Respond
Auto-block, alert, playbook execution
11 response actions all shipped: block_ip · kill_process · disable_account · isolate_file · block_tool · kill_agent · quarantine_session · revoke_skill · reduce_permissions · notify · log_only · Confidence-threshold policy: 90 / 70 / 0.
WHAT THIS LAYER DOES
L6 Respond closes the loop without a human. On a high-confidence detection, RespondAgent picks from 11 actions — block IP, kill process, quarantine skill, revoke credentials, isolate agent session, reduce permissions, send Slack / email / Discord alert, disable account, log-only, or run a custom playbook step.
WHY YOU NEED IT
Detection without response is noise. Agent attacks move in seconds — by the time a SOC analyst reads the alert, credentials are already exfiltrated. The loop has to close without a human for the 90th-percentile case.
HOW IT WORKS
respond-agent.ts dispatches ResponseAction union type. Each action handler is independently testable. Policy engine decides action by confidence threshold (>=90% auto, 70-90% notify + wait, <70% log). Playbook engine composes multi-step responses.
TRY IT NOW
Review default policy + customize per attack class:
pga config set policy.autoRespond 90ATTACKS THIS LAYER CATCHES
Concrete threats, concrete controls
Active agent compromise
CRITICALDetection confidence ≥90% → auto-kill agent session + revoke credentials + alert SOC.
Sustained attack IP
HIGHMultiple failed payloads from same source — auto-block at iptables / pfctl.