The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/with_server.py executes shell commands using subprocess.Popen with shell=True to handle the --server arguments. This is designed to allow starting local development servers (e.g., using cd and &&) but provides a mechanism for arbitrary command execution.\n- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection when interacting with web applications.\n
Ingestion points: The agent is instructed to read the rendered DOM using page.content() and capture browser console logs in examples/console_logging.py.\n
Boundary markers: No specific delimiters or "ignore" instructions are used when the agent processes this untrusted external data.\n
Capability inventory: The skill can execute arbitrary shell commands via scripts/with_server.py and write files to /mnt/user-data/outputs/ and /tmp/.\n
Sanitization: No sanitization or validation of the ingested HTML or log content is performed before the agent analyzes it to identify selectors or debug behavior.\n- [COMMAND_EXECUTION]: The SKILL.md instructions explicitly tell the agent not to read the source code of the helper scripts and to treat them as black boxes. This practice reduces the visibility and oversight of the command execution happening in the background.

webapp-testing