The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it retrieves and processes content from Microsoft SharePoint (e.g., file metadata, list item values, and folder contents) that could be controlled by an external actor. If an agent reads malicious instructions embedded in SharePoint items, it could influence downstream behavior. This is a common characteristic of data-retrieval tools.
Ingestion points: scripts/getItem.ts, scripts/findListItems.ts, scripts/findFiles.ts return data from the SharePoint API.
Boundary markers: The skill does not implement specific delimiters for the retrieved data, relying on the platform's orchestration layer.
Capability inventory: The skill includes powerful write capabilities such as deleteItem, uploadFile, and updateListItem which could be targets for an injection attack.
Sanitization: Data returned from the API is passed back to the agent in structured JSON format without specific content filtering.
[EXTERNAL_DOWNLOADS]: The uploadFile and replaceFile tools are designed to download data from a user-provided fileUrl and upload it to SharePoint. This network activity is the primary, documented function of these tools and is initiated via user or agent intent.
[COMMAND_EXECUTION]: The cli.js entry point uses node:child_process.spawnSync to delegate to the script dispatcher. This is a standard part of the skill's infrastructure for standalone or MCP execution and is not exposed as a general-purpose command execution tool to the agent.

microsoft-sharepoint