hwpx

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection (LOW): The skill provides the capability to ingest, read, and modify existing HWPX files as demonstrated in the evaluation scenarios.
  • Ingestion points: The agent is prompted to open and modify existing files (e.g., assets/report-template.hwpx in evals/evals.json).
  • Boundary markers: Absent. There are no explicit instructions or delimiters used to separate user instructions from potentially malicious content embedded within the HWPX document body.
  • Capability inventory: The skill possesses file write capabilities and XML manipulation via scripts/fix_namespaces.py (using zipfile and os.replace).
  • Sanitization: No sanitization or validation of the text content within the HWPX files is performed before the agent processes or edits them.
  • Unverifiable Dependencies (SAFE): The documentation references the python-hwpx library. While this is an external dependency, it is used for its primary intended purpose and no malicious installation or remote execution patterns were found.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:27 PM