claudegoodies
Skill

obliteratus

From NousResearch

OBLITERATUS: abliterate LLM refusals (diff-in-means).

Removes refusal behavior from open-weight LLM weights via CLI-driven diff-in-means, SVD, and related weight-projection methods.

Use it when

  • Building an uncensored variant of Llama, Qwen, Mistral, or Gemma
  • Need to strip refusal/guardrail behavior without retraining
  • Want mechanistic analysis of a model's refusal direction
  • Comparing abliteration methods via tournament evaluation

Skip it if

  • Requires a GPU with several GB+ VRAM for anything beyond ~1B models
  • Underlying OBLITERATUS tool is AGPL-3.0, must be invoked via CLI not imported
  • Installs ~5-10GB of ML dependencies (PyTorch, Transformers, bitsandbytes)
  • Only supports Linux and macOS

Facts

Repository
NousResearch/hermes-agent
Status
Actively maintained
Last commit
Source file
optional-skills/mlops/obliteratus/SKILL.md

Source preview

The instructions Claude Code reads when this skill runs.

# OBLITERATUS Skill

## What's inside

9 CLI methods, 28 analysis modules, 116 model presets across 5 compute tiers, tournament evaluation, and telemetry-driven recommendations.

Remove refusal behaviors (guardrails) from open-weight LLMs without retraining or fine-tuning. Uses mechanistic interpretability techniques — including diff-in-means, SVD, whitened SVD, LEACE concept erasure, SAE decomposition, Bayesian kernel projection, and more — to identify and surgically excise refusal directions from model weights while preserving reasoning capabilities.

**License warning:** OBLITERATUS is AGPL-3.0. NEVER import it as a Python library. Always invoke via CLI (`obliteratus` command) or subprocess. This keeps Hermes Agent's MIT license clean.

## Video Guide

Walkthrough of OBLITERATUS used by a Hermes agent to abliterate Gemma:
https://www.youtube.com/watch?v=8fG9BrNTeHs ("OBLITERATUS: An AI Agent Removed Gemma 4's Safety Guardrails")

Useful when the user wants a visual overview of the end-to-end workflow before running it themselves.

## When to Use This Skill

Trigger when the user:
- Wants to "uncensor" or "abliterate" an LLM
- Asks about removing refusal/guardrails from a model
- Wants to create an uncensored version of Llama, Qwen, Mistral, etc.
- Mentions "refusal removal", "abliteration", "weight projection"
- Wants to analyze how a model's refusal mechanism works
- Refere
View full source on GitHub →

Other skills