document-converter

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and convert untrusted external documents into Markdown for consumption by the agent, which presents a surface for indirect prompt injection attacks. A document could contain hidden instructions intended to override agent logic or exfiltrate data.
  • Ingestion points: The convert_to_markdown tool and the markitdown CLI command take uri or URL inputs leading to external content.
  • Boundary markers: The instructions do not define boundary markers or delimiters to separate converted content from agent instructions, nor do they include warnings for the agent to ignore embedded instructions.
  • Capability inventory: The skill possesses capabilities for local file system access (file:///), network fetching (https://), and shell command execution via the CLI fallback.
  • Sanitization: There is no evidence of sanitization, filtering, or validation of the converted Markdown text before it is returned to the agent context.
  • [COMMAND_EXECUTION]: The skill provides instructions for a CLI fallback, which involves the direct execution of shell commands (markitdown document.pptx).
  • [EXTERNAL_DOWNLOADS]: The skill installs the markitdown package from Microsoft's official channels and the markitdown-mcp utility. It also supports downloading and processing documents from remote HTTP/HTTPS URLs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 07:41 AM