verdictlisted

Architect-grade technology decision briefs — "which option should we pick, and why?" Takes a decision question (X vs Y vs Z for a stated context: library, database, framework, platform, buy-vs-build) plus your constraints, fans out parallel sub-agents to research each option's project health, community sentiment, technical claims, and operational story, then scores options across weighted criteria and emits an ADR-style recommendation with a full scorecard and cited evidence. Keyless; the recommendation is advisory — you decide. Use when the user must choose between technologies or approaches — e.g. "/verdict postgres vs mongodb for <use case>", "should we use Kafka or NATS", "compare X and Y", "buy vs build", "which framework for this". Also supports "/verdict revisit <brief>" to cheaply re-check a past verdict's revisit-when triggers. For community buzz on a single topic use pulse; for exhaustive fact-checked research use deep-research.
duthaho/skillhub · ★ 6 · API & Backend · score 75

Install: claude install-skill duthaho/skillhub

# verdict — technology decision briefs `/verdict <option A> vs <option B> [vs <option C>] for <context> [constraints]` `/verdict revisit <path-to-previous-brief>` — re-check an old verdict (see Revisit mode) Answer one question: **given this context, which option should we pick, and what evidence supports that?** You research every option the same way, score against explicit weighted criteria, and write the decision up so it can be defended in a design review — or revisited later when the facts change. ## Step 0 — Frame the decision Extract from the request (ask **1–2 questions only if** genuinely blocking): - **Options** (2–4). If the user names only one ("should we use X?"), add the status-quo/do-nothing option and the 1–2 strongest alternatives — say you did. - **Decision context:** what is being built, scale (users/QPS/data volume), team size and existing skills, current stack, timeline. - **Constraints & dealbreakers:** licensing limits, budget, compliance, must-run-on-prem, language lock-in, etc. - **Criteria weights** — defaults below; honor plain-language overrides ("ops burden matters most", "we optimize for hiring"). State every assumption you make in the brief header. A verdict against the wrong context is worse than no verdict. ## Step 1 — Fan out: parallel research sub-agents (per option) Spawn one research sub-agent **per option**, concurrently (one message, multiple `Agent` calls, `subagent_type: "Explore"`). Each covers five angles and returns