webapp-testing
Fail
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/with_server.pyusessubprocess.Popenwithshell=Trueto execute commands passed directly from the--servercommand-line argument. This allows for arbitrary shell command injection if untrusted or unvalidated input is provided to the script. - [PROMPT_INJECTION]: The documentation in
SKILL.mdcontains instructions that explicitly direct the agent to avoid reading the source code of the scripts provided with the skill ("DO NOT read the source until you try running the script first"). This is a deceptive tactic designed to bypass the agent's security review process and prevent it from discovering the dangerous implementation details in the helper scripts. - [DATA_EXPOSURE]: The skill facilitates capturing browser screenshots and console logs from local web applications. This capability could be used to exfiltrate sensitive data or credentials displayed on internal testing sites if the agent is directed to targets containing private information.
Recommendations
- AI detected serious security threats
Audit Metadata