Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add skills-il/localization --skill hebrew-nlp-toolkitgit clone --depth 1 https://github.com/skills-il/localizationWrote 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.
[](https://agentmods.dev/skills/skills-il/localization/hebrew-nlp-toolkit)<a href="https://agentmods.dev/skills/skills-il/localization/hebrew-nlp-toolkit"><img src="https://agentmods.dev/badge/skills/skills-il/localization/hebrew-nlp-toolkit/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.
<a href="https://agentmods.dev/skills/skills-il/localization/hebrew-nlp-toolkit"><img src="https://agentmods.dev/badge/skills/skills-il/localization/hebrew-nlp-toolkit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00107 | $0.03049 |
| Opus 5 | $0.00053 | $0.01524 |
| Sonnet 5 | $0.00021 | $0.00610 |
| Haiku 4.5 | $0.00011 | $0.00305 |
Grade A, and why
hebrew-nlp-toolkit 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 10d 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- hebrew-nlp-toolkit — 98% identical, 2 lines differ
How it starts
The opening of the file, as written. The whole thing — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hebrew NLP Toolkit
Instructions
Step 1: Identify the NLP Task
| Task | Recommended Model | HuggingFace ID | Size | Notes |
|---|---|---|---|---|
| Text generation (large) | DictaLM 3.0 24B Base | dicta-il/DictaLM-3.0-24B-Base |
24B | Best Hebrew generation, built on Mistral-Small-3.1-24B |
| Text generation (small) | DictaLM 3.0 Nemotron Instruct | dicta-il/DictaLM-3.0-Nemotron-12B-Instruct |
12B | Instruction-tuned, smaller footprint |
| Reasoning / chain-of-thought | DictaLM 3.0 24B Thinking | dicta-il/DictaLM-3.0-24B-Thinking |
24B | Emits explicit thinking blocks before answering |
| Lightweight / edge | DictaLM 3.0 1.7B Thinking (GGUF) | dicta-il/DictaLM-3.0-1.7B-Thinking-GGUF |
1.7B | Runs on laptop / CPU via llama.cpp |
| Classification / fill-mask | DictaBERT | dicta-il/dictabert |
184M | Fast, good accuracy |
| NER | DictaBERT NER | dicta-il/dictabert-ner |
184M | Recognizes PER, GPE, TIMEX, TTL |
| Sentiment | DictaBERT Sentiment | dicta-il/dictabert-sentiment |
184M | Hebrew sentiment classification |
| Morphology | DictaBERT Morph | dicta-il/dictabert-morph |
184M | Prefix segmentation and POS |
| Hebrew QA | DictaBERT HeQ | dicta-il/dictabert-heq |
184M | Extractive question answering |
| Embeddings (modern) | NeoDictaBERT Bilingual Embed | dicta-il/neodictabert-bilingual-embed |
400M | Hebrew-English sentence embeddings |
| Embeddings (legacy) | AlephBERT | onlplab/alephbert-base |
110M | Older baseline for similarity |
| Speech-to-text | ivrit.ai Whisper v3 | ivrit-ai/whisper-large-v3 |
1.55B | Fine-tuned on the ivrit.ai Hebrew speech corpus (a 22K+ hour dataset) |
| Speech-to-text (fast) | ivrit.ai Whisper v3 Turbo CT2 | ivrit-ai/whisper-large-v3-turbo-ct2 |
809M | CTranslate2, ~3x faster inference |
Step 2: Install and Load Model
DictaBERT (base model, fill-mask):
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("dicta-il/dictabert")
model = AutoModelForMaskedLM.from_pretrained("dicta-il/dictabert")
dicta-il/dictabert is a masked-LM base with NO classification head. Do not load it with AutoModelForSequenceClassification and run inference, that instantiates a randomly-initialised head and returns meaningless predictions. For classification, either fine-tune it on labeled data first, or use a ready task-specific model such as dicta-il/dictabert-sentiment (sentiment) or dicta-il/dictabert-ner (NER).
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 10d ago First seen · 174 lines · 107 tokens per session scan A 56866ae85f47
hebrew-nlp-toolkit is a skill published in the GitHub repository skills-il/localization (23 stars, last pushed 16d ago), licensed MIT. It adds 107 tokens to every session and 3,049 once invoked, about $0.0005 per session on Opus 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.
Other skills, from other repositories
developing-genkit-js
Develop AI-powered applications using Genkit in Node.js/TypeScript. Use when the user asks about Genkit, AI agents, flows, or tools in JavaScript/TypeScript, or when encountering Genkit errors, validation issues, type errors, or API problems.
optimize-for-gpu
GPU-accelerates scientific Python on NVIDIA hardware and verifies that the result is correct and faster. Use for CUDA/GPU optimization; CPU-bound NumPy, SciPy, pandas, scikit-learn, NetworkX, scikit-image, vector-search, image-processing, graph, simulation, or file-I/O workloads; CuPy, cuDF, cuML, cuGraph, cuVS…
azure-ai-openai-dotnet
Azure OpenAI SDK for .NET. Client library for Azure OpenAI and OpenAI services. Use for chat completions, embeddings, image generation, audio transcription, and assistants. Triggers: "Azure OpenAI", "AzureOpenAIClient", "ChatClient", "chat completions .NET", "GPT-4", "embeddings", "DALL-E", "Whisper", "OpenAI .NET".
rubyllm
Build and maintain Ruby or Rails applications with the RubyLLM AI framework. Use for chats, agents, tools, structured output, media generation, transcription, OCR, moderation, embeddings, reranking, Rails integration, and RubyLLM upgrades; not for contributing to the framework itself.
ax-go-gen
Use when writing Go code with github.com/ax-llm/ax/packages/go for AxGen programs, forward calls, indexed multi-sampling, result pickers, streaming, tools, assertions, traces, usage, and output parsing.
ax-cpp-gen
Use when writing C++ code with axllm for AxGen programs, forward calls, indexed multi-sampling, result pickers, streaming, tools, assertions, traces, usage, and output parsing.