contract-to-billing

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the airwallex CLI binary to perform billing operations including the creation of invoices, products, and subscriptions. This is a legitimate use of vendor-provided tools for the stated purpose.
  • [EXTERNAL_DOWNLOADS]: The skill references standard Python libraries, specifically pdfplumber and python-docx, for document parsing. It also describes a capability to fetch files from remote URLs if shell access is available to the agent.
  • [PROMPT_INJECTION]: The skill processes untrusted data from user-provided documents, establishing an indirect prompt injection surface. This is addressed by a mandatory evidence chain: 1. Ingestion points: documents read in Phase 1, Step 1; 2. Boundary markers: the requirement to flag uncertainty with [?] and present five validation tables; 3. Capability inventory: CLI access for billing operations and file system reads; 4. Sanitization: strict human-in-the-loop verification of all payloads and explicit approval for finalization and writes.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 09:07 AM
Security Audit — agent-trust-hub — contract-to-billing