rules-distill

To extract common rules from installed skill files, this Skill uses deterministic scripts to inventory skills and an LLM to cross-read clusters, proposing actionable rule candidates. It centralizes recurring practices scattered across individual files into formal rule proposals.

Cross-cutting rule distillation is the process of scanning installed skill files to identify recurring principles and converting them into actionable rule candidates. It formalizes ad-hoc knowledge by promoting shared patterns to repository-level rules.

Yes, you need jq installed as a dependency to run the deterministic inventory scripts that enumerate installed skills and rules files. The scripts rely on jq to produce the complete factual index required for the LLM cross-read judgment.

You should run a rules maintenance scan after installing new skills, during periodic monthly maintenance, or when audits reveal recurring patterns across multiple skills. This surfaces principles that should be promoted to repository-level rules.

No, the LLM does not automatically apply changes to rules files. The Skill presents verdicts like Append, Revise, or New File as candidates and requires explicit user approval before any rule modifications are executed.

The best way to audit recurring patterns is using deterministic scripts to inventory skills and rules, followed by an LLM cross-read to produce verdicts. This approach ensures a complete factual index before proposing any rule modifications.