Domain capability, always authorised.
Skills give an authorised agent the domain capability to do real work, and every skill call runs through the Authorisation Layer before it executes. Capability and control ship together.
Banking, legal, project management and fraud agents ship with skills tuned to their field. Each skill is a bounded capability the agent can call as part of a plan.
A skill is not a bypass. When an agent invokes a skill, the action is intercepted and evaluated against your rules, exactly like any other action.
New skills can be added for your own workflows, and they inherit the same authorisation path, so extending an agent never means loosening control.
Related capabilities on the same authorisation layer. Every one is enforced before an action runs and sealed to the Provenance Vault.
Put one workflow behind Xybern and watch every agent action authorised, and sealed to the vault.
Xybern