israeli-client-payment-chaser
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as invoice details and client communication history (Step 2 and Step 5) and interpolates this information into reminders and demand letters. There are no explicit boundary markers or instructions to sanitize this imported content, creating a surface where malicious instructions embedded in invoice notes or messages could potentially influence the agent's actions. * Ingestion points: Invoice data import (Step 2) and recording payment promises/communications (Step 5) in SKILL.md. * Boundary markers: Not present in the instructions for handling external text. * Capability inventory: The skill has access to the 'WebFetch' tool for network operations and 'Bash' for calculations, and it writes state to 'payment-chaser-data.json'. * Sanitization: No specific sanitization or escaping of external strings is mentioned before they are used in prompts.
- [EXTERNAL_DOWNLOADS]: The skill fetches statutory interest rate data from the Israeli government's official open data API (data.gov.il) and retrieves calendar information from HebCal. These references target well-known and official services for their intended functional purpose.
Audit Metadata