Migration from header datasets or shared folders

Create one dedicated folder per report and move its .rmd plus needed data/style/media there. Keep original bytes and update paths relative to that new root. For each old data: {id: sales} block, copy that dataset’s inline mapping directly into the block, or replace it with the same csv/json relative file reference. Repeated consumers can reference the same local file. Remove frontmatter data only after every consumer has migrated. Preserve row order, field names, types, nulls and exact values; migration is an authoring change, not a transform. Regenerate into Rendered and compare the resulting values before replacing old output. The task-owned examples demonstrate this layout; historical artifacts and peer worktrees are unchanged.