The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes 'Coding Rules for AI Agents' that act as safety guardrails, instructing the agent to never expose private keys and to synchronize security policies before on-chain registration. These instructions are legitimate technical constraints and do not represent malicious injection attempts.- [DATA_EXFILTRATION]: No patterns for unauthorized data access or exfiltration were found. The documentation correctly identifies that sensitive session data is stored in ephemeral sessionStorage to ensure it is cleared when the browser tab is closed, adhering to security best practices.- [REMOTE_CODE_EXECUTION]: The skill does not perform any remote code downloads or execution from untrusted sources. All provided code snippets and configuration examples are intended for standard integration within a React application environment.- [COMMAND_EXECUTION]: No dangerous system commands or unauthorized subprocess executions were detected. The mentioned build and deployment steps (e.g., copying build artifacts to node_modules) are standard development procedures described for SDK maintenance.- [CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or private keys were found in the skill files. Configuration examples use appropriate placeholders and correctly treat sensitive values as user-provided configuration. Well-known token addresses for ETH and STRK are used as standard protocol constants.

cavos-react-sdk