Build Agents
Agents should be configured for clear responsibilities and controlled execution.
What this means
Agent configuration is part of governance design. Each agent must have explicit purpose and bounded authority.
Why it matters
Without role boundaries, workflows become hard to audit and unsafe to scale.
How Subbasis handles it
Agent setup combines:
- role definition
- skill pack attachment
- permission scope
- policy guardrails
- model resource preferences
- approval dependencies
Agent build checklist
- Role and task boundaries
- Skill Pack assignment
- Model preferences
- Permission scope
- Approval dependencies
Example scenario
Agency operations:
- Research Agent gathers account intelligence.
- Outreach Agent drafts communication.
- Compliance Agent validates policy before send.
- Human gate approves external message when required.
What to configure
- allowed connectors per role
- context visibility by role
- fallback behavior if action denied
- evidence fields required by role
Limits and deployment notes
- Agent count and skill profile limits depend on plan.
- Advanced orchestration controls may vary by deployment.
Related docs
Continue with Model resources.