manage-vietnam-accounting
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves processing external e-invoice XML data which represents an inherent attack surface for instructions embedded within the data.
- Ingestion points: External e-invoice XML files are parsed per instructions in
SKILL.mdand thereferences/e-invoice-risk-playbook.mdguide. - Boundary markers: The skill enforces an
AI-ACCOUNTING-GUARDRAILrequiring human Chief Accountant review for all generated entries and filings. It also mandates that restricted data (PII, credentials) be kept out of prompts and agent memory. - Capability inventory: The skill is configured with broad capabilities including
write_fileandexecute_command(e.g., for querying government tax portals). - Sanitization: The skill explicitly mandates the use of hardened XML parsers with external entity resolution disabled to defend against XXE attacks.
Audit Metadata