ce-decidelisted

# Decide **Invocation input:** Decision to weigh (and optional `--evidence` mode / a source artifact path): $ARGUMENTS Take **one decision** and **two or more candidate technical options**, and produce an **opinionated, evidence-tagged recommendation**: score each option against a fixed seven-axis engineering rubric with anchored 1-10 bands, **derive the axis weights from the stated situation and show the derivation**, run each option through knockout gates and a falsifiable kill-condition, aggregate by **gate-then-weight** (never a bare sum), and render an **Adopt / Adopt-with-mitigations / Spike-first / Reject** recommendation naming one option — then draft a **proposed ADR** the human promotes. > **A verdict tool — on the engineering side of the line.** The framework keeps > *product, security, and scope* verdicts with the human (`/core-engineering:ce-plan` decomposes, `/core-engineering:ce-brief` > elicits, `/product-discovery:ce-market-scan` frames but never decides; `/product-discovery:ce-idea-score` is the one *product* > verdict exception — both `/product-discovery:ce-market-scan` and `/product-discovery:ce-idea-score` now ship in the companion > `product-discovery` plugin). Choosing among **technical options is engineering judgment the > agent already exercises** — `/core-engineering:ce-spec` ends a material-decision prompt with > `Recommendation: <A/B>`, `/core-engineering:ce-plan` runs a Sizing Gate. This skill makes that judgment > *rigorous and standalone*. I