The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill introduces an indirect prompt injection surface because it reads and processes untrusted inputs from third-party recipients and subscriber records without specifying isolation safeguards.
Ingestion points: In references/provider-mechanics.md, the skill monitors incoming data using gmail_messages_list to fetch unread email replies, and ingests untrusted fields via beehiiv_create_subscription using arbitrary custom_fields values.
Boundary markers: No explicit delimiters, data-typing rules, or instructional barriers are provided to separate the incoming email body text or subscriber inputs from the agent's core instructions.
Capability inventory: The skill has extensive write capabilities, including sending outbound emails (gmail_messages_send, resend_send_email, klaviyo_send_campaign) and configuring multi-step email automations (resend_create_automation).
Sanitization: There are no provisions for escaping content, sanitizing HTML, or validating input payloads prior to the data being parsed by the agent workflow.

email-lifecycle