RedHub AI
How the roll-up gate works
How the Tool-Input Schema gate works
This tool grades every tool your AI agent can call - graded per parameter against the guards its kind and consequence require. It reads 5 tools and rolls them into one verdict — the worst, never the average — then a gate that can only make it worse.
The 5 tools
A parameter whose consequence is money, identity, or destructive must declare every guard its kind requires. Miss one and it's OPEN DOOR (rank 2) instead of LOOSE (rank 1) - the same missing guard is judged harsher because of what the parameter can do, not how it's typed.
Why: 12 parameters across 5 tools, 7 fully typed. But 3 high-consequence parameters are missing a guard their consequence requires - issue_refund.amount (money, no cross_field_rule), delete_records.filter (destructive, no max), and lookup_customer.email (identity, no pattern) - and 2 more are only LOOSE. Guard coverage reads 58%, but coverage is context only: one open door on a money, identity, or destructive parameter is enough to make the whole registry UNGUARDED, worst-not-average.
No single tool is the crisis — the clustering is. The roll-up takes the worst, the gate escalates when trouble compounds, and it names the one thing to fix first. Grades the parameter contracts you declare; it calls no tool and executes nothing. Not legal or security advice.