The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The installation steps involve cloning a repository from https://github.com/ArcadeAI/blueprint-mcp.git and installing the arcade-mcp package via pip. While these are central to the skill's functionality, they originate from a third-party source.
[COMMAND_EXECUTION]: The skill instructions include running arcade-mcp deploy after cloning the external repository, which involves executing code from the downloaded source.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by integrating data from several sources into prompts without visible sanitization.
Ingestion points: Reads local files, GitHub repository structures, HubSpot deal information, and Google Drive document content.
Boundary markers: The provided prompt examples do not include delimiters or instructions to ignore potential commands within the interpolated data.
Capability inventory: The skill has capabilities to write files to the local workspace (open with wb mode) and execute deployment commands.
Sanitization: There is no evidence of content validation or escaping for the data fetched from external APIs before it is used in diagram generation prompts.

blueprint-mcp-diagram-generation