claudeclaw-group-agent
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The system implements seven safety guards (WA-SAFE-01 to WA-SAFE-07) that manage rate limits, message deduplication, and daily budgets to prevent resource exhaustion.
- [SAFE]: Authentication is managed securely via environment variables and per-tenant BYOK (Bring Your Own Key) logic, ensuring no sensitive credentials are hardcoded.
- [SAFE]: The architecture utilizes standard, well-known libraries such as the official @anthropic-ai/sdk for direct API communication.
- [SAFE]: Indirect prompt injection risks from untrusted WhatsApp input are addressed through documented input guardrails, identity anchoring, and jailbreak detection mechanisms.
Audit Metadata