form-5498-sa

Pass

Audited by Gen Agent Trust Hub on Oct 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external sources, specifically tax forms (Form 5498-SA, Form 8889, and W-2), which may be controlled by third parties or contain untrusted content.
  • Ingestion points: Data is ingested through user input during form mapping in SKILL.md (Steps 3, 4, and 7) and the verification steps in filing.md.
  • Boundary markers: The instructions do not define specific delimiters or instructions for the agent to ignore or sanitize embedded text that could mimic instructions.
  • Capability inventory: The skill describes a 'browser-automation playbook' in filing.md for assisting with Form 1040-X amendments, which involves interactions with tax filing services.
  • Sanitization: There are no documented procedures for sanitizing or escaping the data strings extracted from the documents before they are used by the agent.
  • [SAFE]: The skill incorporates a robust 'Security rules' section in filing.md that strictly prohibits the persistence of sensitive identifiers like Social Security Numbers (SSN) and dates of birth, and mandates that the agent must obtain explicit user approval before any filing operations occur.
  • [SAFE]: External links are restricted to the official IRS website (irs.gov) and the vendor's own documentation on jupid.com, which are considered trusted or well-known vendor resources.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 4, 2026, 03:19 PM