trellis-continue

Run the trellis-continue skill, which calls get_context.py to load the current task and phase index, then checks task_progress.py for saved progress. It routes you to the correct workflow step based on task status and which artifacts exist.

It reads the task status field and checks for artifacts like prd.md, design.md, and implement.md. For example, status=planning with only prd.md routes to step 1.4 for lightweight tasks, while status=in_progress with implementation done routes to the check step 2.2.

Completed tasks are not resumed into Phase 2 or 3.3. Instead the flow enters the trellis-push completed-task preflight, which prepares publication recovery, points to trellis-finish-work, or blocks on ambiguous evidence.

Yes, after explicit user confirmation, run task_progress.py reopen with the task name. Only completed to in_progress transitions are valid; reopening clears completedAt but preserves the auditable progress record.

Saved progress is treated as advisory evidence only and never overrides task status, planning artifacts, workflow ordering, or Git publication evidence. This prevents resuming from stale or ambiguous state after compaction or session changes.