The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions that explicitly guide the user or agent to bypass or override built-in security permission systems. Key examples include enabling bypassPermissions in Claude Code, setting approval_policy = "never" in Codex CLI, and using the --dangerously-skip-permissions command-line flag.
[COMMAND_EXECUTION]: The skill provides shell script templates (e.g., ralph-safety-check.sh) and instructions to make them executable via chmod +x. These scripts are intended to be used as system hooks to intercept and validate commands at runtime.
[EXTERNAL_DOWNLOADS]: The documentation references external configuration guidelines and repositories, such as the official Google Gemini GitHub repository, to assist with platform-specific setup.

ralphmode