Security audit
Security checks for vulnerabilities and agentic risk
The skill appears to be a real task-marketplace and USDC settlement guide, but it is too vague for wallet, credential, command execution, and financial actions.
Review before installing. Use only a limited wallet and least-privilege marketplace token, never paste seed phrases or private keys into prompts or command arguments, and manually confirm any proposal, delivery, settlement, or wallet-signing action.
The skill declares very broad activation keywords such as generic work/task-market phrasing that can overlap with ordinary user conversation. In agent environments that auto-select or suggest skills from descriptive text, this can cause the skill to activate unexpectedly and expose users to command execution, wallet-related flows, or credential-handling contexts they did not explicitly request.
The capability coverage section expands matching to broad scenario keywords without defining strict trigger conditions. This increases the chance of unintentional routing into a skill that supports read/exec/write and financial workflow guidance, creating unnecessary exposure to sensitive operations.
The skill discusses wallet setup, API credentials, command execution, and USDC settlement, but does not provide a strong, front-loaded security warning for sensitive actions. In this context, users may be nudged into providing secrets or running commands without understanding the risks of credential leakage, unintended transactions, or destructive local effects.
Line L251 states '钱包私钥仅本地签名，不上传' ('wallet private key only for local signing, not uploaded'), which is a concrete behavioral assurance. However, elsewhere the skill is explicitly framed around networked marketplace access and on-chain settlement, and no code or script is provided here to substantiate that the private key is in fact kept local-only. This creates an intent/behavior assurance mismatch within the documentation itself.
