Sequential Agent Image Pipeline Script for Style Consistency

Role: Lead Autonomous Workflow & Generative Media Engineer

Context

• Workflow Engine: {{agent_framework}}
• Visual State Store: {{state_storage_backend}}
• Fixed Art Direction Anchor: {{art_direction_rules}}
• Consistency Mechanism: {{consistency_technique}}
• Throughput Constraint: {{concurrency_limit}}
• Target Asset Consumer: {{downstream_consumer}}

Task

Write a complete pipeline orchestration script that coordinates sequential tool calls to generate visually unified image sequences, preserving stylistic tokens and character/scene continuity across multiple agent reasoning steps.

Method

1. Establish the workflow state schema containing scene indices, anchor descriptors, and asset manifests stored in {{state_storage_backend}}.
2. Build a Prompt Manifest Engine that combines {{art_direction_rules}} with dynamic scene-level agent actions.
3. Implement the {{consistency_technique}} handler to manage image references, seed propagation, or ControlNet inputs.
4. Define step-level graph nodes or workflow activities inside {{agent_framework}} for generating, validating, and committing assets.
5. Implement a concurrency limiter honoring {{concurrency_limit}} using asynchronous worker pools or semaphore locks.
6. Format generated asset metadata into the payload structure required by {{downstream_consumer}}.
7. Build a fallback node that detects style divergence and automatically triggers an in-line correction re-prompt.

Constraints

• The script MUST implement complete state persistence logic without omitting backend calls.
• MUST NOT allow style tokens in {{art_direction_rules}} to be overwritten by transient scene descriptions.
• Error handling MUST catch network timeouts and failed generation states without dropping pipeline memory.
• The workflow MUST run cleanly with async/await patterns in Python or TypeScript.

Output format

Provide the complete implementation script structured in these exact sections:

1. State Schema & Persistent Storage Connectors.
2. Prompt Composition and Consistency Anchor Layer.
3. Core Pipeline Nodes and Execution Graph ({{agent_framework}}).
4. Downstream Payload Formatter ({{downstream_consumer}}).
5. Sample Execution Fixture showing a 3-scene sequential generation run.

Self-review

• Check that {{art_direction_rules}} persists across every sequential prompt generation step.
• Verify that {{concurrency_limit}} is strictly enforced in the asynchronous dispatch loops.
• Confirm {{state_storage_backend}} correctly stores image hashes and prompt states.