The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from both user inputs and external web sources to generate marketing deliverables and verify facts.
Ingestion points: Data enters the agent context through user-provided customer win metrics, social assets, and external information retrieved via web_search and web_fetch as defined in references/operating-principles.md.
Boundary markers: The skill provides instructions to maintain deterministic accuracy and explicitly flags assumptions using an Assumption: prefix to help differentiate between confirmed facts and AI-generated interpretations.
Capability inventory: The skill invokes specialized content tools (e.g., content_repurposer, case_study_generator) via a hosted MCP server and has the ability to browse the web.
Sanitization: Rule 4 in references/operating-principles.md establishes a four-tier source hierarchy to prioritize authoritative primary sources and reputable research, which serves as a validation mechanism for external data.
[EXTERNAL_DOWNLOADS]: The skill relies on external infrastructure and software packages associated with the vendor.
Evidence: The skill configuration targets a hosted MCP endpoint at https://craft-content.gtmhelix.com/mcp and references the npm package @shashwatgtmalpha/craft-content-mcp. These resources are consistent with the identified author identity 'shashwatgtm'.

helix-mcp-craft-content