The Agent Skills Directory

[SAFE]: The skill consists entirely of documentation and instructional content in Markdown format. It does not contain executable scripts, automated package installations, or privilege escalation commands. The instructions align with established UX best practices and aim to improve interface accessibility and usability.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external and untrusted data, such as reviewing user-provided URLs or frontend code for usability and accessibility issues. While this creates a surface for indirect prompt injection (e.g., hidden instructions within the analyzed code), the skill itself does not provide dangerous capabilities like network exfiltration or file system writing that could be abused. The risk is inherent to the agent's core task of content analysis and is not introduced by malicious intent within the skill's instructions.
Ingestion points: SKILL.md instructs the agent to review screenshots, URLs, and code provided by the user.
Boundary markers: The instructions do not define specific delimiters for untrusted content, relying on the agent's general context management.
Capability inventory: The skill performs analysis and provides critiques; it does not request or use capabilities like network access, file writing, or subprocess execution.
Sanitization: No specific sanitization logic is provided as the skill is purely instructional text.

ux-designer