The Agent Skills Directory

[Remote Code Execution] (CRITICAL): The skill instructs users to install the Cursor CLI using curl https://cursor.com/install -fsS | bash. Piped execution of remote scripts from non-trusted domains is a critical security risk as the script content is not verified before execution.
[Indirect Prompt Injection] (HIGH): The skill is designed to process untrusted data (logs, screenshots, code reviews) and possesses high-privilege capabilities including file modification and command execution.
Ingestion points: Untrusted data enters via the -p prompt flag, @ context selection, and log/screenshot analysis commands in SKILL.md.
Boundary markers: Absent. There are no instructions for delimiting or ignoring instructions embedded within the data being analyzed.
Capability inventory: The skill facilitates arbitrary file modification, code refactoring, and automated command execution, particularly when using the --force flag in CI/CD environments (referenced in SKILL.md).
Sanitization: Absent. There is no evidence of input validation or sanitization before external content is processed by the agent.
[Privilege Escalation] (HIGH): The skill documentation recommends using sudo apt install tmux to facilitate automation. Encouraging the use of sudo for dependency installation increases the risk of system-level compromise if the environment is malicious.
[Persistence & System Modification] (LOW): The skill instructions include modifying shell configuration files (~/.zshrc, ~/.bashrc) to update the system PATH. While common for developer tools, this represents a persistent modification of the user's environment.

cursor-agent