Policy Algebra

Turns natural language policy statements into frozen Starlark governance blocks that you can drop into skills, plans, or rule files. It interviews you when you give it a vague intent string (deep mode) or extracts invariants from an existing document (shallow mode), drafts the rules, runs them through a review panel, and hands back a fenced block ready to paste. The verify flag does canonicalized drift detection between a frozen block and a candidate, useful for catching rule decay over time. It's self-hosted, meaning its own governance rules are encoded as a frozen Starlark block inside the skill file. Opinionated about keeping functions under five lines and grouping by subject, but the real value is making policy code reviewable and diffable instead of buried in prose.

Install to Claude Code

npx -y skills add camacho/ai-skills --skill policy-algebra --agent claude-code