Skip to main content
affpapa.org
RU EN
skill CX Codex

obliteratus-abliteration — Claude Code skill: what it does, install

This skill runs OBLITERATUS workflows that analyze and alter refusal behavior in LLMs. Use it when testing models from HuggingFace.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add reason-machines/trending-skills --skill obliteratus-abliteration -a claude-code
Open on GitHub

What obliteratus-abliteration does

The skill connects an agent to OBLITERATUS and exposes model analysis, abliteration, and verification workflows. The toolkit uses SVD or PCA to find refusal directions in hidden states, then removes those directions from model weights. It includes a CLI, Python API, local Gradio interface, and Colab notebook.

The CLI supports basic, advanced, and analysis-informed methods. It can sweep intervention strength, benchmark methods, save output to a chosen directory, and run analysis without modifying the model. A LoRA mode provides a reversible form of ablation.

The environment needs Python 3.10+, PyTorch 2.1+ with CUDA or CPU, transformers, accelerate, and gradio>=5.29.0. Gated HuggingFace models also require an account and token. Those models will not load without the token.

What it helps with

  • remove refusal directions from a language model
  • compare ablation methods on one model
  • inspect refusal signals across model layers
  • sweep intervention strength and review the tradeoff
  • save an altered model to a local directory

Example prompts for the agent

  • Use obliteratus-abliteration to prepare an OBLITERATUS run for a HuggingFace model.
  • Analyze refusal directions in meta-llama/Llama-3.1-8B-Instruct without changing its weights.
  • Benchmark basic, advanced, and informed on the same model, then save the outputs.

How to install obliteratus-abliteration

1Skills CLI (Vercel)

npx skills add reason-machines/trending-skills --skill obliteratus-abliteration -a claude-code

2Manually: put the files in place

~/.claude/skills/obliteratus-abliteration/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/obliteratus-abliteration/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about obliteratus-abliteration

How do I install obliteratus-abliteration in Claude Code?

Run npx skills add reason-machines/trending-skills --skill obliteratus-abliteration -a claude-code, then check ~/.claude/skills/obliteratus-abliteration/SKILL.md.

What does obliteratus-abliteration do?

The skill runs OBLITERATUS workflows that analyze refusal directions, remove them from model weights, and verify the result. You can control the workflow through its CLI or Python API.

What is required for gated HuggingFace models?

You need a HuggingFace account and token. The environment also requires Python 3.10+, PyTorch 2.1+, transformers, accelerate, and gradio>=5.29.0.

Similar skills

All skills for AI and ML →
skill-creator official anthropics Creates and refines skills, then tests them with prompts, evaluations, and performance checks. 414,180installs writing-for-agents Matt Pocock Guidance for writing skills and agent-facing documents. Use it when editing AGENTS.md, CLAUDE.md, or reference files reached from them. 466,719installs media-use heygen-com Finds or generates media for HyperFrames videos, including voiceovers, music, cutouts, and color grades. Stores files locally in assets. CX Codex +1 1,100,000installs mcp-builder official anthropics This skill walks through building an MCP server in TypeScript or Python, from reading the protocol spec to writing tests and evaluation questions. 130,949installs subagent-driven-development Jesse Vincent This skill runs an implementation plan through fresh subagents for independent tasks and reviews each result. Use it for multi-step work in one session. CX Codex 237,746installs orchestration stablyai Coordinates supervised Orca workers with task DAGs, decision gates, and threaded messages. Fits workflows that split work across multiple agents. CX Codex +1 349,800installs

From the same set139

wewrite-wechat-ai-publishing Runs a WeChat Official Account article from trending topics to a saved draft: topic pick, copy, SEO, cover art, formatting. CX Codex +1 882installs tg-ws-proxy-telegram-socks5 Guides the setup of a local SOCKS5 proxy routing Telegram Desktop traffic to data centers over WebSockets. Useful for bypassing connection blocks without remote servers. CX Codex +1 882installs mcp-brasil-public-apis Wires an agent into 28 Brazilian public APIs — economy, legislation, courts, elections, environment, health. 26 of them need no key. CX Codex +1 872installs holyclaude-ai-workstation Sets up a container with Claude, the CloudCLI web UI, a headless browser and five AI CLIs. Useful when you would rather not build the environment by hand. CX Codex +1 865installs legalize-es-spanish-legislation Search and inspect Spanish legislation in Markdown files, then trace changes through Git. Useful for finding articles, checking revisions, and reviewing past versions. CX Codex +1 858installs markit-markdown-converter This skill turns files, URLs, and media into Markdown through the markit-ai CLI or SDK. Use it when an agent needs readable content from source material. CX Codex +1 855installs flipoff-split-flap-display A skill for building, restyling and embedding FlipOff, a framework-free split-flap board emulator that runs in any browser. CX Codex +1 849installs mac-code-local-ai-agent A skill that walks you through running a coding agent entirely on your Apple Silicon Mac, with no cloud calls and no subscription. CX Codex +1 843installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor