"author:SkilledDragonCorp" matching MCP servers:

Scores proposed tool calls with a decision model before execution, returning allow, ask, or block decisions with safety, production, and prompt-injection probabilities so agents can gate risky actions and fail closed on scorer errors.