The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains numerous commands for exploiting system vulnerabilities and misconfigurations to gain root or administrator access, such as abusing Sudo permissions, SUID binaries, and cron jobs.
[REMOTE_CODE_EXECUTION]: Provides instructions to download and execute arbitrary scripts from remote servers using PowerShell's Invoke-WebRequest and Invoke-Expression commands.
[DATA_EXFILTRATION]: Includes techniques for extracting highly sensitive information, including user credentials, SSH private keys, and the Active Directory NTDS.dit database.
[CREDENTIALS_UNSAFE]: Hardcodes a sample password in a command-line argument for GPO abuse, which could lead to credential exposure if used literally.
[EXTERNAL_DOWNLOADS]: Directs the agent to download and use a variety of third-party offensive security tools and exploitation scripts from unverified sources.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes user-provided inputs for privilege escalation commands and interacts with system outputs without explicit sanitization or boundary markers. 1. Ingestion points: User commands requesting escalation or system state enumeration in SKILL.md. 2. Boundary markers: Absent; instructions do not include delimiters to separate untrusted system data from agent instructions. 3. Capability inventory: Shell command execution, file system access, and network capabilities are present across all provided examples in SKILL.md. 4. Sanitization: Absent; the skill lacks mechanisms to filter or escape potentially malicious strings returned by shell commands.

Privilege Escalation Methods