social-intel

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

• Third-party content exposure detected (medium risk: 0.30). In the social-intel workflow, the runtime takes the user-provided --topic text and sends it as the POST body.query to two outsider-authored web search endpoints (Exa https://api.exa.ai/search and Tavily https://x402.tavily.com/search), which return page text/snippets that the agent then fuses into a cited brief.

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

• Direct money access detected (high risk: 1.00). The skill explicitly compiles its steps into selat-pay calls and describes real payment settlement: per-step caps and live quotes, a machine-readable payment "manifest", resolving settlement from a funded Circle Gateway balance, and a paid confirm flag (--pay) that "settles real 200s". These are explicit payment/settlement operations (sending money to settle API calls), so the skill grants direct financial execution capability.