gzh-design
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill demonstrates a strong security posture by including automated scripts to validate and lint output for compliance and sanitization.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (documents and text) to generate article layouts, representing a potential attack surface. However, it implements multiple mitigation strategies:
- Ingestion points: The skill ingests data via .md, .docx, .pdf, and .txt files.
- Boundary markers: A mandatory structure confirmation step is required by the workflow before layout generation begins, allowing for human-in-the-loop verification.
- Capability inventory: The skill performs local file reading and writing, and executes contained Python scripts (extract_docx.py, validate_gzh_html.py) for format normalization and validation.
- Sanitization: A mandatory validation step using validate_gzh_html.py enforces a strict whitelist of HTML tags and attributes (blocking scripts, styles, etc.), effectively stripping potentially harmful content before delivery.
Audit Metadata