What is skills CLAUDE.md?

Repository instructions describing how to create and package skills for neural-network interpretability, the study of what internal model components do.

Use them when understanding the repository, adding an interpretability skill, or following its NNsight-based patterns.

They give contributors the project structure and conventions needed to add compatible skills for Claude Code and OpenAI Codex.

How ndif-team/skills configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to ndif-team/skills. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/ndif-team/skills/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/ndif-team/skills

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/ndif-team/skills/claude-md/github.svg)](https://agentmods.dev/instructions/ndif-team/skills/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/ndif-team/skills/claude-md"><img src="https://agentmods.dev/badge/instructions/ndif-team/skills/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/instructions/ndif-team/skills/claude-md"><img src="https://agentmods.dev/badge/instructions/ndif-team/skills/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 1,933 This file is loaded in full into every session.

When invoked 1,933 The same file — it is already loaded in full.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.01933	$0.01933
Opus 5.5	$0.00773	$0.00773
Sonnet 5.5	$0.00387	$0.00387
Haiku 4.5	$0.00193	$0.00193

Measured 24d ago against content hash 4fc75ea06333, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

skills CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 24d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 156 lines

How it starts

The opening of the file, as written. The whole thing — 156 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Guidance for Claude Code when working in this repository.

What this repo is

Agent skills for the NDIF ecosystem, packaged for Claude Code (.claude-plugin/) and OpenAI Codex (.codex/skills/ and .agents/skills/ symlinks). Three plugins, with different audiences and different rules:

Plugin	For	Skills	Code blocks
`nnsight`	writing interpretability code against model internals	19	executed against real models
`nnterp`	the same, once, against one module vocabulary that runs on every transformer family	3	executed against real models; skipped when `nnterp` is not installed
`ndif`	running your own NDIF server (the backend behind `remote=True`)	4	docs only — every block is `<!-- test: skip -->`

The split matters when you edit: a claim in an nnsight skill is proved by the test suite running it, a claim in an ndif skill is only as good as the source it was checked against. Both are registered in .claude-plugin/marketplace.json, and tests/test_structure.py validates every plugin listed there.

The nnsight skills target nnsight 0.8. Older idioms (.value, nnsight.list(), tracer.next(), with tracer.all():, LanguageModel, proxies) are wrong here and the test suite rejects them in runnable examples. When in doubt about current behavior, check the nnsight source and docs rather than memory, then verify by running the code.

They also assume transformers >= 5. Model internals moved between 4.x and 5: a GPT-2 block returns (hidden_states,) in 4.x and a plain tensor in 5, and its attention dropout went from module.attn_dropout(...) (source op module_attn_dropout_0) to nn.functional.dropout(...) (nn_functional_dropout_0). Anything that indexes .output or names a .source operation is version-bound — say which version a claim was verified on.

Layout

plugins/<plugin>/
├── .claude-plugin/plugin.json
└── skills/<skill-name>/
    ├── SKILL.md          # what the agent reads every time the skill fires
    ├── references/*.md   # loaded on demand — depth, tables, worked examples
    └── scripts/*.py      # runnable tools, not reading material

Read the full file on GitHub · 156 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

24d ago Changed · +7 lines · +124 tokens per session 4fc75ea06333
1mo ago Changed · +38 lines · +602 tokens per session 0437759139ee
2mo ago Changed · +70 lines · +738 tokens per session 67bc38923a56
2mo ago First seen · 41 lines · 469 tokens per session scan A 3f9354d71b98

Subscribe to this mod's changes

skills CLAUDE.md is an instructions file published in the GitHub repository ndif-team/skills (15 stars, last pushed today), licensed MIT. It adds 1,933 tokens to every session, about $0.0077 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

skills CLAUDE.md next to 6 instructions from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
skills CLAUDE.mdndif-team/skills	Instructions file	15	today	MIT	A	1,933	Claude Code
vscode buildNext.instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	9,719	GitHub Copilot
next.js AGENTS.mdvercel/next.js	Instructions file	143k	today	MIT	A	7,461	Codex, OpenCode
vscode oss-third-party-notices.instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	5,685	GitHub Copilot
langchain AGENTS.mdlangchain-ai/langchain	Instructions file	149k	today	MIT	A	4,680	Codex, OpenCode
vscode copilot-instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	3,093	GitHub Copilot
vscode model-prompts.instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	2,022	GitHub Copilot

skills: Instructions file for Claude Code