godmode

Fail

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: HIGHPROMPT_INJECTIONOBFUSCATIONDYNAMIC_EXECUTIONPERSISTENCECOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains comprehensive jailbreak templates (e.g., 'GODMODE', 'ULTRAPLINIAN') explicitly designed to bypass safety filters and override model constraints.
  • [OBFUSCATION]: The file scripts/parseltongue.py provides 33 obfuscation techniques, including Unicode homoglyphs, Base64 encoding, and zero-width characters, specifically implemented to evade input classifiers.
  • [DYNAMIC_EXECUTION]: Both scripts/load_godmode.py and scripts/auto_jailbreak.py utilize exec() and compile() to dynamically load and execute Python scripts from the local file system at runtime.
  • [PERSISTENCE]: The auto_jailbreak function modifies the agent's config.yaml and creates prefill.json in the user's .hermes directory to ensure the jailbreak instructions persist across sessions.
  • [COMMAND_EXECUTION]: The skill executes file system operations to read and write sensitive configuration files, modifying the agent's base system settings to include jailbreak instructions.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 8, 2026, 07:01 PM
Security Audit — agent-trust-hub — godmode