session-state-snapshot-reuse

To save and restore conversation state, serialize messages to per-line JSON in JSONL format and write atomically using a .tmp file and rename operation to prevent corruption. Restore by reading the JSONL, parsing each line into Message objects, and applying them with context.setMessages.

Conversation checkpointing is best implemented using a Snapshot → Store → Restore workflow that serializes messages to per-line JSON and uses atomic write discipline with .tmp files and renames. This ensures session state remains uncorrupted during persistence and enables deterministic state restoration.

Atomic write discipline matters for session state persistence because it prevents file corruption by writing to a .tmp file first and renaming it only after successful serialization. This guarantees the JSONL snapshot remains intact even if the process crashes during the write operation.

Yes, you can use session-scoped project directories to store conversation snapshots. This Skill employs portable paths to keep per-session state co-located with the working project, ensuring saved JSONL snapshots remain accessible and organized for subsequent state restoration.

To restore a prior session state from saved JSONL messages, read the JSONL file, parse each line into Message objects, and apply them using context.setMessages. This replaces the current UI state and deterministically reconstructs the conversation from the persisted snapshot.

You need JSONL format for message snapshot persistence when building features that require conversation checkpointing, state restoration, or message manipulation. JSONL provides per-line serialization that allows deterministic saving and loading of individual session state records without parsing an entire JSON document.