Trust Score for AI-generated code, as a GitHub Action.
umbra fix— applies provably-safe fixes (score before → after)umbra setup— one-word installer (pre-commit gate + this Action + agent guard hooks)- Baseline mode — adopt in existing repos, gate only on new findings
- Live badges — set
publish: trueto report scores to the hosted badge service - Agent-config threat linting — prompt-injection and MCP-config rules
Consumers: uses: elberacasa/umbra@v1. Full docs in the README.