start

To restore context for an AI coding session, the start skill reads developer identity, git status, and active task signals from the repository to route to the correct workflow step. It classifies task status and identifies existing PRD, design, or implement artifacts.

Workflow phase routing classifies active task status as planning, in_progress, or none to direct the AI to the appropriate next step. You need it when beginning a new session or resuming paused work to ensure correct phase execution and artifact handling.

You initialize a Trellis coding session by running the start skill, which executes get_context.py to load identity, git status, and task signals. It then routes you to the appropriate planning, brainstorm, or execution step based on repository artifacts.

Yes, session initialization works by discovering relevant package and spec layers, then reading index files to identify existing PRD, design, and implement artifacts. This ensures the AI applies the correct gates and planning steps for the active task.

The best way to bootstrap project context is using session initialization to collect identity, git status, active tasks, and journal location. This loads repository state so the AI can continue coherently and route to the correct workflow phase.

Your AI assistant lacks project context because missing or outdated session information prevents coherent continuation. The start skill resolves this by loading identity, git status, and current task signals, then classifying the next action to restore accurate context.