The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the infsh command-line utility to interface with a remote browser automation service and manage browser sessions.
[REMOTE_CODE_EXECUTION]: Provides an execute function that enables the execution of arbitrary JavaScript code within the browser's context, which is a core feature for modern web automation and scraping.
[DATA_EXFILTRATION]: Includes capabilities to extract sensitive data from browser sessions, such as cookies and localStorage content, as demonstrated in the documentation's authentication patterns.
[PROMPT_INJECTION]: The skill inherently possesses an indirect prompt injection surface (Category 8) due to its function as a web browser interacting with untrusted content.
Ingestion points: Any URL processed through the open or goto actions (e.g., in templates/capture-workflow.sh).
Boundary markers: Absent; the skill does not explicitly frame external content with delimiters to warn the agent about potential instructions in the data.
Capability inventory: JavaScript execution (execute), element interaction (interact), and file uploads (upload).
Sanitization: No sanitization of external website content is performed by the skill's scripts.

agent-browser