linux-kernel-exploits

Fail

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches shell scripts and exploit source code from multiple unverified third-party GitHub repositories, including mzet-, jondonas, xkaneiki, Liuk3r, Notselwyn, and Markakd.
  • [REMOTE_CODE_EXECUTION]: The instructions guide users to download tools like linux-exploit-suggester.sh via curl and execute them locally. It also describes a workflow for pulling and running custom exploits from a user-controlled attackbox to the target system.
  • [COMMAND_EXECUTION]: The skill provides embedded C source code for several kernel exploits, including DirtyPipe and DirtyCow. These are written to local files, compiled using gcc, and executed to gain root privileges.
  • [DATA_EXFILTRATION]: The skill contains commands and payloads specifically designed to access and dump sensitive system files such as /etc/shadow and /etc/passwd as part of the privilege escalation process.
  • [CREDENTIALS_UNSAFE]: The embedded exploit source code for DirtyPipe and DirtyCow contains hardcoded password hashes that are used to create new administrative accounts during a successful exploit.
Recommendations
  • HIGH: Downloads and executes remote code from: http://ATTACKBOX:8080/les.sh, https://raw.githubusercontent.com/mzet-/linux-exploit-suggester/master/linux-exploit-suggester.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 16, 2026, 10:05 AM
Security Audit — agent-trust-hub — linux-kernel-exploits