rules-distill

Extract shared principles across installed skills by running a workflow that inventories all skill contents, clusters them by theme, and applies cross-reading analysis to identify overlapping patterns. This generates candidate rules with clear verdicts.

Distill cross-skill insights into concrete rules by performing deterministic collection of skill contents via scripts, then relying on LLM judgment to generate verdicts. The output is structured rule files with actionable policy changes.

Cross-skill analysis for rule management is the process of exhaustively inventorying installed skills to identify cross-cutting principles, reducing duplication and drift across the skill set. It clusters skills by themes to propose rule changes.

Yes, this rule distillation workflow requires jq as a dependency to ensure deterministic collection and parsing of skill contents. The LLM judgment handles the verdicts, while jq structures the data extraction.

Cross-skill analysis generates policy automation verdicts such as Append, Revise, New File, or New Section. These verdicts maintain a traceable audit trail for proposed rule changes extracted from shared principles.

You should not use LLM judgment for rule distillation if your installed skills lack overlapping themes or if deterministic collection via scripts fails to parse your skill contents. The process relies on finding cross-cutting principles.