twitter-automation

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). These URLs point to an untrusted third-party domain (twitr.sh) that appears to be a typosquat/third‑party service for X/Twitter and is not an official vendor — this makes it higher risk for abuse (unexpected charges, account compromise, or malware distribution) even though the links shown are API/docs rather than direct executables.

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

• Third-party content exposure detected (medium risk: 0.30). At runtime, the workflow’s monitors and webhooks (e.g., x_monitor events like tweet.mention and webhook payloads) ingest outsider-authored tweet text and other event content to drive x_compose/writing actions.

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

• Direct money access detected (high risk: 1.00). The skill explicitly describes an integrated payment flow for API calls: the first POST returns an HTTP 402 with the exact price and "your wallet pays in USDC" (payment/retry behavior). It provides an AgentCash MCP client snippet that "handles the 402, payment, and retry automatically" and instructs the agent to surface the 402 amount before signing. Monitors are described as prepaid (hours) and specify payment rails (Pay on Base x402 or Tempo MPP). Action endpoints list per-action prices. These are concrete wallet/crypto payment integrations (USDC, x402/MPP, AgentCash) that allow the agent to trigger on-chain/off-chain payments — i.e., direct financial execution capability.