integration-tests

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

• Third-party content exposure detected (high risk: 0.80). The workflow for “Integration Tests” explicitly includes “real web requests and LLM calls” (outsider web content / outsider LLM outputs) at runtime, so free-form text from external sources can be ingested into the agent’s LLM context indirectly via mock_web()/mock_llm() being replaced by real calls.

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

• Direct money access detected (high risk: 1.00). The skill explicitly exposes blockchain transaction execution: examples call .transact() to submit transactions and wait for consensus, it discusses transaction receipts and lifecycle, and the configuration includes account private keys and mentions testnets requiring funded accounts. These are specific crypto/blockchain operations (signing/submitting transactions), which constitute Direct Financial Execution authority.