form-2553
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, including legal names, addresses, and Social Security Numbers (SSNs), to generate official tax documentation. This ingestion of external data constitutes an attack surface for indirect prompt injection.
- Ingestion points: Shareholder data collected in SKILL.md prerequisites and Step 4 of the workflow.
- Boundary markers: The instructions do not define specific delimiters (e.g., XML tags or triple backticks) to separate user data from instructions.
- Capability inventory: The skill generates draft tax forms and provides filing instructions, including fax transmission via external services.
- Sanitization: The skill includes robust sanitization requirements, instructing the agent to never store PII in logs or transcripts beyond the active session and to redact SSNs to the last four digits in any persistent records.
- [SAFE]: The skill uses authoritative sources for its primary functionality. It references official IRS forms and instructions (irs.gov) and uses well-known government wage data (bls.gov) for financial analysis. All references to the author's own domain (jupid.com) are documented and consistent with the vendor identity.
Audit Metadata