The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines patterns for workflows that ingest and process data from untrusted external events, presenting a surface for indirect prompt injection.
Ingestion points: Data from external events (e.g., front/inbound_received) is ingested into the workflow context through the event.data object in SKILL.md.
Boundary markers: The provided code snippets do not include explicit delimiters, tags, or instructions to the agent to disregard embedded commands within the processed data.
Capability inventory: The workflow logic demonstrates high-privilege capabilities such as executing actions via executeAction and creating drafts through external APIs.
Sanitization: There is no evidence of sanitization, escaping, or validation of the context variable before it is interpolated into the agent's prompt via the buildPrompt(context) function call.

inngest-workflow