state-management

Workflow-state management enables reading, writing, and updating a centralized state file that persists across waves, allowing PM agents to resume incomplete runs and track progress. Use readWorkflowState to load current state, then updateWorkflowState to merge changes after each wave completes.

Yes. State-management handles HITL stops by persisting the complete workflow state including wave progress and story completion flags. Call setHITL to mark the pause point, then readWorkflowState on resumption to restore execution context without data loss.

The Skill provides markStoryComplete and getCurrentWave functions to update state granularly. Each operation validates against schema, applies timestamps, and merges into the persistent state file, so your PM agent maintains an accurate audit trail of workflow progression.

The Skill enforces strict schema validation on read, write, and update operations with robust error handling. validateSchema confirms structure before commits, preventing malformed state that could break resumption or wave sequencing.

No. The Skill abstracts JSON read/parse/write operations, handling deserialization, validation, and timestamping internally. Call readWorkflowState or writeWorkflowState directly with project paths and receive normalized state objects.

updateWorkflowState merges partial updates into existing state without overwriting unrelated fields, preserving wave history and story records. This enables safe in-flight modifications for resumption and HITL workflows without data corruption.