Real-Time Dynamic Bundling Agent Workflow Specification

Role: Lead E-commerce Merchandising Automation Engineer specializing in agentic systems and dynamic pricing.

Context

• Storefront platform: {{storefront_platform}}
• Target customer segment: {{target_customer_tier}}
• Bundle recommendation engine: {{bundle_rule_engine}}
• Minimum combined basket margin: {{min_basket_margin}}
• Maximum allowable agent latency: {{max_tool_invocation_latency_ms}}
• Human supervisor escalation threshold: {{human_override_threshold}}

Task

Produce an end-to-end technical workflow specification for an autonomous agent that intercepts customer basket events, calls recommendation and pricing tools, and publishes dynamically generated product bundles to {{storefront_platform}}.

Method

1. Define the event hook listener structure capturing cart state changes for {{target_customer_tier}} customers.
2. Construct the tool-calling signature for the candidate generator calling {{bundle_rule_engine}} with current basket IDs.
3. Specify the margin evaluation tool schema that calculates composite discount eligibility against {{min_basket_margin}}.
4. Build the dynamic bundle injection tool definition compatible with {{storefront_platform}} checkout APIs.
5. Map the sequential execution DAG ensuring strict evaluation of margin rules prior to basket injection.
6. Incorporate a timeout budget enforcement step ensuring total chain latency remains under {{max_tool_invocation_latency_ms}}.
7. Establish the risk rating heuristic that pauses autonomous publication and escalates to {{human_override_threshold}} when anomalies arise.

Constraints

• Dynamic bundle discounts MUST NOT breach the {{min_basket_margin}} boundary under any condition.
• The agent MUST terminate execution gracefully if API latency exceeds {{max_tool_invocation_latency_ms}}.
• Tool parameter declarations must strictly enforce integer prices and string SKU formats.
• No bundle may be injected without an automated inventory stock reservation check.

Output format

Provide the completed specification structured in these exact sections:

1. Pipeline Summary (1 paragraph, max 150 words)
2. Agent Tool Definitions (JSON schemas for 3 distinct tools)
3. Workflow State Machine (state transitions and decision branches)
4. Risk Control & Escalation Criteria (table format)

Self-review

• Confirm that latency constraints are enforced at each tool invocation step.
• Validate that human escalation boundaries are distinct, measurable, and complete.
• Check that all parameter schemas match standard JSON schema conventions.