Ruleset Optimization
Optimizes Claude Code configuration files and rulesets using context efficiency principles to minimize baseline token usage while maximizing instruction effectiveness.
The source repository doesn't declare a license. Check its terms before reusing the code.
Key features
- Decision tree for extracting procedural content into domain-specific skills
- Deduplication guidelines for managing personal versus project-specific rulesets
- Token savings calculation formula to prioritize optimization efforts
- Progressive disclosure strategies to reduce baseline context by up to 70%
- Standardized templates for creating effective, non-overlapping skills
Use cases
- Reducing high token usage in large projects with complex instruction files
- Migrating monolithic CLAUDE.md files into modular, auto-activating skills
- Organizing project-specific guidelines to prevent conflict with personal defaults
FAQ
When should I use this skill?
You should invoke this skill when your project configuration files become too large, when you notice Claude is consuming too many baseline tokens, or when you need to organize instructions across multiple projects and languages more effectively.
Can this help manage personal vs. project-specific rules?
Yes. It includes a deduplication strategy to help you manage universal defaults in your global ~/.claude/ directory while allowing project-specific overrides in your local .claude/ folder without overlapping content.
How does it reduce token usage by 70%?
By applying the 'Progressive Disclosure' strategy, the skill identifies procedural 'how-to' content that isn't needed 100% of the time. Moving this content into auto-activating skills keeps the baseline context minimal while maintaining high instruction effectiveness.
What does the Ruleset Optimization skill do?
This skill provides a structured framework for optimizing Claude Code configuration files like CLAUDE.md and settings.json. It uses context efficiency principles to move procedural instructions into domain-specific skills, ensuring Claude only loads what is necessary for the current task.
What is the difference between a 'Policy' and a 'Skill'?
Policies (kept in CLAUDE.md) are 'What/Why' rules that always apply, such as security requirements. Skills are procedural 'How-To' guides for specific domains (like Git or Docker) that are only loaded when relevant, maximizing the signal-to-noise ratio.