A skill for AI coding agents that strips any assumption down to its physics/math/logic floor, calculates the convention tax, and rebuilds from reality.
Works with Claude Code, Claude.ai, OpenAI Codex CLI, Gemini CLI, Cursor, and Amazon Kiro.
Most of what any industry accepts as fixed is a set of agreements nobody has challenged recently enough.
This skill runs a 6-stage destruction process on any statement, business model, process, or industry belief:
- Assumption Inventory — finds every hidden assumption (5–15 typical)
- Reality Floor — strips convention; what do physics/math/logic actually require?
- Convention Tax Calculator — what are you paying to maintain each assumption?
- Assumption Removal Test — what breaks, what costs more, what turns out to be purely optional?
- Rebuilt Version — reconstruct from reality, not habit
- First Mover Question — if this rebuilt version is possible, why doesn't it exist yet?
"Enterprise software has to have a long sales cycle"
The destructor finds 9 hidden assumptions, calculates the convention tax (avg 14 months and $280k CAC per deal), removes non-load-bearing ones, and rebuilds a product-led growth model with 3-week time-to-close.
./install.sh --claude-globalThen in any Claude Code session: /first-principles [your statement]
./install.sh --claude-project./install.sh --gemini./install.sh --codex./install.sh --cursorAdds .cursor/rules/first-principles.mdc — auto-activates when you ask to challenge an assumption.
./install.sh --kiroAdds .kiro/steering/first-principles.md — applied to every Kiro session in this project.
./install.sh --all-global./package.shThen go to claude.ai > Customize > Skills > + > Create skill and upload first-principles.skill.
Copy SKILL.md to your agent's skills directory:
| Agent | Global skills path |
|---|---|
| Claude Code | ~/.claude/skills/first-principles/ |
| Gemini CLI | ~/.gemini/skills/first-principles/ |
| OpenAI Codex CLI | ~/.codex/skills/first-principles/ |
| Agents Standard | ~/.agents/skills/first-principles/ |
Once installed, trigger by saying:
/first-principles [statement or belief]"Run first principles on: [idea]""Destroy this assumption: [assumption]""What's the convention tax on [process]?""What's the reality floor for [belief]?"
first-principles-destructor/
├── SKILL.md # Universal (Claude Code, Codex, Gemini CLI)
├── .cursor/
│ └── rules/
│ └── first-principles.mdc # Cursor
├── .kiro/
│ └── steering/
│ └── first-principles.md # Amazon Kiro
├── install.sh # Installer for all platforms
├── package.sh # Build .skill ZIP for claude.ai upload
└── README.md
| Platform | Format | Status |
|---|---|---|
| Claude Code | SKILL.md |
Supported |
| Claude.ai (browser) | .skill ZIP |
Supported |
| OpenAI Codex CLI | SKILL.md |
Supported |
| Gemini CLI | SKILL.md |
Supported |
| Cursor | .cursor/rules/*.mdc |
Supported |
| Amazon Kiro | .kiro/steering/*.md |
Supported |
| GitHub Copilot | SKILL.md (partial) |
Partial |
MIT