This commit adds a comprehensive data-driven action rules engine, enhancing system safety with per-category confidence thresholds and trust-score modifiers. We've introduced a Blast Radius Analyzer that uses static service configurations to override LLM guesses, along with retry limit guardrails to prevent recursive alert loops. These changes harden the worker pipeline against unpredictable behavior, ensuring safer automated remediation.