The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of a Python script (scripts/screenshot.py) via the command line. It encourages the agent to interpolate user-provided parameters such as URLs, file paths, and CSS selectors directly into shell commands, creating a potential vector for command manipulation.\n- [DATA_EXFILTRATION]: The implementation uses Playwright to navigate to URLs without restricting schemes, potentially allowing the use of the file:// protocol to capture screenshots of sensitive local files (Exposure). Additionally, the output_path can be set to arbitrary locations, which could be exploited to overwrite files in sensitive directories.\n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the playwright package and its associated browser binaries. These are fetched from well-known and official sources.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted web content without adequate boundary markers or sanitization. \n
Ingestion points: The url and CSS selector arguments passed to scripts/screenshot.py.\n
Boundary markers: No delimiters or safety instructions are used to separate untrusted web content from the agent's instructions.\n
Capability inventory: The skill has the capability to write to the filesystem and access the network via Playwright.\n
Sanitization: There is no validation or filtering performed on the user-provided URL or the output path.

browser-screenshot