mailbox
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill's instructions focus on establishing a communication protocol between agent sessions. No attempts to override safety guidelines or bypass instructions were detected.
- [DATA_EXFILTRATION]: No network activity was identified. The skill operates exclusively on the local filesystem within a controlled root directory (/tmp/agent-mailbox). It does not access sensitive user directories such as SSH or AWS configurations.
- [REMOTE_CODE_EXECUTION]: The logic is entirely contained within local shell scripts using standard utilities. There are no patterns involving downloading and executing code from external URLs.
- [COMMAND_EXECUTION]: Command usage is limited to routine filesystem operations (mkdir, mv, find) and text processing (sed, awk, grep). Input is properly quoted or escaped, particularly in the JSON registry management logic.
- [CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or secrets were found. The skill uses generated UUIDs for identity management, which are not sensitive secrets.
Audit Metadata