The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Dropbox files and shared links, creating a vulnerability surface where data could contain instructions for the agent.
Ingestion points: DROPBOX_READ_FILE, DROPBOX_GET_SHARED_LINK_FILE, and DROPBOX_EXPORT_FILE as described in SKILL.md.
Boundary markers: Absent. The instructions do not specify delimiters or prompts to ignore instructions within the retrieved file content.
Capability inventory: The skill possesses significant write and management capabilities including DROPBOX_UPLOAD_FILE, DROPBOX_DELETE_FILE_OR_FOLDER, and DROPBOX_MOVE_FILE_OR_FOLDER which could be targeted by an injection attack.
Sanitization: Absent. The skill instructions do not mention sanitizing or validating external content beyond decoding base64 transport formats.
[EXTERNAL_DOWNLOADS]: The skill references and interacts with external infrastructure to provide its functionality.
Mentions adding the MCP server at https://rube.app/mcp as part of the setup process.
Includes the DROPBOX_SAVE_URL tool which allows the agent to fetch data from any public URL into the Dropbox environment.

dropbox-automation