ap2-risk-signals

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

• Third-party content exposure detected (high risk: 1.00). The skill's "Before writing code" steps explicitly require fetching live docs from public URLs (https://ap2-protocol.org/specification/ and https://ap2-protocol.org/topics/privacy-and-security/) and performing web searches (e.g., ap2 protocol risk signals, site:github.com google-agentic-commerce AP2 risk), meaning the agent will ingest untrusted third‑party web content that it must read to decide payload structure and implementation behavior, enabling indirect prompt injection.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The SKILL instructs fetching live docs at https://ap2-protocol.org/specification/ and https://ap2-protocol.org/topics/privacy-and-security/ at runtime to obtain the "exact risk payload" and guidance, so remote content would directly control agent instructions and is treated as a required dependency.