Description

gpt-crawler through 1.5.1 fails to validate the outputFileName parameter in the POST /crawl endpoint, allowing unauthenticated attackers to write arbitrary files to any filesystem path. Attackers can supply absolute paths or parent-directory segments to overwrite existing files with content sourced from attacker-controlled URLs.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "8.6",
  "pubDate": "2026-08-28T20:20:20.093Z",
  "pubdate": "2026-08-28T20:20:20.093Z",
  "executiveSummary": "The gpt-crawler software, through version 1.5.1, is susceptible to an arbitrary file write vulnerability within the POST /crawl endpoint. This flaw stems from a lack of input validation regarding the outputFileName parameter, enabling unauthenticated remote attackers to manipulate the file system write operations.\nBy supplying crafted input—specifically absolute paths or directory traversal sequences (e.g., ../)—an attacker can bypass intended file system boundaries to overwrite sensitive configuration files, source code, or system binaries with attacker-controlled content. This vulnerability presents a critical security risk, as it permits unauthorized file modification on the host system without requiring prior authentication. Successful exploitation leads to remote code execution or complete system compromise, depending on the file targeted for overwriting. Because the application processes user-supplied URLs to populate the content of the target file, attackers can effectively inject malicious payloads into arbitrary locations within the server's file system, significantly undermining the integrity and availability of the affected environment.",
  "technicalDetails": "The vulnerability resides within the request handling logic of the POST /crawl endpoint. When the application receives a request, it fails to sanitize or validate the outputFileName parameter, which is intended to define the destination path for the crawled data. The root cause is an improper neutralization of special elements used in pathnames, commonly referred to as Path Traversal (CWE-22).\nAn unauthenticated attacker can exploit this by manipulating the outputFileName parameter to include directory traversal sequences such as '..'. By combining these segments with absolute file paths, the attacker can break out of the designated output directory and target any location on the file system accessible to the process executing the gpt-crawler application. The application subsequently writes the result of the crawl operation—which is sourced from an attacker-controlled URL—directly to the malicious path specified in the parameter.\nThe attack flow proceeds as follows: First, the attacker identifies the target server running gpt-crawler. Second, the attacker constructs a malicious payload for the POST /crawl request, setting the outputFileName parameter to a critical path, such as an SSH authorized_keys file, a web application configuration file, or a system crontab. Third, the attacker provides a URL pointing to a malicious resource. Upon receipt, the application retrieves the content from the provided URL and writes it to the location defined by the manipulated outputFileName. Finally, the application overwrites the legitimate file with the injected content, leading to persistent code execution or system reconfiguration.\nThe vulnerability affects gpt-crawler versions through 1.5.1. The flaw is exploitable over the network without any authentication or elevated privileges. Because the application performs the file write operation with the permissions of the user running the service, the scope of the impact is strictly limited to the file system reach of that user. However, in many containerized or default configurations, this is sufficient to achieve full system compromise."
}