office-to-markdown

Pass

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from Office documents (Word, Excel, PDF, PowerPoint) for conversion into AI-readable Markdown. This creates a surface for Indirect Prompt Injection where malicious text within a document could influence the behavior of an AI agent processing the output.
  • Ingestion points: The batch_convert and create_ai_corpus functions in SKILL.md use markitdown.convert() to read external files from the local file system.
  • Boundary markers: The provided code snippets do not include boundary markers or explicit instructions for the AI to ignore instructions embedded within the converted text.
  • Capability inventory: The scripts possess capabilities for file system read and write operations, as well as directory creation.
  • Sanitization: No content sanitization or validation of the extracted document text is performed before it is output to Markdown or JSON files.
  • [EXTERNAL_DOWNLOADS]: The skill recommends the installation of the markitdown package and the anthropic SDK. markitdown is an official open-source project from Microsoft, and anthropic is the official library for accessing Anthropic's AI services. These are considered trusted, well-known dependencies.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 31, 2026, 02:03 AM
Security Audit — agent-trust-hub — office-to-markdown