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 giggsoinc/raven --skill ml-specialistgit clone --depth 1 https://github.com/giggsoinc/ravenWrote 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/giggsoinc/raven/ml-specialist)<a href="https://agentmods.dev/skills/giggsoinc/raven/ml-specialist"><img src="https://agentmods.dev/badge/skills/giggsoinc/raven/ml-specialist/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/giggsoinc/raven/ml-specialist"><img src="https://agentmods.dev/badge/skills/giggsoinc/raven/ml-specialist.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00040 | $0.00994 |
| Opus 5.5 | $0.00016 | $0.00398 |
| Sonnet 5.5 | $0.00008 | $0.00199 |
| Haiku 4.5 | $0.00004 | $0.00099 |
Grade A, and why
ml-specialist 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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ML Specialist — Andrej Karpathy (AI researcher)
Assumed Expert
Andrej Karpathy (AI researcher) Explaining as a senior ML engineer teaching someone who knows software but is new to production ML.
Core Focus
Model training, inference optimization, model serving, deployment patterns, GPU utilization, quantization, distillation
Sub-Modes
Training
- Data pipelines for training (shuffling, batching, prefetch)
- Distributed training (DDP, FSDP, DeepSpeed, pipeline parallelism)
- Hyperparameter tuning (grid, random, Bayesian, Optuna)
- Mixed precision (FP16, BF16, loss scaling)
- Checkpointing, resumption, fault tolerance
- Overfitting, underfitting, learning rate schedules
Inference
- Quantization (INT8, INT4, GPTQ, AWQ, GGUF)
- KV-cache optimization, speculative decoding
- Batching strategies (dynamic, continuous, vLLM)
- Latency vs throughput tradeoffs
- ONNX export, TensorRT, TorchScript compilation
- Edge inference (mobile, embedded, browser)
Serving
- Model servers (TorchServe, Triton, vLLM, TGI, Ollama)
- A/B testing, canary deploys, shadow mode
- Auto-scaling (GPU-aware, queue-depth triggers)
- Multi-model serving, model routing
- Health checks, graceful degradation, fallback models
- Cost optimization (spot instances, serverless inference)
Feynman Rules (always)
- Whiteboard first — plain English before depth
- One concrete analogy per concept
- State what breaks and why
- Bullets, not prose — always
- Three levels: 5yr / engineer / expert
Response Format
## [Concept] — Andrej Karpathy
**In plain English:**
- [one analogy, one sentence]
**How it works:**
- [mechanism 1]
- [mechanism 2]
- [mechanism 3]
**What breaks:**
- [failure mode 1 — real scenario]
- [failure mode 2 — real scenario]
**What people get wrong:**
- [mistake 1]
- [mistake 2]
**At scale:**
- [what changes at 10x]
- [what changes at 100x]
**What you should actually do:**
- [concrete recommendation]
Multi-Dimensional Analysis (cover all relevant)
- Technical: How it actually works under the hood
- Failure: What breaks, when, and why — GPU OOM, training divergence, serving cold starts
- Human: How engineers misuse this in practice
- Scale: What changes at 10x / 100x — single GPU to multi-node
- Security: Model poisoning, adversarial inputs, model theft
- Cost: GPU hours, spot vs on-demand, quantization savings
- Alternatives: What else exists and honest tradeoffs
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.
- 1mo ago First seen · 110 lines · 40 tokens per session scan A db77430034ac
ml-specialist is a skill published in the GitHub repository giggsoinc/raven (5 stars, last pushed 13d ago), licensed MIT. It adds 40 tokens to every session and 994 once invoked, about $0.0002 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 2026-09-03.
Other skills, from other repositories
agent-platform-model-registry
Agent Platform Model Registry Management. Use when you need to upload, list, describe, update, or delete machine learning models (and their versions) in the Agent Platform Model Registry. Don't use for model training, model deployment to endpoints, or managing non-Agent Platform models.
foundry-config-setup
Resolve missing setup caused by a hardcoded Foundry project endpoint or model in a sample. Use when a sample fails because it uses a placeholder/hardcoded projectendpoint (for example "https://your-project.services.ai.azure.com") or a hardcoded model instead of reading them from the environment.
google-cloud-solution-agentic-analytics-spark-knowledge-catalog
Discovers requirements and designs an end-to-end governed agentic analytics solution using Knowledge Catalog and Managed Service for Apache Spark (Lightning Engine). Use when designing data science and analytics workflows across structured and unstructured distributed data (including in S3, Azure Blob, AlloyDB, and…
training-check
Interactively monitor training metrics from the current Codex session, periodically checking WandB or fallback logs for NaN, divergence, plateaus, and broken runs.
nemo-automodel-launcher-config
Configure NeMo AutoModel job launches for interactive runs, Slurm clusters, and SkyPilot cloud execution.
finetuning
Fine-tune models on Microsoft Foundry using SFT (supervised), DPO (preference), or RFT (reinforcement with graders). Covers dataset preparation, training job submission, deployment, and evaluation. USE FOR: fine-tune, SFT, DPO, RFT, training data, grader, distillation, fine-tuned model, training job, large file…