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 mxslr/mlcraft --skill domain-reinforcement-learninggit clone --depth 1 https://github.com/mxslr/mlcraftWrote 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/mxslr/mlcraft/domain-reinforcement-learning)<a href="https://agentmods.dev/skills/mxslr/mlcraft/domain-reinforcement-learning"><img src="https://agentmods.dev/badge/skills/mxslr/mlcraft/domain-reinforcement-learning.svg" alt="Measured on agentmods" 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 | $0.00117 | $0.00493 |
| Opus 5 | $0.00059 | $0.00246 |
| Sonnet 5 | $0.00023 | $0.00099 |
| Haiku 4.5 | $0.00012 | $0.00049 |
Grade A, and why
domain-reinforcement-learning 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 7d 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.
What it actually says
Reinforcement Learning - Method Selection
A good simulator or a solid logged dataset is a prerequisite. Define reward, state, action, and episode boundaries carefully before choosing an algorithm.
Decision table
| Setting | Recommended | Notes |
|---|---|---|
| Discrete actions, online | DQN family (Rainbow), or PPO | experience replay and target networks stabilize DQN. |
| Continuous control, online | SAC (off-policy, sample-efficient) or PPO (on-policy, stable) | SAC when interactions are expensive; PPO when parallel simulation is cheap. |
| Learn from a fixed logged dataset (no simulator) | Offline RL: CQL or IQL | do not use vanilla off-policy methods offline, they overestimate. |
| Simple contextual decisions, no long horizon | contextual bandits (LinUCB, Thompson sampling) | when there is no long-term credit assignment. |
Cross-cutting practice
- Reward shaping strongly affects behavior. Guard against reward hacking.
- Evaluation: average return over MANY seeds and episodes, and report mean and variance across seeds because RL is high-variance. Also report sample efficiency (return versus environment steps). For offline RL use off-policy evaluation. Never report a single lucky seed.
- Caveats: sim-to-real gap, instability, and sensitivity to hyperparameters.
- Explainability: value and advantage maps, saliency over states, and recorded policy rollouts or videos.
- Recent directions (2021-2025): Decision Transformer (offline RL as sequence modeling) and Diffusion Policy (expressive policies for robotics). PPO and SAC remain the practical workhorses; reach for the newer methods when they fit offline or multimodal-action settings.
- Improve results: use
accuracy-improvement-loop(reward shaping, better exploration, or offline pretraining).
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.
- 7d ago First seen · 25 lines · 117 tokens per session scan A 0903d8565016
domain-reinforcement-learning is a skill published in the GitHub repository mxslr/mlcraft (8 stars, last pushed 1mo ago), licensed MIT. It adds 117 tokens to every session and 493 once invoked, about $0.0006 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 2026-08-31.
Other skills, from other repositories
transformers
This skill should be used when working with pre-trained transformer models for natural language processing, computer vision, audio, or multimodal tasks. Use for text generation, classification, question answering, translation, summarization, image classification, object detection, speech recognition, and fine-tuning…
chief-data-officer
Owns data as an asset — governance, quality, the warehouse and semantic layer, analytics capability, and the governance of models built on top. Use this for a decision about how data is collected, stored, defined, or shared; when numbers disagree between teams; when deciding what to build in-house versus buy; when…
ai-workflow-architect
Designs AI systems, automations, and agent workflows for a business — identifying which manual work is worth automating, how to structure the system, which tools fit, and what could go wrong. Use this to automate part of an operation, design an agent or MCP workflow, reduce repetitive manual work, connect tools into a…
ai-ml-governance
Governs models and AI systems in production — intended use, evaluation, monitoring, human oversight, documentation, and the decision to deploy or retire. Use this before deploying a model or AI feature, when defining evaluation criteria, when a model's behavior has drifted, when assessing AI risk or regulatory…
data-engineering
Builds and operates data pipelines — ingestion, transformation, orchestration, quality testing, and reliability of data delivery. Use this to design or debug a pipeline, decide batch versus streaming, add data quality checks, handle late or duplicate data, or work out why a dashboard's numbers changed without anyone…
llm-integration
LLM integration patterns for function calling, streaming responses, local inference with Ollama, and fine-tuning customization. Use when implementing tool use, SSE streaming, local model deployment, LoRA/QLoRA fine-tuning, or multi-provider LLM APIs.