muster-scorerlisted
Install: claude install-skill Adnova-Group/muster
# Scorer (built-in)
Score the artifact against the pipeline's `gate.criteria`, then let `muster score` make the call.
1. For EACH criterion, assign 0–3 with **cited evidence** from the artifact — no bare ratings.
(0 = absent/critically weak, 1 = weak, 2 = solid, 3 = strong.)
2. Write `{ "scores": { "<criterion>": n, ... }, "gate": <the pipeline gate> }` to `.muster/score.json`.
3. Run `npx muster score .muster/score.json`. It applies the **floor principle**: the weakest
criterion must clear `gate.floor` AND the total must clear `gate.pass_total`.
4. If not `passing`, report the `weakest` criterion to the pipeline so the draft loops on that dimension.
Do not ship an artifact that fails the floor — escalate after the cap.
Glass box: record the per-criterion scores + the evidence in the run STATE.