ai-human-docx

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill minimizes its attack surface by restricting its own tool access to a single Bash command used to verify the presence of the uv Python tool.
  • [SAFE]: The conversion logic is isolated within a local Python script that uses only standard libraries, avoiding risks associated with external dependencies or remote code downloads.
  • [SAFE]: Strong security boundaries are enforced through both the AI instructions and the script logic, which explicitly check for and refuse file operations that escape the intended working directory.
  • [SAFE]: Protection against indirect prompt injection is provided by instructing the AI model to never read the raw content of the input markdown files, delegating the parsing entirely to the underlying script.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 02:43 PM