The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data provided by the Klaviyo API (e.g., campaign content, flow definitions, and subscriber profiles), which could theoretically contain malicious instructions.
Ingestion points: Data is retrieved through the klaviyo CLI commands such as list-campaigns, list-flows, and generic call-tool read operations specified in SKILL.md.
Boundary markers: The instructions mandate that the agent "read back the audience, message content, and schedule before send_campaign", which serves as a manual verification step for the user, although no specific prompt delimiters are defined.
Capability inventory: The skill possesses capabilities for creating, updating, and deleting sensitive marketing data; however, these are strictly governed by the manifest.yaml which sets ask as the default for all write and delete operations.
Sanitization: There are no explicit requirements in the instructions for the agent to sanitize or filter input received from the external Klaviyo service before processing.

klaviyo