Live: Claude Code Opus 4.6 builds a FastAPI project. Quality lands at 6772.
Not because the agent can't do better — because without a sensor, it doesn't know what to improve.
// The problem nobody talks about
You start a project with Claude Code or Cursor. Day one is magic.
The agent writes clean code, ships features fast.
Then something shifts. The agent starts hallucinating functions
that don't exist. It puts new code in the wrong place. You're
spending more time fixing the agent's output than writing it yourself.
Everyone assumes the AI got worse. It didn't. Your codebase did.
The code your AI writes today is the context it reads tomorrow.
Every session silently degrades your architecture. Without a sensor,
nobody notices until it's too late.
// 5 root cause metrics. One score.
Geometric mean of 5 independent graph-theoretic dimensions.
Ungameable by design — Nash Social Welfare theorem (1950).
modularity
Newman 2004 · graph community detection
Do files cluster into independent modules? Replaces coupling, cohesion, and god file detection.
acyclicity
Martin 2003 · Acyclic Dependencies Principle
Are there circular dependencies? Tarjan's strongly connected component detection.
depth
Lakos 1996 · levelization
How deep are dependency chains? Deep chains amplify change propagation.
equality
Gini 1912 · inequality coefficient
Is complexity evenly distributed? God files are the #1 source of AI agent confusion.
redundancy
Kolmogorov 1963 · complexity theory
Dead or duplicate code. Every unnecessary line is noise for the AI agent.
// Install
terminal
# macOS brew install sentrux/tap/sentrux
# Linux curl -fsSL https://sentrux.dev/install.sh | sh
# Windows cargo install sentrux
# or build from source (any platform) git clone https://github.com/sentrux/sentrux && cd sentrux && cargo build--release
// MCP integration
Your AI agent calls sentrux directly. The feedback loop closes automatically.