groq-data-handling
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data to perform PII redaction and auditing.\n
- Ingestion points: The messages array processed in
sanitizeMessagesandsafeCompletionwithinreferences/implementation.md.\n - Boundary markers: No explicit boundary markers or ignore-instructions are present in the snippets.\n
- Capability inventory: The skill utilizes the Groq SDK to execute AI inference requests via
groq.chat.completions.create.\n - Sanitization: The skill implements regex-based redaction for email addresses, phone numbers, SSNs, credit card numbers, and IP addresses.\n- [EXTERNAL_DOWNLOADS]: The skill references the Groq SDK for installation via npm.\n
- Evidence:
npm i groq-sdkmentioned inSKILL.md.\n - Note: This is an official package for the Groq inference service.
Audit Metadata