session-end

To save project state before ending a session, this skill captures all current work details and overwrites them into a single canonical PROJECT_STATE.md file, ensuring a clear resume point and preventing data loss across context resets.

The best way to persist workflow state is by overwriting a canonical PROJECT_STATE.md file, which guarantees a single source of truth for iterative experiments and long-running research sessions without ambiguity.

State persistence overwrites the save file to prevent ambiguity in subsequent sessions. Appending multiple entries would create conflicting or duplicated records, so overwriting guarantees a single canonical source of truth for project state.

Yes, you can use this session management approach for multi-agent coordination. It preserves work state across sessions, which is critical when multiple agents need state continuity to resume iterative tasks without losing context.

You need to save project state to a markdown file whenever you end a session or run a /clear context reset. This ensures the next session can resume without ambiguity, especially when multiple steps or datasets exist.