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 agentmods add skills/squadcodercom/squadcoder/hebrew-ml-datasets-navigatornpx skills add squadcodercom/squadcoder --skill hebrew-ml-datasets-navigatorgit clone --depth 1 https://github.com/squadcodercom/squadcoderWhat 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 | $0.00274 | $0.06038 |
| Opus 5 | $0.00137 | $0.03019 |
| Sonnet 5 | $0.00055 | $0.01208 |
| Haiku 4.5 | $0.00027 | $0.00604 |
Grade A, and why
hebrew-ml-datasets-navigator 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 2d 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-ml-datasets-navigator — 95% identical, 8 lines differ
How it starts
The opening of the file, as written. The whole thing — 288 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hebrew ML Datasets Navigator
Problem
The Israeli ML community punches above its weight, but the datasets and models are scattered. ivrit.ai publishes world-class Hebrew speech corpora on one HuggingFace org, Dicta publishes Hebrew LLMs and BERT variants on another, the Israeli National NLP Program maintains benchmarks under HebArabNlpProject, and classic resources like AlephBERT live elsewhere. Licenses vary from fully commercial-friendly to research-only. Hebrew register coverage varies dramatically: some corpora are all modern standard, others are half religious texts, others are spoken colloquial. A researcher trying to pick the right combination for "fine-tune a Hebrew sentiment classifier on customer support chat for a commercial product" has to hunt across five orgs and read every dataset card to understand what they can actually use.
Instructions
Step 1: Identify the task
Different Hebrew ML tasks need different datasets. Match your task to a dataset family before searching.
| Task | Primary data type | Dataset families to check first |
|---|---|---|
| Speech-to-text (Hebrew ASR) | Audio + transcripts | ivrit.ai (crowd-transcribe, crowd-recital, audio-v2) |
| Text-to-speech (Hebrew TTS) | Text + studio audio | Public-domain audio with permissive licenses (limited; often requires custom recording) |
| Hebrew LLM pre-training | Large Hebrew text corpus | Dicta's corpora, allenai/MADLAD-400 Hebrew subset, oscar-corpus/OSCAR-2301 Hebrew, uonlp/CulturaX Hebrew slice, HuggingFaceFW/fineweb-2 heb_Hebr filter, mC4 (Hebrew quality is weak), Hebrew Wikipedia, Knesset Plenums |
| Hebrew LLM instruction tuning | Prompt-response pairs in Hebrew | Dicta instruction datasets, translated Alpaca-style datasets, custom |
| Reading comprehension / QA | Text + Q&A pairs | HeQ (Etelis/HeQ_v1 HF mirror, canonical at github.com/NNLP-IL/Hebrew-Question-Answering-Dataset); omrikeren/ParaShoot (~3K few-shot QA examples) |
| Sentiment classification | Hebrew text + labels | HebrewSentiment (HebArabNlpProject/HebrewSentiment) |
| Natural language inference | Hebrew premise-hypothesis pairs | HebNLI (HebArabNlpProject/HebNLI) |
| Named entity recognition | Hebrew text + entity tags | Dicta NER datasets, historical NNLP-IL releases |
| Morphological analysis | Hebrew text + morph tags | Dicta morph datasets |
| Diacritization (nikud) | Unvocalized + vocalized Hebrew | Dicta nikud datasets |
| Paraphrase detection | Hebrew text pairs | NNLP-IL Hebrew paraphrase dataset (9,750 pairs) |
| Summarization | Hebrew article + summary | biunlp/HeSum (10K article-summary pairs from Hebrew news, BIU NLP), HebArabNlpProject/HebSummaries |
| General knowledge benchmarking | MCQ + answers | HEBREW-MMLU (Hebrew-translated MMLU subset; verify the active HF mirror, multiple community translations exist) |
| Hebrew-English translation | Parallel corpora | NeuLabs-TedTalks, OPUS Hebrew subsets |
| Yiddish ASR | Yiddish audio + transcripts | ivrit.ai Yiddish models (yi-whisper) and crowd datasets |
| Yiddish text | Yiddish corpora | ivrit.ai crowd-whatsapp-yi, crowd-recital-yi |
What ships with it
9 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.
- 2d ago First seen · 288 lines · 274 tokens per session scan A c387243c0999
hebrew-ml-datasets-navigator is a skill published in the GitHub repository squadcodercom/squadcoder (11 stars, last pushed 2mo ago), licensed MIT. It adds 274 tokens to every session and 6,038 once invoked, about $0.0014 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
tool-prompt-optimization
Optimize the description prompts an AI agent reads to learn its built-in tools (the .md files under prompts/tools/). Two halves: (1) measure how much of a prompt is already inferable from the tool's JSON parameter schema + name, to prune redundancy with evidence; (2) house authoring rules for what belongs in a tool…
semantic-compression
Re-encode verbose prose into a dense telegraphic register — punctuation as connectives, label frames, verbless assertions — without losing normativity or precision. Use when compressing system prompts, tool/function descriptions, skill bodies, or agent instructions; reducing token count or context bloat; making…
docker-extend
Use when: User wants to extend Docker with custom tools, personalize the Docker environment, or set up user-specific Docker customization. Triggers: 'extend docker', 'docker-extend', 'add tools to docker', 'customize docker', 'add my tools to the container', 'personalize docker setup', 'docker user setup', 'install…
ppt-generation
Use this skill when the user requests to generate, create, or make presentations (PPT/PPTX). Has TWO workflows: (1) Primary — AI-generated full-slide images composed via scripts/generate.py; (2) Fallback — python-pptx programmatic slides (all text editable, better for reports/project management). The fallback…
runjam-defaults
Default constraints for every RunJam session. Defines output path conventions, dependency checking rules, fallback strategies, and file management discipline. This skill is auto-injected into every session — do not remove. Current session working directory: /Users/guizhan/work/code/runjam.
database-migrations
Database migration best practices for schema changes, data migrations, rollbacks, and zero-downtime deployments across PostgreSQL, MySQL, and common ORMs (Prisma, Drizzle, Django, TypeORM, golang-migrate). Use when planning or implementing database schema changes.