ai-ecommerce-expert-gpt-image-2-ecommerce-image

SUSPICIOUS. The skill’s purpose and data flow broadly match an ecommerce image-generation integration, but it relies on a newly observed, unpinned npm CLI and forwards the MCP token plus user assets through that external tool. The domain relationship looks plausible, which argues against confirmed malware, yet the install trust and credential-forwarding footprint are disproportionate enough to treat it as medium-high risk.

No direct evidence of malware (no obfuscated payloads, no command execution beyond running a known external package, no direct data theft routines). However, the module dynamically executes an npm dependency via npx using the unpinned '@latest' tag and forwards MCP_TOKEN and API_URL to that dependency. This creates a significant supply-chain and credential-handling risk: if the npm package is compromised or changes behavior, it can access the token and act on the user's behalf.

This snippet does not contain overt malware, but it creates a meaningful supply-chain execution risk: it uses npx to run an unpinned '@latest' package (non-deterministic, changeable at runtime) and it passes a sensitive token via environment variables to that external dependency. If the dependency or its supply chain is compromised, the attacker would likely gain the token and the ability to perform arbitrary actions on the host and/or make authenticated API calls. Pin versions, verify integrity, minimize token exposure (scoping/least privilege), and restrict outbound network access for the spawned process.