The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing the claude CLI via shell commands. It emphasizes best practices such as using the -p flag for non-interactive automation to prevent execution delays.- [PROMPT_INJECTION]: The skill documents an interface where user-provided prompts are passed to the claude CLI via the shell. This constitutes a potential surface for indirect prompt injection. The skill provides clear instructions on permission levels and warns that bypass modes lack injection protection, recommending they be used only in isolated environments.- [CREDENTIALS_UNSAFE]: The instructions cover the handling of authentication tokens and API keys (e.g., ANTHROPIC_API_KEY). It follows standard security practices by recommending the use of environment variables and CLI-native authentication flows rather than hardcoding sensitive information.

claude-code