dunning-cohort-builder
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external API, creating a potential surface for indirect prompt injection attacks.
- Ingestion points: Fetches subscription records from
ZohoBilling_get_subscriptions_dunning_reportinSKILL.md(Step 1). - Boundary markers: Absent; data fields like
gateway_error_messageare processed and passed downstream without specific delimiters to distinguish between data and instructions. - Capability inventory: The skill aggregates and summarizes data, then passes a structured JSON cohort to the
dunning-action-plannertool. - Sanitization: Includes basic
lowercaseandtrimoperations on gateway codes and messages in Step 2, but does not implement advanced filtering for natural language instructions. - [SAFE]: The tool calls are limited to the vendor's own infrastructure (Zoho Billing), and the logic consists of standard data normalization and arithmetic scoring. No obfuscation, persistence, or privilege escalation patterns were found.
Audit Metadata