← ClaudeAtlas

dependency-auditlisted

Runs a comprehensive third-party dependency health check across every package manager and manifest in a project (npm/yarn/pnpm, pip/poetry/Pipenv, NuGet, Maven/Gradle, Cargo, Go modules, Composer, RubyGems, CocoaPods/SwiftPM, and container base images where relevant), then reports the results as one table (check, area, status, evidence, recommendation). Covers known vulnerabilities/CVEs (native audit tooling per ecosystem), outdated packages (patch/minor vs. major, migration risk), license compliance (copyleft/GPL/AGPL/LGPL conflicts with closed-source distribution, missing/unknown licenses), unused and dead dependencies, lockfile integrity and manifest/lockfile drift, reproducibility of installs, and supply-chain risk signals (unmaintained packages, single-maintainer risk, suspicious/low-download additions, install/postinstall scripts as an attack vector, typosquatting-risk names), plus whether automated update tooling (Dependabot/Renovate or equivalent) is configured. Use this whenever the user asks for a "
finnley07/AI-SKILLHUB · ★ 0 · AI & Automation · score 72
Install: claude install-skill finnley07/AI-SKILLHUB
# Dependency Audit A structured, evidence-based check of a project's third-party dependencies — vulnerabilities, staleness, license risk, dead weight, lockfile integrity, and supply-chain hygiene — across every package manager present in the repo. This is a deeper, broader complement to `cybersecurity-check` (which covers dependency vulnerabilities as a single line item, S16); this skill is the one to reach for whenever dependencies themselves — not the rest of the application surface — are the actual subject of the question. ## Ground rules - **Evidence or it didn't happen.** Every row needs a concrete pointer: an actual command and its actual output, an actual version number, an actual CVE/advisory ID, an actual file:line. Never write "dependencies look fine," "packages appear up to date," or "no vulnerabilities found" as a standalone claim — report what tool ran, against what manifest, and what it printed. - **A ✅ means "the audit tool ran and reported clean" — not "there are definitely no vulnerabilities."** Audit tools have real false-negative rates, especially for vulnerabilities disclosed in the last few days (the advisory database hasn't caught up yet) and for transitive/indirect dependencies some tools don't fully walk. Say this caveat once, up front, so the user doesn't over-trust a clean run. Phrase findings as "`npm audit` reported 0 vulnerabilities as of <date>," not "this project has no vulnerable dependencies." - **Don't invent scope you can'