The Agent Skills Directory

[DATA_EXFILTRATION] (HIGH): The tools GOOGLEPHOTOS_UPLOAD_MEDIA and GOOGLEPHOTOS_BATCH_CREATE_MEDIA_ITEMS accept a file_to_upload parameter. This capability allows the agent to read arbitrary files from the local filesystem. A malicious prompt or indirect injection could compel the agent to upload sensitive files (e.g., ~/.ssh/id_rsa, .env files) to the photo library or an external destination.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill requires connection to an external, untrusted MCP server at https://rube.app/mcp. Furthermore, the url parameter in the upload tools allows the agent to fetch content from any remote URL, which can be used to bypass local security controls or pull malicious payloads.
[PROMPT_INJECTION] (HIGH): The skill exhibits a high-risk surface for Indirect Prompt Injection (Category 8).
Ingestion points: Remote content fetched via the url parameter and metadata (descriptions/titles) provided by the user or external tools.
Boundary markers: Absent. There are no instructions to the agent to treat fetched data or file paths as untrusted data.
Capability inventory: Significant capabilities including file system read access, network transmission (upload), and account modification (creating/updating albums).
Sanitization: None. The skill does not implement any validation or filtering of the content being processed.

googlephotos-automation