inngest-agent-evals
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends installing the 'inngest' package from the official npm registry. This is a well-known service for workflow orchestration and the source is a standard, reputable registry.
- [PROMPT_INJECTION]: The skill describes patterns for building workflows that ingest and process external event data, constituting an indirect prompt injection surface.
- Ingestion points: Untrusted data enters the agent context via 'event.data' in Inngest function triggers as documented in the scoring and experiment sections.
- Boundary markers: Implementation examples show direct interpolation of event data into strings (e.g., in 'if' conditions for event matching) without explicit delimiters or warnings to ignore embedded instructions.
- Capability inventory: The documented patterns leverage powerful SDK capabilities including 'step.run' for execution, 'step.score' for evaluating outcomes, and 'inngest.send' for event emission.
- Sanitization: The provided code snippets do not demonstrate validation, escaping, or filtering of the incoming event data before it is used in logic or interpolation.
Audit Metadata