The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill processes untrusted data from an external source (Cloudinary assets, tags, and folder names). Ingestion points: CLOUDINARY_SEARCH_FOLDERS, CLOUDINARY_GET_RESOURCES_BY_ASSET_FOLDER, and CLOUDINARY_GET_RESOURCE_BY_PUBLIC_ID fetch data that can be influenced by anyone with write access to the Cloudinary account. Capability inventory: The skill can create webhooks (CLOUDINARY_CREATE_TRIGGER), modify asset metadata (CLOUDINARY_EXPLICIT_RESOURCE), and retrieve account configuration (CLOUDINARY_GET_CONFIG). Sanitization: No evidence of sanitization or boundary markers for the data fetched from Cloudinary before it is used in subsequent agent reasoning.
[External Downloads] (MEDIUM): The skill requires the rube MCP server from https://rube.app/mcp. This is an unverified external source not included in the Trusted Sources list.
[Data Exposure] (MEDIUM): CLOUDINARY_GET_CONFIG allows the agent to fetch environment configuration details. Depending on how Cloudinary is configured, this might expose sensitive internal settings or metadata.
[Data Exfiltration] (LOW): CLOUDINARY_CREATE_TRIGGER allows the agent to set up webhooks to arbitrary URLs. While a core feature, it provides a native mechanism for data exfiltration if the agent is subverted via prompt injection.

Cloudinary Automation