scoutlisted

# The tooling scout A session biased toward solving in-context will hand-author three failed attempts at a solved problem before anyone searches the ecosystem. This skill is the search — and the registry that makes each search's answer permanent. **Advise, do not implement.** The deliverable is candidates and an explanation: what each package does, how closely it fits the task shape, and what adopting it would displace. Doing the task with the chosen tool is the session's work, not this skill's. **Project facts come from `.claude/WSS.WORKFLOW.json`**: `WSS.record.toolbelt` is the registry, fallback `WSS.TOOLBELT.md`. An absent file is an empty registry, not an error. This skill is its **sole writer** — [`WSS.OWNERSHIP.md`](../../wss/workflow/WSS.OWNERSHIP.md). ## Job 1 — Consult, before any search Read the registry first. A row whose task shape matches the request answers it: name the package, point at its decision entry, stop. The search below is expensive and deliberate; the lookup costs nothing and must happen every time. ## Job 2 — Scout When the registry has no answer and the task is capability-shaped — convert, parse, render, diff, OCR, anything that smells like a solved problem — search the stack's public registries before anyone writes an implementation. Candidates are judged on fit to the task shape, maintenance signal, and how much they displace; present the top few with a recommendation, and say plainly when building in-context genuinely is the better call.