[FEATURE]: Agent Interface and Events Bounds #575

Summary

Add generic AURA specific bounds to an agent trait. Trait should be its own bounded context and have the following:

GOALS:

• Export only (outbound) event schema as an enum for events the agent can produce such as completion chunks, tool calls, errors, MCP progress bounds, HITL signals, OTEL spans and others (Rig requires us to have OTEL spans as their own outbound events for now - in an ideal world more generic rich events could be emitted without being OTEL specific and easily translate to Otel adapters as a consumer).
• CRUD api for driving agent creation and evaluating prompts + history.
• Consider a light hooks implementation for things that plug into the agent loop like the turn depth nudge (make it easier to add such features going forward)
• Continue to support cancellation hooks at the interface layer that propagate down to MCP driven actions
• Clear delineation from AgentConfig (structures hydrated from toml) and Agent (actual running process that claims a session id)

STRETCH GOALS:

• Mock agent to simplify some tests and reduce our token burden in CI

NON GOALS:

• Avoid trifurcation of single agent, coordinator agent, and workers. The agent layer should drive a common interface - workers/plan Dag etc are just tools

Additional Context

No response

Upload screenshots

No response

Searched Issues

• No similar issues found

Code of Conduct

• I agree to follow this project's Code of Conduct