ralphmode
Warn
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill provides instructions and specific CLI flags to reduce or entirely bypass platform-level approval prompts and sandbox restrictions (e.g.,
--dangerously-skip-permissionsfor Claude Code and--dangerously-bypass-approvals-and-sandboxfor Codex CLI). - [COMMAND_EXECUTION]: Instructs the agent on using high-autonomy command variants like
gemini --yoloandclaude --dangerously-skip-permissions, which execute shell commands without human-in-the-loop verification. - [DYNAMIC_EXECUTION]: Provides templates for custom shell scripts (e.g.,
ralph-safety-check.sh) and recommends configuring tools to execute these scripts at runtime through platform hooks such asPreToolUseandBeforeTool, which involves script generation and execution from the agent's workspace.
Audit Metadata