payroll-tax-analysis

Warn

Audited by Gen Agent Trust Hub on Feb 14, 2026

Risk Level: MEDIUMPROMPT_INJECTIONNO_CODE
Full Analysis
  • [Indirect Prompt Injection] (MEDIUM): The skill is designed to process untrusted external data, including contractor contract details and entity relationship information, to determine tax compliance.
  • Ingestion points: Wage data, contractor contracts, and entity grouping assessments entered into the agent context.
  • Boundary markers: Not present in the skill definition or instructions.
  • Capability inventory: Performs multi-state tax calculations and references an external logic engine at lib/analysis/payroll-tax-engine.ts.
  • Sanitization: No input validation, escaping, or sanitization mechanisms are described for the data being analyzed.
  • [Missing Logic] (MEDIUM): The core execution logic is contained in an external file (lib/analysis/payroll-tax-engine.ts) which was not provided. In an 'assume-malicious' posture, this hidden logic cannot be verified for safe file access, network calls, or command execution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 14, 2026, 03:58 PM