← ClaudeAtlas

scaffolding-submoduleslisted

Scaffold a brand-new Core or AWS submodule with no docs/reference page or src dir yet: module folder, failing placeholder test, barrel re-export, docs stub — hands off to implementing-submodules. Use for "add"/"create"/"scaffold" a new library module, even named casually ("a cache module"). Already has a spec page → implementing-submodules directly.
monte3l/m3l-automation · ★ 0 · AI & Automation · score 56
Install: claude install-skill monte3l/m3l-automation
# scaffolding-submodules Scaffold a new submodule under the library's `core` or `aws` namespace. This is the _greenfield_ entry point for a **net-new** module — one beyond the bootstrap catalog that has no `docs/reference` page and no `src/` directory yet. It creates the seams (folder, failing test, barrel re-export, doc stub) and updates the status tracker, then hands the actual implementation to `implementing-submodules`. The package `exports` map stays at its three namespace entries (`.`, `./core`, `./aws`) — new submodules are surfaced through the namespace barrel, **never** a new subpath entry (that would be a semver event; see `docs/contributing/contributing.md`). The sole exception is ADR-0004's dated Update (`./core/errors`, admitted only for a proven browser-safe leaf a browser-target package needed a real value import from) — it is not a precedent for scaffolding a new submodule's own subpath. The skill's name is about adding a new module, not a new `exports` subpath — the two are deliberately kept separate. ## Role boundaries (hub-and-spoke) This skill runs in the **hub** (main agent) and only lays down scaffolding + tracker rows. It does **not** implement the module or review code — that happens in the spoke subagents (`test-author`, `code-implementer`, `code-reviewer`, `spec-conformance-reviewer`) orchestrated by `implementing-submodules`. The agent that writes the implementation is never the one that reviews it; keep that separation intact here by handing off