Skip to main content

dependency-audit

Audit a project's dependencies for known vulnerabilities (CVEs) and outdated packages across npm/pnpm/yarn, Go, Python, and Cargo, then produce a prioritized, read-only remediation report. Use when the teammate asks to audit or check dependencies, scan for vulnerable or outdated packages, run a supply-chain / dependency security check, or mentions npm audit, govulncheck, pip-audit, or cargo audit. Read-only — it never edits files or upgrades anything.

Informações da origem

Repositório
PramodTKodag/claude-code-skills
Última atividade na origem
29 de setembro de 2026 às 11:21
Idioma detectado do SKILL.md
inglês
Estrelas
2
Forks
0

dependency-audit: report vulnerable and outdated

Inspect a project’s dependency ecosystems, run their vulnerability and outdated-package tools, and produce a prioritized remediation report.

Prerequisites

The source covers npm/pnpm/yarn, Go, Python and Cargo projects. Each detected ecosystem needs its corresponding audit tool; unavailable tools are reported as not run.

How to use

Provide the project to audit or ask to check its dependencies. The workflow detects manifests and lockfiles, runs the native tools once, and summarizes severity, installed and fixed versions, advisory IDs and dependency paths. It separates patch or minor recommendations from changes needing review.

Limitations

The Skill reports and proposes; it does not install upgrades or apply fixes. A missing audit tool is not a passing result. Advisory IDs, severities and fixed versions must come from the actual tool output.

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

Explorador de arquivos
2 arquivos

Exibindo SKILL.md

SKILL.md
Instruções da origem · Visualização somente leitura
name
dependency-audit
description
Audit a project's dependencies for known vulnerabilities (CVEs) and outdated packages across npm/pnpm/yarn, Go, Python, and Cargo, then produce a prioritized, read-only remediation report. Use when the teammate asks to audit or check dependencies, scan for vulnerable or outdated packages, run a supply-chain / dependency security check, or mentions npm audit, govulncheck, pip-audit, or cargo audit. Read-only — it never edits files or upgrades anything.
user-invocable
true
allowed-tools
["Read","Grep","Glob","Bash(npm audit*)","Bash(npm outdated*)","Bash(pnpm audit*)","Bash(pnpm outdated*)","Bash(yarn audit*)","Bash(yarn outdated*)","Bash(govulncheck*)","Bash(go list*)","Bash(pip-audit*)","Bash(pip list*)","Bash(uv*)","Bash(cargo audit*)","Bash(cargo outdated*)","Bash(jq*)"]
effort
medium
# Dependency audit Scan the dependency graph for **known vulnerabilities** and **outdated packages**, then hand back one prioritized report. **Read-only**: run audit tools, never modify manifests, lockfiles, or code — the teammate (or `/dev-flow`) applies fixes. ## Procedure 1. **Detect ecosystems.** Glob for manifests — `package.json`, `pnpm-lock.yaml`/`yarn.lock`, `go.mod`, `requirements.txt`/`pyproject.toml`, `Cargo.toml`. Report which were found; audit only those. 2. **Run each ecosystem's native tools once** — the exact audit + outdated commands, their `jq` extraction filters, and install hints are in [references/ecosystems.md](references/ecosystems.md). Prefer the tool the lockfile implies (`pnpm` over `npm` when `pnpm-lock.yaml` exists, etc.). 3. **Missing tool = "not run", never a silent pass.** If an audit tool isn't installed, list it under *Tools not run* with the one-line install from the reference; do not infer results from the manifest. 4. **Synthesize** the outputs into the report below. Then stop — propose, don't apply. ## Token discipline (required) - **Never** paste raw `audit`/`outdated` JSON into the conversation. Always pipe it through the `jq` filter in the reference to keep only `{package, severity, id, installed, fixed, path}`, then work from that. - Run each tool **once**; capture the filtered summary and discard the raw output. - Read manifests with `grep`/targeted reads, never whole large lockfiles. - Load `references/ecosystems.md` only when you need a command you don't already have. ## Output Report, most severe first: - **Summary:** `N vulnerabilities (c critical / h high / m moderate / l low), M outdated` across the ecosystems scanned. - **Vulnerabilities** — table: `severity · package · installed → fixed · advisory/CVE · dependency path`. - **Outdated** — table: `package · current → latest · jump (patch / minor / major)`. - **Prioritized remediation:** 1. **Safe now** — patch/minor upgrades that fix a vuln or are low-risk, each with the **exact command**. 2. **Needs review** — major/breaking upgrades, flagged separately with why (API break, transitive pin). - **Tools not run** — any ecosystem whose audit tool was missing, with the install hint. State clearly that nothing was changed. If the teammate wants the fixes applied, recommend running them through `/dev-flow` (or manually) — this skill does not upgrade. ## Guardrails - Read-only: no `Edit`/`Write`, no `npm install`/upgrade, no lockfile changes. - Report severities and versions **as the tools report them** — do not guess CVE IDs or fixed versions. - If a scanned ecosystem has zero findings, say so explicitly rather than omitting it.
Ver no GitHub