The Agent Skills Directory

[SAFE]: The skill provides implementation templates for Convex functions. It includes instructions to enforce authentication via ctx.auth.getUserIdentity() and authorization via ownership checks, which are security best practices.
[EXTERNAL_DOWNLOADS]: Mentions documentation and resources from the official Convex GitHub repository. As Convex is a well-known cloud service, these references are considered safe.
[PROMPT_INJECTION]: The skill provides an example of an action that processes user prompts through an LLM. This highlights an attack surface for indirect prompt injection, but it is presented as a standard implementation pattern rather than a malicious vector.

function-creator