The Agent Skills Directory

[Remote Code Execution]: The skill suggests a one-line installation method for a utility.
Evidence: The command curl -fsSL https://raw.githubusercontent.com/pproenca/agent-tui/master/install.sh | sh in SKILL.md downloads and executes a script from a third-party source without a manual review step.
[External Downloads]: The skill references several external sources for tool installation.
Evidence: It suggests using npm, pnpm, and bun to install the agent-tui package, as well as cargo install from the GitHub repository https://github.com/pproenca/agent-tui.git.
[Command Execution]: The skill provides instructions for managing system processes and files to facilitate its daemonization logic.
Evidence: The script snippet in SKILL.md uses tmux kill-session, agent-tui daemon stop, and rm -f /tmp/agent-tui* to prepare the execution environment.
[Indirect Prompt Injection]: The skill is designed to automate terminal interactions based on visual feedback, which introduces an attack surface where untrusted data could influence behavior.
Ingestion points: Terminal screen content is captured via agent-tui screenshot (referenced in SKILL.md).
Boundary markers: The instructions do not define delimiters to separate terminal content from agent instructions.
Capability inventory: The skill can perform actions via agent-tui type, agent-tui press, and agent-tui run (referenced in SKILL.md).
Sanitization: The skill does not describe mechanisms to sanitize or validate terminal content before it is processed by the agent.

agent-tui