The Agent Skills Directory

[SAFE]: The skill provides high-quality documentation and patterns for durable execution. It specifically includes a 'Sharp Edges' section that warns developers about security and reliability risks, such as missing idempotency keys in payment flows and non-deterministic behavior in workflows.
[PROMPT_INJECTION]: The 'Inngest with AI Orchestration' code example demonstrates a common pattern where untrusted event data (event.data.task) is interpolated into an LLM prompt. While presented as a standard implementation example for orchestration, it constitutes a surface for indirect prompt injection. Production implementations should include sanitization and boundary delimiters.

workflow-automation