a2a-protocol

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt includes examples that embed credentials verbatim in requests (e.g., push_notification_config.authentication.credentials: "webhook-token-here") and otherwise shows passing API keys/tokens in request bodies or headers, which would require the model to output secret strings directly.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly instructs clients to fetch and parse public agent cards from well-known URIs (e.g., /.well-known/agent-card.json as shown in SKILL.md and references/agent-card.md) and to rely on fields like capabilities, security_schemes, and skills/examples to decide streaming, authentication, endpoints, and workflow, so untrusted third-party agent-card JSON can materially influence agent actions.