The Agent Skills Directory

[Remote Code Execution] (HIGH): The skill instructs the agent to download and execute scripts directly from external URLs without verification or integrity checks.
Evidence: curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.0/install.sh | bash in WSL2 setup.
Evidence: curl -fsSL https://deb.nodesource.com/setup_24.x | sudo -E bash - in Linux setup.
Evidence: /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)" in macOS setup.
[Privilege Escalation] (HIGH): Several instructions require the use of sudo, which grants the setup process full administrative access to the system.
Evidence: sudo apt install -y nodejs and subsequent piping of remote scripts to sudo bash.
[External Downloads] (HIGH): The skill clones and runs code from a third-party GitHub repository ('AndyMik90/Auto-Claude') that is not on the trusted organizations list.
Evidence: git clone https://github.com/AndyMik90/Auto-Claude.git followed by npm run install:all.
[Indirect Prompt Injection] (LOW): The skill serves as a gateway to external data (README, guides, and repositories) that could contain malicious instructions processed by the agent during installation.
Ingestion points: SKILL.md (instructions), AndyMik90/Auto-Claude (cloned repository content).
Boundary markers: Absent; no clear separation between instructions and data-driven commands.
Capability inventory: Full shell execution, privilege escalation via sudo, network access (curl/git), and file system modification.
Sanitization: None; commands are executed as provided in the markdown guide.

auto-claude-setup