RCE learning path

Orientation and first principles — what the bug class is and how it behaves.

1. Command injection in Python: examples and prevention
   10 min readsnyk.io2026
   Library for preventing command injection in Python, detailing vulnerabilities arising from insecure use of `os.system`, `subprocess.run` with `shell=True`, dynamic command construction, and `eval()`. It highlights examples like MLflow and PaddlePaddle, and emphasizes mitigation through strict input validation, sanitization, and the use of parameterized queries to protect against unauthorized command execution and system compromise.
2. Understanding command injection vulnerabilities in Go
   7 min readsnyk.io2026
   Library for mitigating command injection vulnerabilities in Go applications. It details how insecurely handling user input with packages like `os/exec` can lead to attackers executing arbitrary commands, potentially causing data breaches or system compromise. The library emphasizes secure coding practices such as rigorous input validation and sanitation, preferring safe APIs (like `github.com/disintegration/imaging` for image manipulation) over direct system command execution, and refactoring vulnerable code to use `exec.Command` with separate arguments instead of constructing shell strings.
3. How I Found a Bug Worth $3,500 — In a Feature Nobody Was Watching.
   6 min readinfosecwriteups.com2026
   Writeup detailing the discovery of a storage-exhaustion flaw (CWE-434, CWE-770) and a stored XSS vulnerability (CWE-79) within a B2B SaaS platform's file upload feature. The storage exhaustion occurred due to a spoofed file size field, allowing for resource consumption. The XSS vulnerability leveraged an unsanitized filename that executed within an admin's browser, leading to potential session hijacking and platform compromise. The report emphasizes the importance of scrutinizing metadata, understanding the downstream impact of vulnerabilities, and re-testing "boring" features.
4. Exposed JDWP Exploited in the Wild: What Happens When Debug Ports Are Left Open
   5 min readwiz.io2026
   Library for detecting and preventing attacks exploiting exposed Java Debug Wire Protocol (JDWP) interfaces. This protocol, often unintentionally exposed in popular applications like TeamCity, Jenkins, and Spring Boot, allows for remote code execution. Attackers can leverage this misconfiguration to deploy malware, establish persistence through various methods including systemd services and cron jobs, and execute stealthy cryptomining payloads like customized XMRig. The library aids in identifying exploitation attempts and the subsequent malicious activities, such as those observed in the wild with rapid exploitation of JDWP ports.
5. Defense at AI speed: Microsofts new multi-model agentic security system tops leading industry benchmark
   15 min readmicrosoft.com2026
   Library for agentic AI-driven vulnerability discovery, codename MDASH, utilizes over 100 specialized agents and an ensemble of models to find and prove exploitable bugs. This system orchestrated across frontier and distilled models achieved top scores on industry benchmarks, including identifying 16 new vulnerabilities in Windows networking and authentication, four of which were Critical remote code execution flaws in components like the TCP/IP stack and IKEv2 service. MDASH's end-to-end pipeline includes stages for preparation, scanning, validation, deduplication, and proof, demonstrating a move towards production-grade, enterprise-scale AI vulnerability defense.

Real testing methodology, tooling, and writeups that show the work.

1. Enterprise Tech In, Shell Out (Progress Kemp LoadMaster Uninitialized Heap to Pre-Auth RCE CVE-2026-8037) - watchTowr Labs
   13 min readlabs.watchtowr.com2026
   Writeup on CVE-2026-8037, an uninitialized heap to pre-authentication Remote Code Execution vulnerability in Progress Kemp LoadMaster. The vulnerability arises from improper handling of user input within the `escape_quotes` function, which fails to null-terminate an allocated buffer. This can lead to information disclosure or code execution when the uninitialized memory is later processed by the API. The writeup details the vulnerability's mechanism and the specific code changes introduced in the patch.
2. AutoJack: How a single page can RCE the host running your AI agent
   16 min readmicrosoft.com2026
   Technique AutoJack demonstrates how a single web page can achieve remote code execution on the host running an AI agent, specifically targeting AutoGen Studio. By exploiting three weaknesses in the Model Context Protocol (MCP) WebSocket — an origin allowlist easily bypassed by an agent, opt-out authentication middleware, and unvalidated `server_params` from the URL—an attacker can trick the browsing agent into executing arbitrary commands on the host. This crosses the localhost trust boundary, turning the agent into a delivery vehicle for RCE.
3. Remote Code Execution in Ghost CMS (CVE-2026-29053)
   8 min readhiddeninvestigations.net2026
   Writeup on CVE-2026-29053, a remote code execution vulnerability in Ghost CMS versions 0.7.2 through 6.19.0. The flaw arises from unsafe expression evaluation within the theming system, where specially crafted themes can exploit a dependency chain involving the `jsonpath` and `static-eval` libraries. Exploitation requires an administrator to upload and activate a malicious theme, leading to arbitrary JavaScript execution on the server during page rendering, potentially impacting supply-chain trust and admin-targeted deception.
4. Ni8mare: Unauthenticated Remote Code Execution in n8n (CVE-2026-21858)
   22 min readcyera.com2026
   Writeup of CVE-2026-21858, an unauthenticated remote code execution vulnerability in n8n discovered due to a Content-Type confusion bug. Attackers can exploit this flaw by crafting a malicious request that manipulates the `req.body.files` object, allowing them to read arbitrary local files and achieve full takeover of n8n instances. This issue impacts over 100,000 servers globally and has a CVSS score of 10.0. Users should upgrade to n8n version 1.121.0 or later for remediation.
5. SSTI Exploitation with RCE Everywhere | YesWeHack
   7 min readyeswehack.com2026
   Writeup detailing advanced Server-Side Template Injection (SSTI) exploitation techniques for achieving Remote Code Execution (RCE) without quotes or external plugins. It covers payloads for Jinja2, Mako, Twig, Smarty, Blade, Groovy, and FreeMarker, demonstrating how to bypass auto-escaping and exploit built-in functions like `chr`, `popen`, `passthru`, and `execute` across various languages and frameworks.

Novel research, edge cases, and the techniques that push the class forward.

1. I went looking for a managed-Postgres provider. Instead, I found a vulnerability in a 4-star PostgreSQL extension available everywhere! and turned it into code execution at NeonDB, Supabase, Xata and many other PostgreSQL service companies
   29 min readmehmetince.net2026
   Library for discovering and exploiting vulnerabilities in managed PostgreSQL instances. It details a memory corruption bug in the popular PostGIS extension's `address_standardizer` function, chaining it with a separate memory disclosure to achieve RCE on providers like NeonDB, Supabase, and Xata. The research also encompasses broader systemic risks within the managed PostgreSQL industry, including further extension vulnerabilities and core PostgreSQL RCEs.
2. Chaining Security Bugs in Discuz! X5.0: from Race Condition to Pre-Auth RCE
   28 min readkarmainsecurity.com2026
   Library for chaining vulnerabilities in Discuz! X5.0, demonstrating a pre-authentication RCE attack. The exploit combines a Cross-Context Token Reuse leading to a Race Condition and Authentication Bypass, a custom OCR model for CAPTCHA bypass, and an administrative Local File Inclusion (LFI) vulnerability to achieve full server control.
3. MeshCentral: From XSS to RCE
   8 min readtechanarchy.net2026
   Writeup detailing a full RCE exploit chain against MeshCentral, starting with an XSS vulnerability. The analysis demonstrates how a large language model, Claude Opus, was used to discover and develop a Proof of Concept, which was then refined to allow an attacker to impersonate an existing agent by extracting credentials from local files. The article contrasts this with a failed attempt using a local, uncensored LLM, highlighting the practical application of AI in security research and exploit development.
4. Popping Root on UniFi OS Server: Unauthenticated RCE Chain Detection & Analysis
   18 min readbishopfox.com2026
   Tool for detecting unauthenticated RCE chains on UniFi OS Server, specifically addressing CVE-2026-34908, CVE-2026-34909, and CVE-2026-34910. This vulnerability allows attackers to bypass authentication, perform path traversal, and achieve command injection leading to root privileges. The tool aids defenders in identifying exposed systems and recommends immediate patching, network segmentation, and secret rotation, as exploitation grants access to sensitive data and control over managed devices.
5. React2Shell: Node.js RCE Against a Production Next.js App
   52 min readhunt.io2026
   Analysis of CVE-2025-55182, "React2Shell," details a Node.js Remote Code Execution vulnerability in Next.js applications utilizing React Server Components. The exploit leverages the Flight protocol's unsafe deserialization to trigger `child_process.spawnSync()`, allowing arbitrary shell commands with server process privileges. The report reconstructs a six-stage attack campaign, including C2 communication across multiple servers and the use of Lachlan Davidson's "02-meow-rce-poc" for RCE confirmation, despite defensive measures like container restrictions limiting further attacker progression.
6. Complete Defense Against Node.js RCE: Real-World Exploit Analysis
   10 min readseekerslab.com2026
   Analysis of Node.js RCE vulnerabilities, including CVE-2022-24329, details how attackers exploit `child_process.exec` misuse and improper input validation to achieve command injection. The article contrasts vulnerable code patterns, such as direct passing of user input to `exec`, with secure alternatives like `spawn` or `execFile` and emphasizes strict input validation and sanitization to prevent shell meta-character interpretation. It also discusses the need for an integrated security approach, combining SAST/DAST, cloud workload security with SeekersLab's FRIIM CNAPP, and real-time threat detection via Seekurity SIEM/SOAR, augmented by KYRA AI Sandbox for analyzing suspicious code.
7. Leading the Blind to Light! - A Chain to RCE
   5 min readblog.zsec.uk2017
   Writeup detailing a Remote Code Execution chain on Oracle E-Business Suite. The exploit begins with an authentication bypass, leading to blind XXE and information disclosure. This disclosure helps identify an internal endpoint, which through further fuzzing, reveals an SQL injection vulnerability. By re-enabling `xp_cmdshell` via SQL injection, the attacker achieves command execution with administrator privileges.
8. Before the first prompt: Code execution paths in trusted coding-agent projects
   6 min readsecuritylabs.datadoghq.com2026
   Library for identifying code execution paths in trusted coding-agent projects, demonstrating how configurations like Codex's Model Context Protocol (MCP) and Claude Code's project-controlled PATH can trigger attacker-controlled processes before the first user prompt, bypassing typical hook reviews. The library highlights that vulnerabilities extend beyond malicious hooks and skills, encompassing editor tasks, environment settings, and runtime startup files.