The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/with_server.py enables the execution of arbitrary shell commands provided via the --server argument. It uses subprocess.Popen with shell=True, which allows for complex shell syntax and command chaining (e.g., cd dir && npm start). The script also executes the final test command using subprocess.run.
[DYNAMIC_EXECUTION]: The skill's workflow encourages the agent to dynamically generate and execute Python Playwright scripts to automate browser interactions. This runtime generation and execution of code is a core capability that could be misused if the agent's decision-making is compromised.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with and processes data from local web applications, which serves as an ingestion point for untrusted content. A malicious application could attempt to influence the agent's behavior through hidden text or manipulated output.
Ingestion points: External data is ingested through Playwright functions like page.goto(), page.content(), and the console log listener in examples/console_logging.py.
Boundary markers: The skill does not employ boundary markers or specific instructions to the agent to ignore instructions embedded within the web pages it tests.
Capability inventory: The agent possesses the capability to execute shell commands via scripts/with_server.py, write files to local storage (e.g., /mnt/user-data/outputs/), and perform complex browser automations.
Sanitization: There is no evidence of content sanitization or validation before the agent interprets the rendered DOM or logs for reconnaissance.

webapp-testing