← Zurück zu Entwickler interner Systeme

Skill · Version 1.0.0

Sichere Uploads und Downloads

Schützt Dateieingang und Auslieferung durch Inhaltsprüfung, Ressourcenrechte und Browserisolation statt Vertrauen in Namen oder Links.

secure-upload-download

Teammitglieder mit diesem Skill

Entdecke, wie der Skill hilft

Wann ist der Skill hilfreich?

Nutze ihn bei Uploads, Remoteimporten oder privaten Download-URLs für Anwendungsdatensätze.

Wobei hilft er dir?

  • Prüft tatsächlichen Inhalt, Archivpfade und entfernte Importe.
  • Prüft Mandanten- und Elternressourcenrechte vor URL-Signierung.
  • Trennt Dateiauslieferungsursprung und beschränkt Browserausführung.

Das könntest du fragen

Weise getarnte HTML-Uploads und unberechtigte signierte Downloadlinks zurück. Benenne aktives HTML als Bild um und verwende eine fremde Dateikennung. Inhaltsprüfung und Autorisierung müssen vor Ausgabe eines signierten Links scheitern.

Ausführliche Dokumentation

Hier findest du die ursprüngliche Anleitung, Beispiele und Prüfschritte des Skills in ihrer Originalsprache.

Details…Weniger anzeigen

Secure Upload & Download

User-uploaded files are untrusted binary inputs that can carry active scripts (XSS via SVG/HTML), path traversal payloads, decompression bombs, SSRF vectors or malware. Enforce defense in depth at both ingress (upload) and egress (download/preview).

1. Untrusted Input Validation (Beyond Extensions)

  • Never trust File.name extension or the client's Content-Type header:
    • Inspect magic bytes (file signatures) from the first chunk of the actual uploaded bytes (e.g., file-type / libmagic) and verify that the detected format matches the allowed allowlist for that feature.
    • For images, decode headers/dimensions with a battle-tested parser before accepting.
  • High-risk formats:
    • SVG, HTML, XML, XHTML, SWF: contain executable JavaScript (<script>, onload=, foreignObject) and XXE vectors. Never allow raw SVG/HTML as user avatars or inline images served from the application origin; rasterize to PNG/WebP/AVIF or sanitize and serve only as an attachment from a sandboxed domain.
    • Archives (.zip, .tar.gz, .docx, .xlsx): enforce strict Zip Slip path validation (path.resolve(destDir, entryName).startsWith(destDir + path.sep)), reject symlinks, and cap total uncompressed byte count, compression ratio and file count to prevent zip bombs.
  • Remote URL imports (SSRF Prevention):
    • If users can import files by URL, block private/loopback/link-local/cloud-metadata IP ranges (127.0.0.0/8, 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16, 169.254.169.254, IPv6 ::1, fc00::/7, fe80::/10) at DNS resolution and socket connection time, disable redirects to internal hosts, and enforce strict byte and timeout limits.

2. Resource-Level Authorization & Presigned URLs

  • Verify tenant & resource permissions on every operation: check that the authenticated user is authorized to upload to, view, share or delete the target parent resource (project, task, message, invoice) before issuing a signed URL or streaming bytes.
  • Scoped Direct-to-Object-Store Uploads:
    • Never expose bucket-wide credentials to the browser.
    • Issue short-lived presigned PUT or POST policies (e.g., 5–15 minutes) locked to a single server-generated object key, exact Content-Type and Content-Length / content-length-range.
  • Signed Download URLs Are Bearer Tokens:
    • Remember that a presigned S3/Blob GET URL cannot be individually revoked once issued until its TTL expires. Keep signed download URL TTLs short (e.g., 60–300 seconds) and generate them on demand after checking current authorization, or proxy through an authenticated edge/application endpoint when instant revocation is required.

3. Safe File Serving & Browser Execution Isolation

When serving user files back to browsers:

  • Always set X-Content-Type-Options: nosniff so the browser never guesses executable HTML/JS from a text or image payload.
  • Serve from an isolated origin/bucket domain (e.g., files.example-usercontent.com) separate from the main application origin so an inline file can never read application cookies, localStorage or CSRF tokens.
  • Force Content-Disposition: attachment; filename="..."; filename*=UTF-8''... for downloads and any file type not explicitly safe for inline rendering, and apply Content-Security-Policy: sandbox on file responses.

Enthaltene Dateien 1

  • SKILL.md