check-mcp-tool-drift

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This skill explicitly instructs posting complete MCP tool catalogs and metadata to a pinned external origin (including payment flow to a fixed recipient), creating a clear data-exfiltration and third-party gating vector that could be abused.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls the external runtime API https://bountyverdict-agent-production.mimirslab.workers.dev/api/mcp-drift and uses its JSON verdicts (e.g., UNCHANGED, BREAKING, BLOCK_CURRENT) as the gating decision that directly controls the agent's behavior, so this is a required runtime dependency that controls instructions.

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

• Direct money access detected (high risk: 1.00). The skill explicitly requires performing a crypto payment flow: it mandates sending the request to an x402 PAYMENT-REQUIRED flow, specifies blockchain network (eip155:8453), a canonical USDC contract and recipient address, an allowed amount, and instructs sending a payment signature to the production origin (with guidance about wallet secrets and signing). These are explicit crypto/payment instructions (wallet signing / transaction authorization), which constitute direct financial execution capability.