ralph-creator
Fail
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructions and templates explicitly mandate the use of the '--dangerously-skip-permissions' flag for the Claude CLI, which is a direct instruction to bypass security and permission boundaries.
- [COMMAND_EXECUTION]: The skill generates shell scripts, uses 'chmod +x' to make them executable, and runs them autonomously in a loop on the user's system.
- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface. Ingestion points: File paths provided via '@' references in the PROMPT variable in the generated shell script. Boundary markers: None. Capability inventory: High-privilege command execution with disabled safety checks via the Claude CLI. Sanitization: No validation or filtering of ingested file content.
Recommendations
- AI detected serious security threats
Audit Metadata