The Agent Skills Directory

[PROMPT_INJECTION]: The skill interacts with untrusted external websites, creating an indirect prompt injection surface (Ingestion points: browser_open, browser_get-text in SKILL.md). Mitigation is present through the mandatory use of the aidefence_is_safe safety gate which serves as a boundary marker and sanitization step before content is processed by the agent. Capability inventory includes Bash, Write, and browser_eval as specified in SKILL.md. \n- [COMMAND_EXECUTION]: The skill references framework-specific commands such as trajectory-end, rvf compact, and /ruflo-browser replay for managing test session lifecycle and artifacts. \n- [REMOTE_CODE_EXECUTION]: The skill utilizes browser_eval for executing JavaScript assertions within the browser context, which is standard functionality for automated UI testing tools.

browser-test