sag-setuplisted

# sag-setup — first-run setup (manual trigger only) This skill **does not load itself**. It runs when the user types `/sag-setup` or gives a direct order to set the plugin up. It logs in, can create a SAG source, and writes config files — do not infer permission for any of that from an unrelated request. ## Ask for these, in this order Stop and ask. Do not guess a URL, invent a source name, or pick include globs by reading the repo without confirmation. | # | Ask | Needed when | Example | |---|---|---|---| | 1 | **SAG URL** | always | `http://sag.home` | | 2 | **Login name** | always | `vuong` | | 3 | **Is there already a scope to join?** | always | see below | | 4 | **`source_id`** | JOIN only | `a3f1…` or the name to look up | | 5 | **Source name** | BOOTSTRAP only | `project-a-knowledge` | | 6 | **State service URL + token** | only if agents run on >1 machine | `http://state.home:9000` | Question 3 decides everything after it. Phrase it concretely: *"Has any other agent or machine already published knowledge for this project into SAG — do you have a `source_id`?"* - **They have a `source_id`** → JOIN. Nothing is created. The scope already has documents, a manifest, and possibly a shared state service. - **They do not** → BOOTSTRAP. This machine provisions the scope. Everything else is derived — never ask for it: - `key_format` comes from probing the instance (below), not from a preference. - `source_id` in JOIN mode comes from the user; in BOOTSTRAP from `source