rules-distill

Extracting core principles from installed skills involves running automated scripts to inventory all skills and existing rules, then using LLM judgment for cross-skill pattern recognition to distill recurring principles into formal, structured rules for consistent application and policy maintenance.

Automating rules maintenance uses deterministic bash scripts to gather skill and rule data comprehensively, then applies LLM analysis to identify shared principles across multiple skills, suggesting new rules or edits to consolidate best practices and streamline ongoing policy updates.

Yes, you need jq, node, and bash installed to run the deterministic data collection scripts that gather skill inventory and existing rule files, which are then processed for cross-analysis and rule generation.

Start consolidating best practices by running the scan-skills.sh and scan-rules.sh scripts to collect inventory data, then review the generated suggestions for possible rule updates and formal principle extraction.

Cross-skill pattern recognition analyzes installed skills to identify recurring principles shared across multiple skills, using LLM judgment to find common patterns and distill them into formal, structured rules for consistent content analysis and policy improvement.

The process relies on deterministic scripts for data collection but uses LLM judgment for principle extraction, meaning rule generation quality depends on the LLM's analysis of cross-skill patterns rather than fully deterministic rule consolidation.