sosa-governor

You can run a governance audit by triggering a daily SOSA audit that summarizes JSONL audit logs, token budgets, and trust scores. The report provides totals, impact-level breakdowns, denied actions, and identifies sessions with the most activity.

Tool trust score management reads and displays trust scores from the impact registry, explains the trust formula, and resets or updates scores to control auto-approval behavior. This ensures compliant tool usage by maintaining accurate trust classifications for autonomous agents.

The best way to monitor token budgets is comparing configured budgets against actual usage to flag categories approaching limits. This token budgeting analysis runs during standard governance audits and identifies untracked token spend across autonomous agent sessions.

Yes, you can scan for hardcoded secrets using grep-based security scans that check JSONL audit files and JSON config files. The compliance check also verifies gating hooks, plan-act-verify loops, and evaluates overall SOSA compliance levels for suspicious patterns.

To reclassify tool impact levels, access the impact registry to view high, medium, and low classifications, then apply safe reclassification with immediate effect. This updates tool trust states and controls auto-approval behavior for subsequent autonomous agent actions.

Compliance reports show denied actions and anomalous behavior when the governance audit detects high-impact tool calls, misclassified tool trust, or suspicious patterns in cross-session audit logs. These findings indicate potential security incidents or untracked token spend requiring registry updates.