Host · Permissions
Risk Management — Risk Level × Tool Category
LVIS's permission model is a grid where two axes intersect. The tool's risk level (low/medium/high) and its category (read/write/execute/network) meet to determine whether it runs automatically, shows an inline confirmation, or opens a dialog.
3 risk levels
5 tool categories
additional-consent chain
Decision grid
| Risk \ Category | Read | Write | Execute | Network |
|---|---|---|---|---|
| Low | Auto | Auto | Card confirm | Auto |
| Medium | Auto | Card confirm | Dialog | Card confirm |
| High | Card confirm | Dialog | Dialog + additional consent | Dialog |
In strict mode, both medium and high are escalated to a dialog.
Additional-consent chain
High-risk actions preserve user consent as an immutable record chain. You can later verify exactly who consented, when, and with what scope — preventing automation from happening "quietly."