You are a workflow orchestration expert specializing in AI a...
Prompt
You are a workflow orchestration expert specializing in AI agent systems. I am upgrading "AI Visual Director Production" v7.8.5 → v8.0.0. The current R0–R8 gate controller is STRICTLY LINEAR. Real production requires parallel paths, conditional skips, and feedback loops—but the current architecture forces sequential blocking even when independent work could proceed. CURRENT LINEAR GRAPH: R0(Scope) → R1(Brief) → R2(Concept) → R3(Reference/Taste) → R4(Asset Schema) → R5(Timing) → R6(Generation) → R7(Edit) → R8(Delivery) KNOWN PARALLELISM OPPORTUNITIES: - R3 (reference calibration) and R4 (asset schema) can develop concurrently after R2 - R5 (timing) can begin during R4 if audio assets exist - R6 failures of type "drift" should trigger targeted R4 revision, not full R2 restart - Motion graphics projects may skip R4_character_model_sheet entirely - Client approval sub-gates run parallel to technical QA YOUR TASK: Redesign the gate controller as a conditional DAG with explicit parallel execution semantics. Produce: 1. Complete DAG definition as adjacency list + node metadata (JSON) 2. For each gate node: dependencies, conditional skip rules, parallel siblings, feedback edges, required approvals 3. Conditional logic expressions: machine-evaluable predicates for skip/branch decisions (e.g., project_type, asset_availability, failure_type) 4. Feedback edge specifications: what triggers a backward edge, what state gets preserved vs reset, how cycles are prevented 5. Parallel execution protocol: how agents coordinate shared state when multiple gates run concurrently (conflict resolution, merge semantics) 6. Serialization format: how the DAG embeds in handoff packets so receiving agents can reconstruct execution state 7. Deadlock prevention: formal guarantee that no valid configuration creates unresolvable waits CONSTRAINTS: - Must remain a DAG (no uncontrolled cycles); feedback edges must have bounded iteration limits - Skip conditions must be evaluable from production-state alone (no external knowledge) - Parallel gates must not create race conditions on shared assets - The graph must be traversable by agents with no hardcoded gate logic—pure data-driven execution - Preserve all v7.8.5 compliance guarantees: no downstream polish bypassing upstream failure Do NOT ask clarifying questions. Define the complete DAG architecture. Output both human-readable diagram description AND machine-readable JSON.
