Multi-Agent State and Guardrail Discovery Script

Role: Principal AI Architect and Discovery Specialist in multi-agent orchestration and dynamic runtime guardrails.

Context

• Target Organization: {{target_organization}}
• Agent Coordination Pattern: {{agent_coordination_pattern}}
• Integrated Tool Endpoints: {{third_party_tool_endpoints}}
• Human Escalation Threshold: {{human_in_the_loop_threshold}}
• Failure Recovery Policy: {{failure_recovery_policy}}
• Token and Compute Budget: {{token_budget_constraints}}

Task

Create a discovery session script to evaluate architectural readiness, guardrail parameters, and multi-agent state coordination challenges for {{target_organization}}.

Method

1. Frame the discovery objective around multi-agent reliability and runtime orchestration safety.
2. Draft architecture-probing questions on agent delegation under {{agent_coordination_pattern}}.
3. Script diagnostic inquiries around context sharing, memory persistence, and tool state synchronization.
4. Design boundary questions targeting unexpected tool responses from {{third_party_tool_endpoints}}.
5. Formulate deep-dive probes on fallback mechanisms and {{failure_recovery_policy}}.
6. Script qualification prompts on human handoff criteria using {{human_in_the_loop_threshold}}.
7. Draft cost-governance questions assessing execution loop controls under {{token_budget_constraints}}.
8. Build structured transition statements summarizing architectural gaps and scoping a joint architecture review.

Constraints

• MUST address state persistence, infinite loop prevention, and tool-call failure modes.
• MUST NOT provide generic AI advice or open-ended philosophical prompts.
• Script phrasing must reflect advanced multi-agent engineering patterns.
• Include exact verbatim questions and conversational guidance notes.

Output format

Produce a 4-part discovery script:

1. Architecture Alignment & Agenda Setter (150 words)
2. Agent Topology & State Synchronization Script (350 words)
3. Guardrails, Tool Failures & Escalation Script (350 words)
4. Compute Economics & Technical Next Steps (150 words)

Self-review

• Ensure all variables are directly operationalized in discovery questions.
• Confirm multi-agent failure modes and guardrails are thoroughly investigated.
• Validate that conversational prompts sound natural and technically credible.