gsc-cli

SUSPICIOUS: The skill's purpose is coherent, but its default trust model is weak. It fetches and executes unpinned third-party code from GitHub and forwards Google service-account credentials to it, creating a disproportionate supply-chain and credential-handling risk for an otherwise read-focused SEO helper.

No direct malicious behavior is implemented in this wrapper script (no eval/obfuscation, no explicit exfiltration, no destructive filesystem/network actions in the fragment). The dominant risk is supply-chain execution: the script can dynamically fetch and execute an MCP server using uvx from a default remote git source, and GSC_SOURCE can override that selection. It also injects a service-account credential path into the executed tooling and can enable destructive behavior via GSC_ALLOW_DESTRUCTIVE. This warrants pinning/verifying the fetched source/versions and hardening configuration controls; review the resolved mcp-gsc/mcp2cli code for behavior and trust guarantees.