Use the self-learning loop to improve orbit rounds over time. Records findings, analyzes domain risk weights, detects regressions, and recommends attack priorities for the next round. Use between orbit rounds.
Use this skill between orbit rounds to make each subsequent round smarter. The loop learns which domains produce the most bugs, which fixes regress, and where to focus attack effort.
After R52 completes, record all findings with domain, severity, and fix status.
Query all prior findings. The report includes:
Use the recommended priorities to configure the queen agent's task dispatch. High-priority domains get more attack agents.
Use vector search to check if a new finding matches a previously-fixed issue — this is a regression flag.