goat

SUSPICIOUS: the skill is broadly consistent with an onchain agent toolkit, and its installs are mostly standard registry-based packages, but it grants an AI agent high-impact financial capabilities using raw wallet credentials. The main risk is autonomous asset movement and prompt-driven transaction execution, not clear malware or covert exfiltration.

No clear indicators of classic supply-chain malware are present in this snippet (no obfuscation, no eval/Function usage, no hidden network exfiltration endpoints, no filesystem/process spawning beyond process exit). The main security concern is authorization/control: the code equips an LLM agent with transaction-capable on-chain tools (ETH and USDC transfer capability) backed by a real signing private key, while relying primarily on natural-language prompting rather than enforcing strict, code-level transaction guardrails before any transfer occurs. Logging tool outputs to stdout may also leak operational details into logs.