skills/jkf87/hwpx-skill/hwpx/Gen Agent Trust Hub

hwpx

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill orchestrates its various components using subprocess.run to call internal Python scripts and a vendored Node.js runtime (rhwp_convert.mjs). This is used for document assembly, namespace fixing, and HWP to HWPX conversion.
  • [PRIVILEGE_ESCALATION]: Automated conversion on Windows utilizes PowerShell with the -ExecutionPolicy Bypass flag to run the included convert_hwp_hancom.ps1 script. This is a common practice for enabling local automation scripts within a restricted environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it processes untrusted content from HTML and Markdown files.
  • Ingestion points: scripts/html2hwpx.py (HTML input), scripts/yoyak.py, and scripts/geomto.py (Markdown inputs).
  • Boundary markers: Absent; the tools interpret the structure of the provided data to build the document parts.
  • Capability inventory: The skill can execute local commands via subprocess.run and control the Hancom Office application via Windows COM automation (win32com.client.DispatchEx).
  • Sanitization: The skill employs xml_escape in scripts/hwpx_helpers.py to prevent XML injection, though it does not specifically filter for natural language instructions embedded in the input.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 05:52 AM