The Agent Skills Directory

[Metadata Poisoning] (MEDIUM): The skill provides false information by claiming the official @sinclair/typebox package is deprecated and instructing users to use typebox instead. In practice, @sinclair/typebox is the active and recommended package. Such misleading instructions can lead to dependency confusion or the use of non-standard packages.
[Dynamic Execution] (MEDIUM): The skill demonstrates the use of Compile() with raw TypeScript syntax strings. This involves runtime parsing and code generation (often utilizing new Function() for performance). If an attacker can influence the string passed to Compile(), it may lead to arbitrary code execution within the agent's runtime environment.
[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process schemas from various external libraries (Zod, Valibot). While it functions as a validator, this ingestion surface creates a potential vector for injection attacks if the schemas themselves contain malicious instructions designed to subvert agent logic.

typebox