By ndif-team
Neural network interpretability with nnsight - tracing, patching, steering, and analysis
Causal intervention via activation patching to identify important model components. Use when determining which layers, heads, or positions are causally responsible for model behavior.
Gradient-based approximation to activation patching for scalable circuit analysis. Use when activation patching is too slow or when analyzing many components simultaneously.
Causal mediation analysis to identify which model components mediate specific behaviors. Use when investigating how information flows through the network and which neurons or layers are causally responsible for outputs.
Decode intermediate layer predictions using the Logit Lens technique. Use when analyzing what a model predicts at each layer, understanding information flow, or visualizing layer-wise processing.
Control model behavior through persistent edits and steering interventions. Use when modifying model outputs, applying steering vectors, or creating persistently modified model versions.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Agent skills for neural network interpretability with NNsight.
Compatible with both Claude Code and OpenAI Codex via the Agent Skills Specification.
# 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
# Open OpenAI Codex terminal
codex
# Install skills
skill-installer install https://github.com/ndif-team/skills.git
| 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 |
Once installed, just ask naturally:
The agent will automatically apply the relevant skills.
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
└── ...
npx claudepluginhub ndif-team/skillsBuild and configure neural network architectures
ML engineering plugin: Give your AI coding agent ML engineering superpowers.
Evaluate and compare ML model performance metrics
Local-first resolver for Hugging Face models (GGUF, MLX, safetensors). The agent checks your own storage and any mounted drives before downloading anything.
Dynamic/morphogenetic neural networks - grow, prune, adapt topology during training. Covers continual learning, gradient isolation, PEFT/LoRA, modular composition, lifecycle orchestration. 7 reference sheets, 2 commands, 1 SME agent.
LLM post-training — unified interface for SFT, OSFT, LoRA fine-tuning, and GRPO reinforcement learning