The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents how to build workflows that process external events, which is an inherent attack surface for indirect prompt injection. Evidence: (1) Ingestion points: inngest.createFunction handlers receive untrusted event data (documented in references/00-overview.md). (2) Boundary markers: The documentation advocates for Zod schema validation to define strict data boundaries (documented in references/12-typescript-patterns.md). (3) Capability inventory: Workflows can perform database writes and API calls using step.run (documented in references/03-steps-workflows.md). (4) Sanitization: Best practices for data validation and type-safe event handling are covered throughout the references.
[EXTERNAL_DOWNLOADS]: The maintenance script scripts/check-updates.py performs a network request to registry.npmjs.org to retrieve the latest version of the inngest package. This is a standard maintenance task using a well-known service.

inngest