The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs the user to install the Membrane CLI tool (@membranehq/cli) globally from the official npm registry. This is a vendor-owned resource for managing the skill's environment.
[COMMAND_EXECUTION]: The skill relies on shell commands through the membrane CLI to perform operations such as authentication (membrane login), connection management (membrane connect), and action execution (membrane action run).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it ingests and processes untrusted data from an external file storage service.
Ingestion points: Data is brought into the agent context via actions like list-folder-contents, get-item-by-path, and search-files which return file names, paths, and metadata.
Boundary markers: The instructions lack explicit directives for the agent to use XML-style tags or other delimiters to isolate untrusted external content from the system prompt.
Capability inventory: The skill possesses capabilities to modify the user's environment, including deleting items (delete-item), uploading files (upload-small-file), and renaming items (rename-item).
Sanitization: There is no evidence of data sanitization or validation logic to filter out potentially malicious instructions embedded in OneDrive file data.

onedrive