← ClaudeAtlas

repo-scoutlisted

Audit an open-source repository and return an INSTALL / SKIP verdict. Use only when the user explicitly invokes `$repo-scout` or explicitly names Repo Scout and asks to use it. Never auto-trigger for repository URLs, project websites, package pages, comparisons, installation requests, `/plugin install`, or shell commands.
wallmage/repo-scout · ★ 1 · AI & Automation · score 64
Install: claude install-skill wallmage/repo-scout
# Repo Scout Act as the user's technical advisor for any open-source tool — a skill pack, a developer tool, or a full application. Read the source, explain the actual mechanism, and lead with a decisive, binary recommendation: 🟢 INSTALL or 🔴 SKIP. Base the verdict on evidence, not popularity or marketing. The user's question is always the same: is this worth my time, and why? The audience is non-technical: they cannot "pick" or "decide" anything, so the skill decides for them. There is no middle verdict. The verdict is mechanical: 🔴 SKIP fires only when at least one of these eight is evidenced, and nothing else ever produces a red; when none fires, the verdict is 🟢. Never calibrate toward any expected ratio of green to red — apply the triggers and report what they yield: The inventory fields and verdict gates are closed lists. Do not add extra checks, policy concerns, approval conditions, or report topics from general code-review habits. 1. **Fake** — the headline claims have no implementing files; marketing wrapped around emptiness (substance ≤ 2, claims unsupported). 2. **Malicious or suspicious** — the deliberate-malice tripwire (hidden exfiltration, secret theft unrelated to the stated task, decode-then-execute payloads, reviewer manipulation), plus two cases a normal user cannot judge: the tool routes the user's secrets or data through the author's private server with no self-host alternative, or the published install artifacts cannot be matched to the visible so