CSL MCP Server – Write and Verify AI Safety Policies from Claude/Cursor
Mathematically verified policies enforced outside the model—formal proof replaces prompt engineering.

Encodes patio11's dispute tactics into a reusable Claude Code agent workflow.
Consumers dealing with billing disputes, insurance claims, or bank errors
Claude Code · Cursor Rules
This is a community implementation. No affiliation with patio11, just a fan of his work.
Repo: https://github.com/Tetra-Research/dangerous-professional-plu...
Install: `npx skills add Tetra-Research/dangerous-professional-plugin`
Mathematically verified policies enforced outside the model—formal proof replaces prompt engineering.
Trello-to-PR pipeline with Claude Code, but no clear edge over existing AI coding tools.
TLA+ verification caught production bugs that years of testing missed.
Docker sandbox for Claude Code's dangerous flag when Anthropic won't let you run it bare.
Session types make dangerous actions code-unreachable, not just policy-blocked.
Polished UI for a problem Gamma, Beautiful.ai, and Canva already own.