Capability design
Better skills can improve triage quality without changing tools; better connections can improve evidence without changing the model.
Product
Agent Operating Model
An OpsRabbit agent behaves like an operator with three capability layers: skills shape behavior, tools perform actions, and connections provide authenticated external access. Runtime policy, approvals, and sandboxing decide what the agent may actually execute.
What this page covers
Use skills for playbooks, domain logic, and procedural behavior when output quality is inconsistent.
Use tools and workflows for executable operations, API calls, deterministic automation, and status checks.
Use connections for scoped access to cloud, Kubernetes, SSH, code, data, monitoring, and ITSM systems.
Operational detail
How OpsRabbit supports Agent Operating Model
An OpsRabbit agent behaves like an operator with three capability layers: skills shape behavior, tools perform actions, and connections provide authenticated external access. Runtime policy, approvals, and sandboxing decide what the agent may actually execute.
Runtime boundaries
The agent can answer product usage, configuration, troubleshooting, and supported operational questions while sandbox and tool policy constrain execution.
Feature mapping
Providers influence reasoning, inbounds start work, schedules trigger recurring runs, workflows add determinism, and sandboxing constrains command risk.
Related pages