Agent-Skill-Kit

Ask Conceptual Integrity Sentinel

A Principal-level engineering agent that audits repositories for architectural drift, bloated abstractions, and “dead code.” It enforces Agentic Engineering protocols: Assumption Surfacing, Confusion Management, and Simplicity First.

Purpose

To fight “Code Slop”—the tendency of AI models to over-engineer, over-abstract, and leave dead code behind. This skill helps maintain conceptual integrity by identifying and eliminating unnecessary complexity.

Usage

Activate the sentinel with phrases like:

• “Study this repository and provide a detailed list of improvements.”
• “Audit the auth module for complexity bloat.”
• “Why is the codebase becoming fragile?”

Features

• Complexity Bloat Detection: Detects files where abstraction cost outweighs utility.
• Dead Paths Identification: Identifies exported functions/components that have zero consumers.
• Assumption Surfacing: Prompts the agent to explicitly list architectural assumptions before analyzing.

Tools

• verify_complexity_bloat: Detects files where abstraction cost outweighs utility.
• detect_dead_paths: Identifies exported functions/components that have zero consumers.
• surface_assumptions: Prompts the agent to explicitly list architectural assumptions.

Best Practices

• Run the verify_complexity_bloat tool regularly to catch bloat early.
• Perform detect_dead_paths scans before major refactors.
• Always review the SENTINEL_REPORT.md for critical insights.