The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes workflows for processing external webhook events (e.g., bounces and complaints) from email providers. This data ingestion path constitutes an indirect prompt injection surface where malformed external data could theoretically influence the agent's logic. The skill mitigates this by instructing the agent to verify webhook signatures.
Ingestion points: Webhook endpoint for processing delivery events as described in RELIABILITY.md.
Boundary markers: The skill explicitly instructs the agent to treat signature verification as a security boundary before processing any webhook data.
Capability inventory: The logic involves database operations to update suppression lists and conditional send logic within LISTS-AND-CAPTURE.md and RELIABILITY.md.
Sanitization: The instructions mandate verifying webhook signatures (specifically mentioning svix-id headers) and validating event identities to ensure data integrity.
[NO_CODE]: The skill package is composed entirely of Markdown files providing documentation and instructional guidance for the agent. No executable scripts, binaries, or active code components are included.

email-practices