The Agent Skills Directory

[COMMAND_EXECUTION]: The utility script scripts/with_server.py uses subprocess.Popen(shell=True) and subprocess.run() to manage server processes and execute automation scripts. While this is the intended functionality for local development testing, it provides a powerful capability that could be misused if the agent is influenced by malicious input.
[PROMPT_INJECTION]: The skill contains a significant surface area for indirect prompt injection. \n
Ingestion points: The agent is instructed to use Playwright to navigate to URLs, inspect rendered DOM content, and capture console logs (as seen in SKILL.md and examples/element_discovery.py). \n
Boundary markers: The skill documentation does not provide boundary markers or explicit instructions for the agent to ignore or isolate instructions found within the web pages it interacts with. \n
Capability inventory: The agent has the ability to generate and execute local Python scripts and shell commands through the with_server.py helper. \n
Sanitization: No sanitization or validation steps are included to verify that the content retrieved from web applications does not contain instructions that could subvert the agent's logic.

webapp-testing