zod-best-practices
Warn
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides deceptive metadata by claiming it was authored against 'official Zod documentation version 4.3.6'. As Zod v4.3.6 does not exist, this constitutes metadata poisoning by establishing a fictional basis for the agent's technical output.
- [PROMPT_INJECTION]: The instructions and reference documentation direct the agent to override its internal knowledge of the Zod library with hallucinated 'modern' APIs (e.g.,
z.email(),z.uuid(),z.strictObject(),z.treeifyError()). By framing actual Zod 3.x APIs as 'stale' or 'deprecated', the skill causes the agent to generate non-functional code and provide misleading advice.
Audit Metadata