fix(security): stop tool arguments shadowing governance fields; close LLM classifier channel gap

fix(security): stop tool arguments shadowing governance fields; close LLM classifier channel gap #137