adcp-review

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly requires fetching https://docs.adcontextprotocol.org/llms.txt "fresh, for every review" at runtime and to treat the fetched live docs as the authoritative source for review evidence, so remote content will directly influence the agent's instructions/evidence selection.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill is specifically about the AdCP advertising protocol and explicitly references protocol task APIs that create/modify media buys (e.g. "create_media_buy") and notes that "AdCP mutations commit advertising spend." Those are domain-specific financial operations that can move advertising budgets (i.e., money), so this skill exposes or documents explicit APIs that perform financial execution for ad spend.