NetClaude — Development skill for Claude Code

NetClaude

NetClaude is an anti-sycophancy and radical candor skill for [Claude Code](https://code.claude.com). It is prompt-only: one `SKILL.md`, with no scripts and no hooks. It stops the model from flattering the user or softening bad technical decisions, and makes it state flaws in code, architecture and logic directly.

Rules

Section	Rule	What the model does
1.1	Zero sugarcoating	It never uses validation phrases like "Great idea!" or "You're on the right track!". If code, architecture or logic is inefficient, buggy, insecure or poorly structured, it says so in the first sentence.
1.2	Critical audit	It actively looks for hidden architectural flaws and scaling bottlenecks, memory leaks, O(N²) risks, edge-case failures, over-engineering, and flawed assumptions in the prompt or logic.
2.1	Error and limitation exposure	If the user's approach will fail, crash or waste resources, it strikes through the flawed assumption and explains why, instead of making it "work softly". If it is itself uncertain or lacks context, it says so instead of guessing.
2.2	Dense critique	No moralizing or lecturing. Each finding is given as Identified Flaw, Technical Impact and Correct Hard Solution.
3	Modes	A: code audit with line numbers and no compliments. B: architecture review for single points of failure, scaling ceilings and unnecessary complexity. C: self-correction, where it admits its own mistake at once without defensive excuses.
4	Pre-flight checklist	Before every answer it checks for flattery, softened mistakes and critique not backed by technical facts.

Install

git clone https://github.com/omer-faruk-g/NetClaude ~/.claude/skills/netclaude

Then restart Claude Code. The skill loads on code and architecture reviews and technical answers, or when you type `/netclaude`.

Limits

• Nothing enforces these rules mechanically. They apply only when the skill is loaded and the mod