Support Declarative Agents

[BechtelCanDoIt]

We need to add some kind of Declarative Agent Runner.

This runner would provide the OTel functionality. This would make it so that non-developer uses could add in agents easily.

Currently Supported:
❌ Declarative agents are closer to AI-native thinking such as Agent Flavored Markdown. You're specifying intent and constraints, not procedure. The LLM figures out the how. This is the direction Anthropic is pushing — Claude's behavior emerges from instructions, not from hardcoded logic. This is the YAML/Markdown structure that Claude Code, and other agent harnesses, can execute.

These type of Agent definitions are especially relevant for cron type of system for reporting, dash boarding, and anything else where you want something to run and then stop in a scheduled way.

✅ Imperative agents are closer to AI-enabled thinking in disguise. You've designed the workflow. The LLM fills roles within it. You're still the architect of the control flow — the LLM is a smart function call.

	Declarative (YAML/Markdown)	Imperative (Code)
Who controls flow?	The LLM	The developer
Autonomy	Higher	Lower
Predictability	Lower	Higher
AI-Native alignment	Higher	Lower
Enterprise comfort	Lower	Higher

[jingchang0623-crypto]

$BODY