npm-supply-chain-audit
できること
Audit npm projects for Shai-Hulud-class supply-chain exposure: lifecycle scripts, Git dependency prepare hooks, release-age gaps, publish-token exposure, trusted publishing, signed releases, and verifier docs.
インストールすると、このページがデスクトップ版 AgentsRoom で開きます。アプリが未インストールの場合はダウンロードページに移動します。
SKILL.md
--- name: npm-supply-chain-audit description: Audit npm projects for Shai-Hulud-class supply-chain exposure: lifecycle scripts, Git dependency prepare hooks, release-age gaps, publish-token exposure, trusted publishing, signed releases, and verifier docs. --- # npm-supply-chain-audit Use this skill when reviewing a JavaScript or TypeScript repository for install-time malware exposure, compromised npm maintainer risk, or CI publish-path abuse. ## Triggers - "npm supply chain audit" - "Shai-Hulud" - "malicious npm package" - "trusted publishing review" - "publish token exposure" - "Git dependency prepare script" - "router_init.js" / "tanstack_runner.js" ## Audit sequence ### 1. Dependency source scan Search manifests and lockfiles for exotic sources: ```bash rg -n '"(git\\+|git://|github:|file:|link:)|https?://[^"]+\\.(tgz|tar\\.gz)' package.json package-lock.json ``` Findings to escalate: - `github:` or `git+` dependencies, especially in `optionalDependencies`. - Direct tarballs from non-registry hosts. - `file:` or `link:` sources outside deliberate local workspace development. Run the known-affected package feed scan as a separate hard gate: ```bash npm run lint:affected-packages AIWG_AFFECTED_PACKAGES_CSV=/mnt/ops/users/roctinam/Downloads/22-packages.csv npm run lint:affected-packages AIWG_AFFECTED_PACKAGES_CSV=https://gist.githubusercontent.com/<user>/<gist-id>/raw/22-packages.csv npm run lint:affected-packages ``` Treat exact package/version hits as incident evidence. Preserve the package name, version, published timestamp, detected timestamp range, and feed source URL/path in the finding. ### 2. Lifecycle-script review Inspect root package scripts and nested package manifests: ```bash rg -n '"(preinstall|install|postinstall|prepare)"' package.json package-lock.json . ``` Treat install-time scripts as code execution on every developer machine and CI runner. Remove them unless the package cannot function without them. If one must remain, document the exact reason and what it can access. ### 3. Release-age gate Check for `.npmrc`: ```ini min-release-age=7 ``` Confirm contributors and lockfile-changing CI jobs use npm 11.5+. Use 10 days or higher for release-prep dependency churn and security- sensitive branches. ### 4. Publish-path hardening Review release workflows for: - npm trusted publishing or another OIDC publish path. - No long-lived npmjs.org publish token once trusted publishing is active. - `permissions: id-token: write` scoped only to the job that publishes. - Signed tag verification before build, pack, publish, or asset upload. - SHA-pinned actions and digest-pinned containers in release jobs. - No publish step on fork pull requests or untrusted PR triggers. ### 5. Evidence and user verification For packages users install from a registry, require: - npm provenance attestation where supported, - signed git tag, - signed release tarball or equivalent artifact signature, - SBOM, - user-facing verification docs. The verifier docs should avoid commands that run lifecycle scripts while checking an artifact. Prefer `npm install --package-lock-only --ignore-scripts <pkg>@<version>` followed by `npm audit signatures`. ## Incident-response trigger If a known malicious version was installed or a suspicious lifecycle script ran, assume secrets reachable from that environment are exposed. Rotate npm tokens, GitHub/Gitea tokens, cloud credentials, Kubernetes service account tokens, Vault tokens, and deployment secrets. Then audit recent publishes and workflow runs. If the affected-package feed hit a CI runner or workstation cache, quarantine that cache before reuse. ## Output format Lead with findings, ordered by severity: ```markdown ## Findings - CRITICAL: <file:line> <risk> <fix> - HIGH: <file:line> <risk> <fix> ## Clean Checks - No Git dependency sources found. - No install lifecycle scripts found. ## Follow-up Issues - <title>, <acceptance criteria> ``` ## References - npm trusted publishing: <https://docs.npmjs.com/trusted-publishers> - npm audit signatures: <https://docs.npmjs.com/cli/v11/commands/npm-audit#audit-signatures> - npm `min-release-age`: <https://docs.npmjs.com/cli/v11/using-npm/config#min-release-age>
さらに詳しく
Claude Ads:広告アカウントを監査する Claude Code スキル
Claude Ads は Claude Code 向けのオープンソーススキル。Google、Meta、LinkedIn、TikTok、Amazon 広告など250以上の項目をチェックし、100点満点のスコアと優先順位付きのアクションプランを、わずか10分ほどで出力します。インストール方法、コマンド、限界、そして AgentsRoom での使いこなし方まで解説。
AGENTS.md: すべてのコーディングエージェントに効く唯一のコンテキストファイル(Codex、Antigravity、Claude)
AGENTS.md は、AI コーディングエージェントがコードに触れる前に読み込む、ポータブルな指示ファイルです。何を書くべきか、CLAUDE.md との違い、そして Codex、Antigravity、Claude をまたいで単一のコンテキストを保つ方法を解説します。
AgentsRoomをダウンロード
すべてのAIエージェントを、すべてのプロジェクトで、ひとつのウィンドウから実行。
コンパニオンアプリ:外出先でもエージェントを確認
Claude、Codex、Antigravity CLI、またはその他の AI プロバイダーを使用します。
バグや要望を公開バックログに直接送信できます。