The Agent Skills Directory

[SAFE]: The skill's instructions are focused on legitimate visual testing workflows, such as baseline management and cross-viewport validation. No unauthorized data exfiltration, hardcoded credentials, or malicious command execution patterns were identified.
[PROMPT_INJECTION]: The skill operates by navigating to and interacting with external URLs provided at runtime. This creates a surface for indirect prompt injection, as malicious instructions embedded on a target website could attempt to influence the agent's behavior during the testing process. This is a common characteristic of web-browsing skills and is not indicative of malicious intent within the skill itself.
Ingestion points: Target URLs provided in the test plan and page content read via browser tools like read_page.
Boundary markers: The instructions do not define specific markers to separate external content from agent instructions.
Capability inventory: The skill uses tools for browser navigation (navigate), viewport resizing (resize_window), screenshot capture (computer), and local storage (memory-mcp).
Sanitization: No specific sanitization or filtering logic is described for the data retrieved from external sources.

visual-testing