hwp
Warn
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions and Python scripts perform dynamic installation of external packages from PyPI, including
pyhwp2md,python-hwpx,pyhwp, andsix. - [COMMAND_EXECUTION]: The
scripts/hwp_reader.pyfile executes external command-line binaries (hwp5txt,hwp5html) usingsubprocess.run. TheSKILL.mdinstructions also execute shell commands for file verification and content extraction. - [DYNAMIC_EXECUTION]: The skill implements a runtime detection and invocation logic for package managers (
uvx,pipx,pip) within shell scripts. Thescripts/hwp_reader.pyscript also invokespipat runtime to install dependencies, including the use of the--break-system-packagesflag. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted
.hwpand.hwpxfiles and prints their raw content into the agent's context, creating a surface for indirect prompt injection. - Ingestion points: File content extraction in
scripts/hwp_reader.pyviahwp5txtandHwpxDocument. - Boundary markers: None present to distinguish document content from system instructions.
- Capability inventory:
subprocess.runcalls inscripts/hwp_reader.pyfor package installation and tool execution. - Sanitization: None present for the extracted text.
Audit Metadata