The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/with_server.py uses subprocess.Popen with shell=True to execute server startup commands provided via the --server command-line argument. This allows for the execution of complex shell command strings (e.g., using && or cd) and potentially arbitrary commands depending on how the agent interprets instructions.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted external data from web pages, creating a vulnerability surface for indirect prompt injection.
Ingestion points: examples/element_discovery.py retrieves inner text and attributes from DOM elements; examples/console_logging.py captures and prints browser console messages.
Boundary markers: Absent. The skill does not provide boundary markers or explicit instructions to the agent to treat discovered web content as untrusted or to ignore embedded instructions.
Capability inventory: The skill provides significant capabilities, including arbitrary command execution via scripts/with_server.py and file writing (e.g., logs and screenshots) in the example scripts.
Sanitization: Absent. Content retrieved from the browser is captured and presented to the agent's context without filtering, validation, or escaping.

webapp-testing