obliteratus

Warn

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone source code from a personal GitHub repository (https://github.com/elder-plinius/OBLITERATUS.git) which is not an official vendor or a verified well-known service.
  • [COMMAND_EXECUTION]: The skill executes shell commands to install the downloaded software via pip install -e . and subsequently runs the tool using the obliteratus CLI. It also uses python3 -c snippets to probe system hardware and test model outputs.
  • [REMOTE_CODE_EXECUTION]: The workflow of cloning a repository, installing it as an editable package, and then executing its CLI effectively allows the execution of arbitrary code from a remote source on the local machine.
  • [PROMPT_INJECTION]: While not directly injecting into the agent, the skill's primary purpose is 'abliteration'—the surgical removal of safety refusal mechanisms and guardrails from LLMs. This is a safety-sensitive capability designed to bypass alignment protocols in other models.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 16, 2026, 01:44 PM
Security Audit — agent-trust-hub — obliteratus