[VibeDefend] Agent Access Map & Guardrail Coverage

An AI coding agent can reach almost everything its user can: repositories and files, secrets and environment variables, shell commands, network access, cloud credentials, MCP servers and skills. VibeDefend models this full access surface for each agent and overlays your active guardrails (security rules, Action Guard policies, MCP Security Proxy) to show exactly what is covered and what is not. Uncovered areas are highlighted and prioritized by risk, with recommendations to close each gap.