The Agent Skills Directory

[SAFE]: The skill consists entirely of Markdown documentation and reference files. No executable scripts or active code components are included in the skill package.
[NO_CODE]: The skill contains no executable code files (.js, .ts, .py, etc.) and is purely instructional.
[PROMPT_INJECTION]: The documentation describes library features like ctx.sample() and ctx.elicit() which handle LLM and user input. This represents a surface for indirect prompt injection in applications built with the library. However, the skill provides extensive mitigation guidance, including the mandatory use of Zod schemas for input validation and explicit warnings against processing untrusted data without sanitization.
[REMOTE_CODE_EXECUTION]: Reference guides include an example command for Supabase deployment using curl | bash from url.mcp-use.com. As this targets the vendor's own domain and is documented for deployment purposes, it is treated as a safe vendor-resource finding.
[EXTERNAL_DOWNLOADS]: The documentation references several well-known and trusted dependencies, including zod, jose, better-sqlite3, and pg. It also describes using OAuth providers like Auth0, WorkOS, and Supabase, which involve standard external network interactions.
[COMMAND_EXECUTION]: Examples within the guides demonstrate the use of CLI tools such as mcp-use, tsx, and uv for development and server composition (proxying) tasks. These are documented as standard developer workflows.

build-mcp-use-server