competition-jwt-claim-confusion

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires preserving raw headers/claims and embedded key material and producing a "replayable token-to-acceptance" sequence, which forces the LLM to handle and potentially output full JWTs or key material verbatim, creating an exfiltration risk.