humanizer-vi
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references several external URLs in
sources.mdfor style and terminology guidance. These include official Vietnamese government domains (chinhphu.vn, baochinhphu.vn, mst.gov.vn), academic repositories (aclanthology.org), and well-known technology services (Google, Mailchimp, Nielsen Norman Group). These references are for informational purposes and do not involve executable code or data exfiltration.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text as its primary function.\n - Ingestion points: Text submitted by users for editing and style humanization (referenced in
SKILL.mdandreferences/workflow.md).\n - Boundary markers: The instructions do not define explicit delimiters to isolate untrusted user input from the agent's system instructions.\n
- Capability inventory: The skill is purely instructional and does not demonstrate capabilities for file system modification, network exfiltration, or shell command execution.\n
- Sanitization: No explicit sanitization or filtering is described, as the focus is on linguistic transformation while preserving factual integrity.
Audit Metadata