<!-- https://staircase.co/agents/roster/assistant-builder -->
# Assistant builder

# Assistant builder

Creating new agents

<!--source:agent-kits-->
Designs and implements task-triggered agents on the established runtime pattern — one function receiving events from the work-management surface, conversation state held externally, a model doing the reasoning, and tracing on every turn.
<!--/source-->

- **Used when**
<!--source:agent-kits-->
— When a new agent should be triggered by a task or a comment rather than run by hand, which is the shape behind most of the deployed fleet.
<!--/source-->

- **Loads first**
<!--source:agent-kits-->
- The agent-building skill, for the runtime, state and telemetry patterns.
<!--/source-->
<!--source:agent-kits-->
- The shared engineering guidelines.
<!--/source-->

- **Returns**
<!--source:agent-kits-->
- The architecture and its runtime boundaries — one function, its handlers, the external state layout.
<!--/source-->
<!--source:agent-kits-->
- An implementation checklist covering ingress wiring, state, the webhook construct, memory, the model and its tools, and tracing.
<!--/source-->
<!--source:agent-kits-->
- Deployment and configuration requirements, including the exact values a human must collect and who owns the webhook per environment.
<!--/source-->
<!--source:agent-kits-->
- End-to-end verification steps covering the task flow, the traces, the locking and the persistence of conversation memory.
<!--/source-->

- **Hands off**
- Operations agent builder — operations agents for document and media pipelines
- Spend approval — approval workflows that gate model spend
- Audit analyser — analysers that scan processing output for anomalies

## Others in Creating new agents

- Audit analyser
- Operations agent builder
- Spend approval
