The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and executes the package @softeria/ms-365-mcp-server at runtime using npx -y. This dependency is not pinned to a specific version, making it susceptible to supply chain attacks if the package or the maintainer's account is compromised.
[COMMAND_EXECUTION]: The skill uses shell commands to invoke the MCP server through npx. While these are configured with specific tools, they still involve executing external binaries on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill has an indirect injection surface because it processes untrusted data (file names and folder listings) from external Microsoft 365 tenants while possessing write capabilities.
Ingestion points: External data is ingested via list-folder-files and get-drive-item in SKILL.md.
Boundary markers: The instructions include a manual confirmation step ("confirm the target path and filename with the operator before uploading") which acts as a human-in-the-loop boundary, but no technical delimiters are specified for the tool outputs.
Capability inventory: The skill possesses upload-file-content (write) and login capabilities as defined in SKILL.md.
Sanitization: No specific sanitization or filtering of the retrieved file metadata is described.

msgraph-visio