The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The skill processes untrusted data (file names, metadata) from the Uploadcare API via the UPLOADCARE_LIST_FILES and UPLOADCARE_GET_FILE_INFO tools.
Ingestion points: File metadata, specifically file names and descriptions retrieved during listing or inspection (SKILL.md).
Boundary markers: None present. There are no instructions for the agent to ignore or delimit instructions found within file metadata.
Capability inventory: The skill can store files (UPLOADCARE_STORE_FILE) and generate download URLs, providing a path for the agent to act on malicious instructions embedded in file names.
Sanitization: No evidence of metadata sanitization or validation before presenting to the agent.
[External Downloads] (MEDIUM): The skill requires the installation of the rube MCP server from https://rube.app/mcp. This is an external remote dependency from an unverified source not included in the Trusted External Sources list.

Uploadcare Automation