# PROCESS: SYNTHESIS CYCLE (Data-Centric, AI Proposes & User Confirms) ## Process Steps: 1. AI: Verify Schema access. Access Project State & invoking Task Instance. 2. AI: Identify input data within `project_state` for synthesis based on user instruction. Access current Framework State. 3. AI: Perform synthesis and analysis of inputs. 4. AI: Propose specific updates to the draft Framework State object. 5. AI: (Optional) Internal review of proposed updates. 6. AI: Formulate yes/no questions to confirm key synthesis findings and proposed Framework State updates. 7. AI: Present summary of proposed updates and confirmation questions to user. 8. User: Responds Yes/No or provides freeform corrections. 9. AI: If revisions needed, return to step 3. 10. AI: If confirmed, update the `framework_state_data_object` in `project_state.knowledge_artifacts`. Increment version. Update metadata timestamp. 11. AI: Update invoking `task_execution_instance` status to 'Completed', store summary/provenance in `output_data`. 12. AI: Present completion statement. Signal completion to `03-EXECUTE-Task`. ## Example Structured Questions (AI to formulate): * "Synthesis of Tasks [ID1, ID2] suggests [Key Finding]. Does this seem accurate? (Yes/No)" * "Based on this synthesis, I propose updating the definition of Concept '[Concept Name]' in the Framework State to '[New Definition Snippet]'. Is this acceptable? (Yes/No)" * **[IF a proposal was rejected]:** "Understood. Please clarify how the synthesis or proposed Framework State update should be revised." (Freeform input expected) * **[IF completed]:** "Synthesis cycle complete. Project State (Framework State v[new version]) updated." ---