Does it automatically modify my project files?

No. The tool generates a detailed report and requires manual user approval (approve, modify, or skip) for every proposed change to your rule files.

What is the 'Deterministic Collection + LLM Judgment' principle?

It uses reliable shell scripts to gather all factual data (skills and rules) and then uses an LLM to perform high-level reasoning and cross-reading of that data.

How does Rules Distiller decide what becomes a rule?

It follows a strict '2+ skills' requirement, ensuring principles are broad enough to serve as project-wide rules rather than being specific to a single task.

Can I use this with any coding agent?

While optimized for Claude Code and the ECC framework, its logic applies to any agent system using structured rule files like Cursor, Codex, or OpenCode.

Rules Distiller

Name: Rules Distiller
Author: affaan-m

byaffaan-m

•

172,007

•

Developer Tools

Scans and extracts cross-domain principles from installed skills to automate the creation and maintenance of project-wide rules.

Rules Distiller (rules-distill) is a sophisticated meta-utility for Claude Code that optimizes agent performance by consolidating scattered knowledge into actionable project rules. It identifies recurring patterns across multiple skills and uses LLM-driven analysis to propose additions, revisions, or new rule files. By promoting principles that appear in two or more skills to the global rule level, it ensures consistent agent behavior, reduces instruction redundancy, and maintains a clean, high-performance development harness with full user oversight.

Key Features

01LLM-powered verdict generation for appending or revising rules

02Automated deterministic scanning of skill and rule inventories

03Evidence-based reporting with violation risk assessments

04172,007 GitHub stars

05Cross-skill pattern recognition requiring evidence from 2+ sources

06Interactive approval workflow to prevent unauthorized file changes

Use Cases

01Auditing agent behavior to ensure coding standards match actual implementation patterns

02Consolidating repetitive instructions into a centralized project rulebook

03Performing periodic rule maintenance after installing new agent capabilities

Key Features

01LLM-powered verdict generation for appending or revising rules

02Automated deterministic scanning of skill and rule inventories

03Evidence-based reporting with violation risk assessments

04172,007 GitHub stars

05Cross-skill pattern recognition requiring evidence from 2+ sources

06Interactive approval workflow to prevent unauthorized file changes

Use Cases

01Auditing agent behavior to ensure coding standards match actual implementation patterns

02Consolidating repetitive instructions into a centralized project rulebook

03Performing periodic rule maintenance after installing new agent capabilities