The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to clone a repository from an unverified GitHub account at https://github.com/penligent/AI2PentestTool.git.
[REMOTE_CODE_EXECUTION]: After cloning the external repository, the skill requires installing unverified dependencies via pip install -r requirements.txt and executing the application's logic through python main.py.
[COMMAND_EXECUTION]: The tool's primary function is the execution of system commands to install security software. It utilizes subprocess.run and os.system equivalents to interact with package managers (apt, brew) and binary files.
[PRIVILEGE_ESCALATION]: The documentation explicitly instructs the user to provide administrative privileges using sudo for tool installation, allowing the unverified third-party scripts to run with root access.
[DYNAMIC_EXECUTION]: The skill implements an AIAgent class that generates "installation plans" (shell commands) dynamically based on AI responses. These plans are then executed by the ToolInstaller without visible sanitization, enabling potential arbitrary command execution if the AI output is manipulated.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill processes tool definitions and metadata from config.py (the TOOLS_INFO dictionary) to generate prompts for the AI agent.
Boundary markers: There are no visible delimiters or instructions provided to the agent to ignore potentially malicious content within the tool descriptions.
Capability inventory: The skill possesses capabilities for shell command execution (subprocess.run), environment variable access (OPENAI_API_KEY), and network communication.
Sanitization: The skill lacks validation or sanitization of the installation plans generated by the AI agent before they are passed to the shell for execution.

ai2pentesttool-installer