session-aware

Session state across multiple prompts is managed by tracking per-worktree progress in a local session.json file and enforcing explicit mode transitions. This prevents skipped steps and maintains context across different development activities.

Session-aware workflows coordinate multiple AI prompts by enforcing structured session types like vision, plan, and implement. They track progress and ensure proper handoffs between skills to maintain context throughout development phases.

Yes, isolated worktrees are required to store local per-worktree session state at .session-state/session.json. This setup isolates session tracking data and prevents state conflicts across concurrent development activities.

Yes, a YAML SKILL.md frontmatter with name and description is required to coordinate session types and wrap other skills. This metadata allows the orchestrator to properly delegate prompts and enforce mode transitions.

The best way to structure development phases is by using a session orchestrator that enforces transitions across vision, guardrails, plan, brainstorm, whiteboard, and implement modes, automatically announcing mode switches to maintain context.