Optimize security-guard token usage with pre-run relevance gating, lower turn cap, and leaner prompt context#2113

lpcox merged 3 commits intomaingithub/gh-aw-firewall:mainfrom

copilot/optimize-claude-token-costsgithub/gh-aw-firewall:copilot/optimize-claude-token-costsCopy head branch name to clipboard