The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it interacts with untrusted external data.
Ingestion points: Untrusted data enters the agent context via the browser_navigate tool when visiting URLs specified by the user (found in SKILL.md and references/playwright-api.md).
Boundary markers: The instructions do not specify any delimiters or explicit warnings to the agent to ignore instructions embedded within the content of the target websites.
Capability inventory: The skill utilizes browser_navigate, browser_take_screenshot, and browser_resize (across all files), and potentially browser_snapshot or browser_console_messages (mentioned in references/playwright-api.md), which could allow an attacker to influence agent behavior if the agent reads page content or logs.
Sanitization: There is no mention of sanitizing or filtering the page content or console logs before the agent processes them.

playwright-responsive-screenshots