The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions for performing end-to-end web application testing using Playwright. It includes patterns for navigation, form handling, and multi-step verification.
[COMMAND_EXECUTION]: The skill utilizes Playwright MCP tools for browser control. These tools allow for navigation, clicking, typing, and taking screenshots, which are standard operations for the intended use case of automated testing.
[PROMPT_INJECTION]: The skill's primary function involves ingesting untrusted data from web pages via the browser_snapshot tool, which presents an inherent surface for indirect prompt injection.
Ingestion points: Web page accessibility trees and content are ingested using mcp__playwright__browser_snapshot and mcp__playwright__browser_wait_for.
Boundary markers: No specific delimiters or warnings for the agent to ignore embedded instructions within page content are provided in the instructions.
Capability inventory: The skill has capabilities to navigate, click, type, and upload files (mcp__playwright__browser_set_input_files) which could be misused if a malicious page successfully injected instructions.
Sanitization: There is no mention of filtering or sanitizing the content retrieved from the browser before the agent processes it.
Note: This surface is expected and necessary for the functional requirements of a browser automation skill.

dev-test-playwright