← ClaudeAtlas

first-principleslisted

Executable first-principles engineering method for AI agents. Run this on every task to question requirements, delete components, simplify/optimize, accelerate cycle time, and automate.
MasihMoafi/skills · ★ 3 · Code & Development · score 76
Install: claude install-skill MasihMoafi/skills
# First-Principles Engineering Method ## Intro - What is it? This skill translates first-principles engineering into a structured, 5-step agent execution protocol. It forces the agent to verify the correctness of the problem trajectory, aggressively trim code, minimize abstractions, and benchmark speed before writing any automation. ## Core Stance - **Correct Trajectory First:** The most rigorous phase is requirement checking. Finding the right problem to solve prevents wasting cycles on things that shouldn't exist. - **Strict Ordering:** Question -> Delete -> Simplify/Optimize -> Accelerate -> Automate. Never skip a step or optimize a process that should be deleted. - **Sovereign Execution:** Human steers, agent executes. State assumptions and halt at critical design junctions. --- ## The 5-Step Protocol ### 1. Question the Requirements (Finding the Trajectory) - **Goal:** Get the requirements as least wrong as possible. Do not solve a problem that shouldn't exist. - **Procedure:** - Assume every requirement in the issue or prompt is wrong or sub-optimal. - State all assumptions before proposing changes. - If the request has multiple plausible meanings, surface the interpretations instead of silently choosing one. - Ask when ambiguity makes the change risky or likely wrong. - Push back if a simpler or safer approach better serves the core objective. - Stop and name confusion immediately when code, runtime behavior, or requirements disagree. ### 2. Delete th