The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): The skill explicitly instructs against hardcoding secrets (e.g., in references/cursor_rules_netlify.md) and recommends using encrypted environment variables or secrets management systems.
[Indirect Prompt Injection] (LOW): The skill utilizes external data ingestion by instructing the agent to fetch current documentation from official websites (e.g., Next.js, Vue, and Tailwind documentation) using tools like WebFetch. This is a functional requirement for the agent's expertise.
Ingestion points: WebFetch calls to nextjs.org, vuejs.org, and tailwindcss.com specified in persona files.
Boundary markers: None explicitly defined in the prompt templates.
Capability inventory: Access to Bash, Write, MultiEdit, and WebFetch tools.
Sanitization: The skill provides snippets and instructions for using DOMPurify to sanitize user-generated content in built applications.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): While the skill defines workflows involving build tools (Vite, Webpack, npx), it does not include any automated or suspicious remote script executions (e.g., piped bash scripts).

frontend-specialist