rad-repo-scaffold

Rad Repo Scaffold

Create only the structure justified by the project's purpose, technology, workflows, and selected AI tools. Preserve existing work.

Workflow

1. Accept a direct purpose description or one or more .md requirement files and read them completely.
2. For an existing repository, inspect its structure and instructions first. Treat open IDE buffers as authoritative when IDE-aware tools are available.
3. Extract purpose, deliverables, maturity, technologies, package layout, selected AI assistants, documentation and automation needs, source/test layout, constraints, and prohibited changes.
4. Ask only questions whose answers materially change the structure. Do not ask for discoverable information.
5. Read references/structure-catalog.md; also read references/memory-pattern.md when a memory workflow is in scope. Propose only justified paths and give a one-line reason for every top-level entry.
6. Separate paths into required now, optional later, and excluded. Explain exclusions for tempting but unnecessary AI-tool folders. Note, in the report only, any companion AI tool from the catalog's "Companion AI tools" list that plausibly fits the project's scale — never install or configure one without explicit confirmation.
7. Confirm before creating a large structure, modifying existing files, or adding unrequested tool-specific configuration.
8. Save the approved structure as JSON. Run python scripts/apply_structure.py --root <repo> --plan <plan.json>, review its dry run, then rerun with --apply. Prerequisite: this step needs a Python 3 interpreter on PATH — it is the only part of this kit that does; if Python isn't available, create the approved folders/files manually (or via PowerShell) following the same plan JSON instead of skipping the scaffold.
9. Verify and report created, skipped, and conflicting paths.

Plan Format

{
  "directories": ["src", "tests", ".agents/skills"],
  "files": {"AGENTS.md": "# Repository instructions\n"}
}