The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions utilize commands that download and immediately execute shell scripts from an external, unverified GitHub repository (iOfficeAI/OfficeCLI) by piping them to bash or iex. This bypasses security checks and executes arbitrary remote code with the current user's permissions.
[COMMAND_EXECUTION]: The skill's primary operations depend on executing a binary named officecli from an untrusted source. This binary has broad access to the agent's environment and the host file system.
[EXTERNAL_DOWNLOADS]: The skill fetches executable components and installation scripts from a repository belonging to an organization not recognized as a trusted vendor.
[PROMPT_INJECTION]: The skill processes untrusted content from Word, Excel, and PowerPoint files through the view and get commands, creating an indirect prompt injection surface. 1. Ingestion points: Office file content is ingested into the agent context via the officecli tool. 2. Boundary markers: Absent; there are no instructions for the agent to distinguish document data from operational commands. 3. Capability inventory: The skill has extensive shell execution and file-system modification capabilities. 4. Sanitization: Document content is processed and used directly in the provided script examples without any escaping or validation.

officecli-office-automation