Agent capability is scaling faster than trust boundaries.
AgentVerus Scanner v0.7.1 expands high-risk workflow coverage and tightens severity scoring for undeclared browser, network, file, and session behavior.
If you build agents, MCP servers, or skills, I want blunt feedback.
AI agents now install packages, read files, and touch real systems. Trust tooling is behind.
AgentVerus Scanner v0.6.2 improves script detection, capability-contract checks, auth/browser risk inference, and SBOM output.
If you build agents or skills, I want blunt feedback.