Create Rule

This is meta-tooling for when you catch yourself or an implementation agent making the same mistake twice. The idea is simple: if you're saying "you're absolutely right, I should have done it differently," that's your signal to codify the fix as a rule. The skill guides you through creating contrastive examples (incorrect vs correct patterns) that load into every session as behavioral guardrails. It emphasizes token efficiency and path scoping so you're not loading frontend rules when editing backend code. The approach is grounded in research showing that showing both the wrong and right pattern eliminates ambiguity better than positive examples alone. Useful when you notice patterns in agent mistakes rather than one-off issues.

skills.sh installs: 1.3K
GitHub stars: 1.8K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add neolabhq/context-engineering-kit --skill create-rule --agent claude-code

Installs into .claude/skills of the current project.

Files

Create Rule

Guide for creating effective .claude/rules files with contrastive examples that improve agent accuracy.

Overview

Core principle: Effective rules use contrastive examples (Incorrect vs Correct) to eliminate ambiguity.

REQUIRED BACKGROUND: Rules are behavioral guardrails, that load into every session and shapes how agents behave across all tasks. Skills load on-demand. If guidance is task-specific, create a skill instead.

About Rules

Rules are modular, always-loaded instructions placed in .claude/rules/ that enforce consistent behavior. They act as "standing orders" — every agent session inherits them automatically.