bni-121
Fail
Audited by Gen Agent Trust Hub on Apr 30, 2026
Risk Level: HIGHCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The shell scripts
scripts/submit-121.shandscripts/submit-referral.shcontain a command injection vulnerability where user-provided inputs (such as member names) are interpolated directly into a Python command string using unsafe shell expansion (e.g.,python3 -c "print('$NAME'[0])"). A crafted input containing a single quote followed by Python code (e.g.,'); import os; os.system(...)#) would allow arbitrary code execution in the agent's environment.\n- [CREDENTIALS_UNSAFE]: The skill instructions explicitly direct the agent to ask the user for their BNI Connect email and password. Collecting and processing raw credentials in plain text is a high-risk practice that exposes sensitive account information to the agent's context and session logs.
Recommendations
- AI detected serious security threats
Audit Metadata