The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from OneDrive, creating a potential surface for indirect prompt injection attacks where malicious content in files could influence the agent's behavior.
Ingestion points: The skill reads external data via ONE_DRIVE_SEARCH_ITEMS (metadata and content) and ONE_DRIVE_GET_ITEM (item metadata and children).
Boundary markers: The instructions do not define boundary markers or explicit warnings for the agent to ignore instructions embedded in the retrieved files.
Capability inventory: The agent has significant capabilities that could be targeted by an injection, including permission management (ONE_DRIVE_INVITE_USER_TO_DRIVE_ITEM), file sharing (ONE_DRIVE_CREATE_LINK), and file deletion (ONE_DRIVE_DELETE_ITEM).
Sanitization: There are no requirements or procedures defined for sanitizing or validating the content retrieved from OneDrive before processing.
[EXTERNAL_DOWNLOADS]: The skill instructs users to configure an external MCP (Model Context Protocol) server at https://rube.app/mcp and refers to toolkit documentation at composio.dev. While these are necessary for the skill's stated purpose of OneDrive automation, they involve routing agent interactions and data through external third-party infrastructure.

one-drive-automation