The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation provides commands to download and immediately execute shell scripts from an external GitHub repository, a practice that enables arbitrary remote code execution on the host system. Evidence: 'curl -fsSL https://raw.githubusercontent.com/iOfficeAI/OfficeCLI/main/install.sh | bash' and 'irm https://raw.githubusercontent.com/iOfficeAI/OfficeCLI/main/install.ps1 | iex' in SKILL.md.
[EXTERNAL_DOWNLOADS]: The installation and update mechanisms fetch scripts and binaries from the iOfficeAI GitHub organization, which is an external source for executable content.
[COMMAND_EXECUTION]: The skill instructions involve executing a command-line tool (officecli) that performs sensitive file system operations on user documents. Furthermore, the tool's documentation indicates a background auto-update feature, which functions as a persistence mechanism by executing code periodically without direct user invocation.
[PROMPT_INJECTION]: The skill is designed to ingest and process external Office documents (.docx, .xlsx, .pptx), which provides an attack surface for indirect prompt injection where malicious instructions embedded in documents could influence agent behavior. Ingestion points: 'officecli view ' and 'officecli get ' in SKILL.md. Boundary markers: None identified. Capability inventory: The skill can read, write, and batch-process OpenXML document parts. Sanitization: No specific sanitization, escaping, or validation of external document content is described before the data is processed or interpolated.

officeCLI