The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interpolates the user-provided name variable, collected via scripts/gather-plugin-info.cjs, directly into a shell command: claude plugin validate <repo-root>/plugins/<name>. Because the gathering script does not validate or sanitize this input, it is vulnerable to command injection if a user provides a name containing shell metacharacters (e.g., ;, &&, or |).
[COMMAND_EXECUTION]: The name variable is used to construct file system paths for directory and file creation (e.g., plugins/<name>/README.md). Without sanitization, this allows for directory traversal attacks where a name containing ../ sequences could be used to write or overwrite files outside the intended directory structure.

create-plugin