The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/with_server.py uses subprocess.Popen(shell=True) to execute server commands provided via the --server argument. This allows for arbitrary shell execution, including the use of shell metacharacters, piping, and command chaining, which is risky if the command strings are influenced by untrusted input.
[COMMAND_EXECUTION]: The script scripts/with_server.py executes the primary test command provided as trailing arguments using subprocess.run(), which could lead to arbitrary code execution if the command itself is untrusted.
[PROMPT_INJECTION]: The skill captures browser console logs in examples/console_logging.py and test-helper.js, creating a surface for indirect prompt injection where a malicious website could output instructions that the agent might inadvertently follow.
Ingestion points: Browser console messages are collected via page.on("console", ...) in examples/console_logging.py and test-helper.js.
Boundary markers: None identified; logs are captured and processed without delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill can execute shell commands (via scripts/with_server.py), write files to /mnt/user-data/outputs/, and perform browser automation.
Sanitization: No sanitization, escaping, or filtering of the console log content is performed before it is processed or saved.

webapp-testing