The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and interpret data from several external, untrusted sources, which creates a potential surface for indirect prompt injection attacks.
Ingestion points: The agent is instructed to ingest data from CRM platforms (HubSpot, Salesforce), advertising networks (Google Ads, Meta, LinkedIn), and self-reported user surveys as detailed in SKILL.md and references/by-business-type.md.
Boundary markers: The instructions do not define explicit boundary markers or delimiters for the agent to use when reading this external data, nor are there instructions to ignore potentially malicious embedded content.
Capability inventory: The skill utilizes various external tools through the Model Context Protocol (MCP), including PostHog, HubSpot, Salesforce, and Segment, which provide read and write access to external systems.
Sanitization: Although the documentation provides implementation guidance for developers to normalize and protect data in the resulting system, it lacks specific instructions for the agent to sanitize or filter input during its own reasoning and data processing tasks.
[EXTERNAL_DOWNLOADS]: The skill points to external documentation and tool registries located on a third-party GitHub repository.
Evidence: In SKILL.md, the "Tool Integrations" section links to external resources at https://github.com/coreyhaines31/marketingskills/blob/main/tools/REGISTRY.md and related integration guides hosted in the same repository.

attribution