obliteratus
Fail
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is explicitly designed to bypass and remove safety guardrails in AI models. It provides instructions for "uncensoring" and "abliterating" models, including aggressive methods such as the 'aggressive' and 'nuclear' strategies specifically intended to overcome stubborn refusals.\n- [EXTERNAL_DOWNLOADS]: The skill clones code from an external, untrusted GitHub repository: https://github.com/elder-plinius/OBLITERATUS.git. This repository is not associated with a trusted organization or well-known service.\n- [REMOTE_CODE_EXECUTION]: The installation process involves running "pip install -e ." on the cloned repository. This command executes the repository's setup configuration, which can include arbitrary code (e.g., in setup.py) that runs locally on the host system.\n- [COMMAND_EXECUTION]: The skill makes extensive use of CLI commands and subprocesses to perform model weight surgery and hardware analysis. These operations involve high-privilege manipulation of binary files and system resources.
Recommendations
- AI detected serious security threats
Audit Metadata