goblin-dolisted

# goblin-do > **Reference skill — requires a private "second brain" + the `goblin` CLI, both omitted here.** > Published as a **design reference**, not runnable as-is: it reads a per-project second-brain atom > store (`brain/suggestions/*.md`) that is **not included, for privacy**, and it is driven by the > **`goblin` CLI** (a separate project) that is likewise not part of this repo. To reconstruct a > working version, see [`_shared/goblin-second-brain.md`](../_shared/goblin-second-brain.md) — it > documents the atom shape and how to build your own second brain. The design below is unedited. Take ONE chosen goblin atom (a suggestion produced by `/goblin-suggest`, or a UAT task produced by `/goblin-suggest --uat`, persisted under `brain/suggestions/`) and act on it. `goblin do` is the single front door: it resolves the atom, reads its kind + mode, and **mode-dispatches**: - a **`small` suggestion** or a **safe `uat` task** → **EXECUTED**: goblin leases the existing workspace `/build-step` rail against the TARGET project inside a `goblin/do/<slug>` scratch branch, then either **auto-ships** a genuinely-clean result (fast-forwards it into the target's default branch) or **parks** anything borderline on the scratch branch with a one-line `goblin do --land <id>` to ship later. - a **`big` suggestion** or a **not-safe `uat` task** → **HANDED OFF**: prints the `/plan-feature` seed (or a UAT handoff block) + the build-rail next step. A big idea or a not-safe task needs