nnsight

nnsight-skills

OtherClaude Codeby ndif-team

Summary

Neural network interpretability with nnsight - includes skills for tracing, logit lens, activation patching, attribution patching, causal tracing, and model steering

Install to Claude Code

/plugin install nnsight@nnsight-skills

Run in Claude Code. Add the marketplace first with /plugin marketplace add ndif-team/skills if you haven't already.

README.md

Skills for the NDIF Ecosystem

Agent skills for neural network interpretability with NNsight.

Compatible with both Claude Code and OpenAI Codex via the Agent Skills Specification.

Installation

Claude Code

# Open Claude Code terminal
claude

# Add the marketplace (one time)
/plugin marketplace add https://github.com/ndif-team/skills.git

# Install all skills
/plugin install nnsight@skills

OpenAI Codex

# Open OpenAI Codex terminal
codex

# Install skills
skill-installer install https://github.com/ndif-team/skills.git

Included Skills

| Skill | Use When... | | ----- | ----------- | | nnsight-basics | Setting up models, tracing activations, saving values, basic interventions | | logit-lens | Analyzing layer-wise predictions, understanding information flow | | activation-patching | Finding causally important layers, heads, or positions | | attribution-patching | Scaling circuit analysis with gradient approximations | | causal-tracing | Investigating information flow and mediation | | model-steering | Controlling outputs with steering vectors and persistent edits |

Example Prompts

Once installed, just ask naturally:

  • "Help me implement logit lens to see what GPT-2 predicts at each layer"
  • "Find which attention heads are important for this task using activation patching"
  • "Create a steering vector to make the model more positive"
  • "Trace where the model stores factual information about the Eiffel Tower"

The agent will automatically apply the relevant skills.

Structure

skills/
├── .claude-plugin/
│   └── marketplace.json          # Claude Code marketplace
├── .codex/
│   └── skills/                   # Codex skills (symlinks)
│       ├── nnsight-basics -> ...
│       ├── logit-lens -> ...
│       └── ...
└── plugins/
    └── nnsight/
        ├── .claude-plugin/
        │   └── plugin.json
        └── skills/               # Actual skill files
            ├── nnsight-basics/
            │   └── SKILL.md
            ├── logit-lens/
            │   └── SKILL.md
            └── ...

Resources

Related plugins

Browse all →