microsoft-onedrive

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The uploadFile and replaceFile tools download binary data from a user-supplied fileUrl parameter to upload the content into OneDrive. Additionally, the getCopyStatus tool fetches data from a dynamically generated, pre-authenticated monitorUrl returned by the Microsoft Graph API.
  • [COMMAND_EXECUTION]: The entry-point script cli.js uses spawnSync to execute the connector's internal CLI logic (cli.ts or dist/cli.js) using the current Node.js binary. This is a standard architectural wrapper for providing a CLI interface to the skill's scripts.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources which could potentially contain malicious instructions for the agent.
  • Ingestion points: scripts/uploadFile.ts and scripts/replaceFile.ts download data from arbitrary external URLs. Tools like scripts/getItem.ts and scripts/listFolderItems.ts retrieve file names, metadata, and content from the user's OneDrive account.
  • Boundary markers: The skill returns all results within a structured JSON envelope ({ data, meta }), providing clear separation between tool output and system metadata.
  • Capability inventory: The skill possesses the capability to write, move, copy, share, and delete files on OneDrive. It can also perform network requests to user-defined URLs via the fileUrl parameter.
  • Sanitization: The skill employs encodeURIComponent when interpolating identifiers into API path parameters, providing protection against basic path traversal or injection in the API requests.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 10:23 PM
Security Audit — agent-trust-hub — microsoft-onedrive