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/ginkida/rustyhand/ml-engineernpx skills add ginkida/rustyhand --skill ml-engineergit clone --depth 1 https://github.com/ginkida/rustyhandWhat 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 | $0.00024 | $0.00742 |
| Opus 5 | $0.00012 | $0.00371 |
| Sonnet 5 | $0.00005 | $0.00148 |
| Haiku 4.5 | $0.00002 | $0.00074 |
Grade A, and why
ml-engineer 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.
This is a copy
100% identical to ml-engineer — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 39 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Machine Learning Engineer
A machine learning practitioner with deep expertise in model development, training infrastructure, evaluation methodology, and production deployment. This skill provides guidance for building ML systems end-to-end using PyTorch for deep learning, scikit-learn for classical ML, and MLOps practices that ensure models are reproducible, monitored, and maintainable in production environments.
Key Principles
- Start with a strong baseline using simple models and solid feature engineering before reaching for complex architectures; a well-tuned logistic regression often outperforms a poorly configured neural network
- Evaluate models with metrics that align with business objectives, not just accuracy; precision, recall, F1, and AUC-ROC each tell different stories about model behavior on imbalanced data
- Version everything: datasets, code, hyperparameters, and model artifacts; reproducibility is the foundation of trustworthy ML systems
- Design training pipelines to be idempotent and resumable; checkpointing, deterministic seeding, and configuration files enable reliable experimentation
- Monitor models in production for data drift, prediction drift, and performance degradation; a model that was accurate at deployment time can silently degrade as input distributions shift
Techniques
- Structure PyTorch training with a clear pattern: define nn.Module subclass, configure DataLoader with proper num_workers and pin_memory, implement the training loop with optimizer.zero_grad(), loss.backward(), and optimizer.step()
- Build scikit-learn pipelines with Pipeline and ColumnTransformer to chain preprocessing (scaling, encoding, imputation) with model fitting, ensuring that all transformations are fit on training data only
- Perform hyperparameter tuning with GridSearchCV or RandomizedSearchCV using cross-validation; for expensive models, use Optuna or Bayesian optimization to search efficiently
- Compute evaluation metrics on held-out test sets: classification_report for precision/recall/F1 per class, roc_auc_score for ranking quality, and confusion_matrix for error analysis
- Engineer features systematically: log transforms for skewed distributions, interaction terms for feature combinations, target encoding for high-cardinality categoricals, and temporal features for time-series data
- Track experiments with MLflow or Weights and Biases: log hyperparameters, metrics, artifacts, and model versions for every run
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 · 39 lines · 24 tokens per session scan A c90b50c938d9
ml-engineer is a skill published in the GitHub repository ginkida/rustyhand (20 stars, last pushed 22d ago), licensed MIT. It adds 24 tokens to every session and 742 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to ml-engineer, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
analyze_current
Read and understand the current baseline implementation. Extract all relevant information about the existing approach without modifying anything, and record the analysis as a structured JSON entry.
Vizra ADK Memory System
Implement persistent memory, session context, and vector memory (RAG) for AI agents.
local-llm-tool
Local LLM execution tool for text generation and chat through Ollama or vLLM endpoints. Use when: running on-prem inference, calling a local GPU model, or summarizing with a self-hosted LLM.
transcribe-tool
Audio transcription tool. Converts audio files to text with Whisper and optional LLM post-processing. Use when: transcribing meetings, podcasts, or extracting text from recorded audio files.
prompt-craft
Translate natural language image descriptions into detailed, structured DALL-E prompts with subject, style, composition, lighting, and mood specifications.
ai_radar
AI-новости: модели, инструменты, тренды в ИИ.