The Agent Skills Directory

The skill migrate-to-nextjs is designed to automate the migration of legacy websites to Next.js. The analysis reveals the following:

🔴 MEDIUM Findings: • UNVERIFIABLE_DEPENDENCY

Line 28: The skill instructs to npm install -g migent. While migent appears to be a tool related to Next.js migration, it is an external dependency installed globally via npm, and its source is not explicitly listed as a trusted GitHub organization. This introduces a risk of executing unverified code. • UNVERIFIABLE_DEPENDENCY
Line 70: The skill instructs to bun add -D @biomejs/biome. Biome is a reputable tool, but it is an external dependency installed via bun, and its source is not explicitly listed as a trusted GitHub organization. This introduces a risk of executing unverified code. • UNVERIFIABLE_DEPENDENCY
Line 94: The skill configures migent to be run via npx -y migent mcp. This relies on the previously flagged migent dependency. • UNVERIFIABLE_DEPENDENCY
Line 190: The skill conditionally instructs to bun add framer-motion. Framer Motion is a popular animation library, but it is an external dependency installed via bun, and its source is not explicitly listed as a trusted GitHub organization. This introduces a risk of executing unverified code. • UNVERIFIABLE_DEPENDENCY
Line 270: The skill instructs to bunx shadcn@latest init -y. Shadcn/ui is a popular component library, but it is an external dependency installed via bunx, and its source is not explicitly listed as a trusted GitHub organization. This introduces a risk of executing unverified code. • UNVERIFIABLE_DEPENDENCY
Line 280: The skill configures shadcn to be run via npx shadcn@latest mcp. This relies on the previously flagged shadcn dependency. • UNVERIFIABLE_DEPENDENCY
Line 287: The skill instructs to bunx shadcn@latest add ... -y. This relies on the previously flagged shadcn dependency.

🔵 LOW Findings: • TRUSTED_EXTERNAL_SOURCE

Line 20: The skill adds vercel-labs/agent-skills via npx skills add. Vercel Labs is a trusted GitHub organization. • TRUSTED_EXTERNAL_SOURCE
Line 21: The skill adds vercel-labs/next-skills via npx skills add. Vercel Labs is a trusted GitHub organization. • TRUSTED_EXTERNAL_SOURCE
Line 22: The skill adds vercel-labs/next-skills via npx skills add. Vercel Labs is a trusted GitHub organization. • TRUSTED_EXTERNAL_SOURCE
Line 23: The skill adds vercel-labs/agent-skills via npx skills add. Vercel Labs is a trusted GitHub organization. • TRUSTED_EXTERNAL_SOURCE
Line 58: The skill uses bunx create-next-app@latest. create-next-app is an official tool from Vercel/Next.js, which is a trusted source.

ℹ️ INFO Findings: • INDIRECT_PROMPT_INJECTION_RISK: The skill processes a significant amount of external and potentially untrusted content from the user's legacy website, including file system content (e.g., package.json, index.html, .js files, sitemap.xml) and live site content via ir_capture. Malicious instructions embedded within this content could theoretically be interpreted by the AI, leading to unintended actions. Users should be aware of this inherent risk when processing untrusted external data.

No prompt injection attempts, data exfiltration to non-whitelisted domains, obfuscation, privilege escalation, persistence mechanisms, or time-delayed attacks were detected within the skill's instructions themselves.