notify
Fail
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
notify_cli.shis vulnerable to AppleScript injection. It constructs a command string forosascript -eby directly interpolating the$MESSAGE,$TITLE, and$SOUNDvariables. An attacker can provide a payload (e.g., using a closing quote and the&operator) to execute arbitrary commands on the system via AppleScript'sdo shell scriptfunctionality. - [REMOTE_CODE_EXECUTION]: Because the skill's parameters are often derived from untrusted external data processed by the agent, this injection vulnerability allows for remote code execution on the host machine.
- [DATA_EXFILTRATION]: Exploiting the command injection allows an attacker to execute shell commands to read sensitive local files, such as SSH keys or environment configuration files, and send them to a remote server.
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting untrusted data through its parameters without validation.
- Ingestion points: The
title,message, andsoundparameters inSKILL.mdare passed tonotify_cli.shvia stdin. - Boundary markers: Absent; no delimiters or warnings are used to separate user data from the command context.
- Capability inventory: The skill can execute arbitrary system commands on macOS through the
osascriptutility. - Sanitization: Absent; the script fails to escape quotes or other control characters before passing them to the AppleScript engine.
Recommendations
- AI detected serious security threats
Audit Metadata