webapp-testing

Warn

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/with_server.py uses subprocess.Popen with shell=True to execute commands provided via the --server argument. This allows for arbitrary shell command execution, including chained commands (e.g., using && or ;). While intended for starting local servers, this mechanism can be exploited if the input to the command is influenced by untrusted data.
  • [PROMPT_INJECTION]: The SKILL.md file contains instructions telling the agent "DO NOT read the source until you try running the script first" and describing the scripts as "black-box scripts rather than ingested into your context window." These instructions discourage the AI from performing safety inspections of the code it is about to execute, which is a characteristic of behavior-overriding prompts.
  • [COMMAND_EXECUTION]: The examples examples/element_discovery.py, examples/console_logging.py, and examples/static_html_automation.py demonstrate patterns for interacting with local web applications and the file system. These scripts have the capability to read and write files (e.g., /mnt/user-data/outputs/console.log, /tmp/page_discovery.png) and capture browser data.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 10, 2026, 03:23 AM