The Agent Skills Directory

[SAFE]: The skill follows robust security practices for blockchain application development. It includes non-negotiable security rules that prohibit hardcoding secrets, require the use of environment variables, and mandate human-in-the-loop confirmation for all financial transactions.
[EXTERNAL_DOWNLOADS]: The skill references legitimate, well-known industry libraries such as viem and bip39 for mnemonic generation and Ethereum interactions. It also utilizes the official vendor SDK @circle-fin/modular-wallets-core, which is appropriate for its stated purpose.
[CREDENTIALS_UNSAFE]: The skill proactively warns against storing sensitive data like mnemonics or passkey credentials in insecure locations such as localStorage for production environments. It provides clear guidance on using secrets managers and environment variables instead.
[DATA_EXFILTRATION]: No patterns of unauthorized data access or external exfiltration were detected. All network communication is directed toward official Circle RPC endpoints and standard blockchain infrastructure.
[PROMPT_INJECTION]: The instructions are purely functional and do not contain any patterns intended to bypass safety filters, override agent behavior, or extract system prompts.

use-modular-wallets