The Agent Skills Directory

[DYNAMIC_EXECUTION]: The documentation in interop.md mentions z.compile() (available in Zod 4.5+), which utilizes dynamic code generation for performance. The skill provides a security warning advising developers to check Content Security Policy (CSP) and schema features before adoption.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to facilitate the ingestion and processing of untrusted external data, which inherently provides an attack surface for indirect prompt injection if the processed data is later used to influence agent behavior.
Ingestion points: Explicitly targets "untrusted data" and "external data" as noted in SKILL.md and boundaries.md.
Boundary markers: The skill prescribes clear boundaries using safeParse and strictObject to separate untrusted input from the application core.
Capability inventory: Mentions dynamic execution capabilities (z.compile()) and extensive data transformation logic.
Sanitization: Focuses on Zod-based validation, custom refinements, and type-safe transforms to ensure data conforms to expected contracts.

zod