okx-ai

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md routes free-form peer/user content into task sub-sessions and forwards it verbatim via okx-a2a session send / pending-decisions-v2 resolve-prompt, so outsider-authored chat text can end up as LLM-readable context through the “free-form task intent” path.

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

• Direct money access detected (high risk: 1.00). The skill explicitly orchestrates on-chain financial actions: it requires wallet preflight, routes wallet login/balance and token/contract safety checks to okx-agentic-wallet, and directs broadcasting raw transactions and staking/unstaking (evaluator staking) flows. These are specific crypto/blockchain transaction capabilities (signing/broadcasting/staking), not generic tooling — so it grants direct financial execution authority.