The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands provided by the user, specifically skills add and skills remove. It also uses PowerShell for inventory and validation tasks, such as filtering skill lists using $_.name -match '<pattern>'.
[REMOTE_CODE_EXECUTION]: The skill explicitly references npx skills add. The npx utility downloads and executes packages from the NPM registry at runtime, which is a vector for remote code execution.
[EXTERNAL_DOWNLOADS]: The skill manages external dependencies and directs the agent to interact with a centralized skills repository (~/.agents/skills) and various agent-specific directories.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied patterns to identify targets for uninstallation and inventory. These patterns are interpolated into PowerShell commands (-match '<pattern>'), creating a surface for indirect injection if user input contains malicious regex or shell escape sequences.
[PRIVILEGE_ESCALATION]: The skill performs high-privilege file system operations, including creating symbolic links (symlinks), junctions, and performing backups/replacements of directories in system-sensitive paths like ~/AppData/Roaming/.

install-skills