • •How do completely different domains solve similar problems?
  • •What becomes possible now that we've dropped [convention]?

Propose 2-3 radically different approaches. Lead with your recommendation and reasoning, but present alternatives. Discuss trade-offs conversationally.

Phase 6: Validate

Before committing, stress-test the new approach.

Explore:

  • •Does this actually solve the core problem from Phase 1?
  • •What new assumptions have we introduced?
  • •What could go wrong?
  • •Is this actually implementable?

Be willing to loop back if something doesn't hold up.

Socratic Toolkit

Keep these in your back pocket for any phase:

When you need to...Ask...
Clarify"What do you mean by...?" / "Can you give an example?"
Challenge"What if that weren't true?" / "Who says?"
Probe evidence"How do we know?" / "Are there counterexamples?"
Shift perspective"How would an outsider see this?" / "What's the opposite?"
Test consequences"What happens if...?" / "What could go wrong?"

After the Analysis

Document it:

  • •Write to docs/plans/YYYY-MM-DD-<topic>-first-principles.md
  • •Capture: problem, assumptions challenged, fundamentals, new approach
  • •Commit to git

If implementing:

  • •Create plan from the new mental model
  • •Watch for old assumptions creeping back in during implementation

Key Principles

  • •Question everything - Especially the "obvious"
  • •One question at a time - Depth over breadth
  • •Follow threads - If something's interesting, explore it
  • •Seek fundamentals - Physics, logic, true requirements
  • •Ignore convention - "How it's done" isn't a reason
  • •Validate rigorously - New ideas need scrutiny too