The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The documentation in README.md and install.sh encourages installation via a piped shell command (curl -fsSL ... | bash). This script is hosted on the author's personal GitHub repository and is designed to clone or update the skill's file structure.
[EXTERNAL_DOWNLOADS]: The developer utility scripts/generate-references.sh performs a shallow clone of the official hyprwm/hyprland-wiki repository to pull the latest documentation into the references/ directory.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface (Category 8):
Ingestion points: Documentation files in references/*.md are ingested into the agent's context based on user queries.
Boundary markers: Absent; SKILL.md instructs the agent to read the reference files directly without specifying delimiters or warnings regarding embedded instructions.
Capability inventory: The skill is designed for system configuration, implying the agent will likely have shell access and command execution capabilities to assist the user.
Sanitization: Absent; external markdown content is processed without filtering or validation.
[COMMAND_EXECUTION]: The reference files contain extensive examples of shell commands for package management (pacman, zypper, apt), system service control (systemctl), and hardware configuration (chvt). While these are contextually relevant for Hyprland setup, they represent a broad set of capabilities the agent might be prompted to execute.

hyprland