21 agents
The canonical registry is designed around 21 specialized agents.
ProSystems: a skill-based operating architecture for multiple AI agents
ProSystems organizes specialized agents, reusable skills, business context and security boundaries as a maintainable AI workspace.
19 skills
A shared catalogue contains 19 authoritative skill entry points.
Security by design
Secret exclusion, permission boundaries and rollback are explicit parts of the architecture.
Why an agent OS?
As the number of agents grows, prompt-only systems become difficult to maintain. Roles, skills, tool permissions, business context and runtime rules need their own structure.
Skill-centric architecture
ProSystems separates agent identities from reusable capabilities. An agent-to-skill map defines which skills each agent can use, so capability logic can be maintained centrally.
Business context
Company knowledge lives in dedicated context areas instead of being mixed into universal instructions. Sensitive secrets are explicitly excluded from those files.
Security and rollback
Permission files are protected, structural changes are backed up, and runtime precedence rules prevent stale notes from overriding current deployment information.
Transparent runtime status
The file architecture and skill catalogue have been implemented and verified. Runtime activation for persisted agents is kept separate and requires supported authorized mechanisms.
Define the next step
Use the configurator or one of the free tools to define a controlled starting point for a pilot.
Start AI Configurator