Not in the top 1,000 by GitHub stars
itsmostafa/llm-engineering-skills
LLM Engineering Claude Skills
History
Review weekly changes to SKILL.md before updating. Dates mark the weeks in which we observed each version; checks apply to the recorded commits.
View weekly changes
skills/agents/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/context-engineering/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/lora/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/mlx/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/prompt-engineering/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/pytorch/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/qlora/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/rlhf/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
skills/transformers/SKILL.md
Week of Nov 2, 2026 · Line counts unavailable
Before Not checked yet → After Network or broad shell
First observation → c364a15d5138
No earlier version was observed for comparison.
View on GitHubAdd to my skills
Verified commit c364a15d5138e3240a360b919a757249ab3e5cd9
git clone --filter=blob:none --no-checkout https://github.com/itsmostafa/llm-engineering-skills.git .hotskills/itsmostafa-llm-engineering-skills && git -C .hotskills/itsmostafa-llm-engineering-skills checkout --detach c364a15d5138e3240a360b919a757249ab3e5cd9
# Copy each selected skill directory from .hotskills/itsmostafa-llm-engineering-skills into .claude/skills/<skill-name>/24GitHub stars
0stars in 7 days
9skills in this repository
0commits in 30 days
Heat
0.4- Star growth, last 7 daysNot available
- Star growth, last 24 hoursNot available
- Rank movement, last 7 daysNot available
- Commits, last 30 daysNot available
- Accounts asking about it through MCP, last 7 daysNot available
- Hot votes this weekNot available
Heat never changes the order by stars, and sponsorship never touches it. How heat works
Stars over time
30-day change not yet available: not enough history.
Individual skills in this repository
- agents
skills/agents - context-engineering
skills/context-engineering - lora
skills/lora - mlx
skills/mlx - prompt-engineering
skills/prompt-engineering - pytorch
skills/pytorch - qlora
skills/qlora - rlhf
skills/rlhf - transformers
skills/transformers
How it works
agents
When your agent uses it
Patterns and architectures for building AI agents and workflows with LLMs. Use when designing systems that involve tool use, multi-step reasoning, autonomous decision-making, or orchestration of LLM-driven tasks.
What it loads into context
Frontmatter
46 tokensAlways available to the agent
SKILL.md
2,633 tokensWhen the skill is activated
Reference documents
47,073 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Agents vs Workflows
- Workflow Patterns
- Agent Architectures
- ReAct Pattern
- Tool Design
- External Context Protocols
- Best Practices
- References
Files
../../.claude-plugin/marketplace.jsonasset815 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytesSKILL.mdcore12,279 bytes../context-engineering/SKILL.mddoc13,293 bytes../context-engineering/reference/evaluation-strategies.mddoc9,533 bytes../context-engineering/reference/summarization-patterns.mddoc11,242 bytes../lora/SKILL.mddoc13,200 bytes../lora/reference/advanced-techniques.mddoc8,914 bytes../mlx/SKILL.mddoc9,103 bytes../mlx/reference/fine-tuning.mddoc8,928 bytes../mlx/reference/quantization.mddoc5,995 bytes../prompt-engineering/SKILL.mddoc9,824 bytes../pytorch/SKILL.mddoc10,473 bytes../pytorch/reference/debugging.mddoc7,610 bytes../pytorch/reference/training-patterns.mddoc7,574 bytes../qlora/SKILL.mddoc12,705 bytes../rlhf/SKILL.mddoc14,189 bytes../rlhf/reference/direct-alignment.mddoc9,221 bytes../rlhf/reference/policy-optimization.mddoc10,470 bytes../rlhf/reference/reward-modeling.mddoc7,884 bytes../transformers/SKILL.mddoc13,654 bytes../transformers/reference/fine-tuning.mddoc12,531 bytes
Read SKILL.md
# Building Agents
Agents are systems where LLMs dynamically direct their own processes and tool usage. This skill covers when to use agents vs workflows, common architectural patterns, and practical implementation guidance.
## Table of Contents
- [Agents vs Workflows](#agents-vs-workflows)
- [Workflow Patterns](#workflow-patterns)
- [Agent Architectures](#agent-architectures)
- [ReAct Pattern](#react-pattern)
- [Tool Design](#tool-design)
- [External Context Protocols](#external-context-protocols)
- [Best Practices](#best-practices)
- [References](#references)context-engineering
When your agent uses it
Strategies for managing LLM context windows effectively in AI agents. Use when building agents that handle long conversations, multi-step tasks, tool orchestration, or need to maintain coherence across extended interactions.
What it loads into context
Frontmatter
46 tokensAlways available to the agent
SKILL.md
2,809 tokensWhen the skill is activated
Reference documents
4,352 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Principles
- Context Management Strategies
- System Prompt Design
- Tool Design for Context Efficiency
- Long-Horizon Task Patterns
- Implementation Patterns
- Best Practices
- References
Files
SKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesSKILL.mdcore13,293 bytesreference/evaluation-strategies.mddoc9,533 bytesreference/summarization-patterns.mddoc11,242 bytes
Read SKILL.md
# Context Engineering
Context engineering is the discipline of curating and maintaining the optimal set of tokens during LLM inference. Unlike prompt engineering (crafting individual prompts), context engineering focuses on what information enters the context window and when.
## Table of Contents
- [Core Principles](#core-principles)
- [Context Management Strategies](#context-management-strategies)
- [System Prompt Design](#system-prompt-design)
- [Tool Design for Context Efficiency](#tool-design-for-context-efficiency)
- [Long-Horizon Task Patterns](#long-horizon-task-patterns)
- [Implementation Patterns](#implementation-patterns)
- [Best Practices](#best-practices)
- [References](#references)lora
When your agent uses it
Parameter-efficient fine-tuning with Low-Rank Adaptation (LoRA). Use when fine-tuning large language models with limited GPU memory, creating task-specific adapters, or when you need to train multiple specialized models from a single base.
What it loads into context
Frontmatter
54 tokensAlways available to the agent
SKILL.md
3,416 tokensWhen the skill is activated
Reference documents
2,175 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Concepts
- Basic Setup
- Configuration Parameters
- QLoRA (Quantized LoRA)
- Training Patterns
- Saving and Loading
- Merging Adapters
- Best Practices
- References
Files
SKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesSKILL.mdcore13,200 bytesreference/advanced-techniques.mddoc8,914 bytes
Read SKILL.md
# Using LoRA for Fine-tuning
LoRA (Low-Rank Adaptation) enables efficient fine-tuning by freezing pretrained weights and injecting small trainable matrices into transformer layers. This reduces trainable parameters to ~0.1% of the original model while maintaining performance.
## Table of Contents
- [Core Concepts](#core-concepts)
- [Basic Setup](#basic-setup)
- [Configuration Parameters](#configuration-parameters)
- [QLoRA (Quantized LoRA)](#qlora-quantized-lora)
- [Training Patterns](#training-patterns)
- [Saving and Loading](#saving-and-loading)
- [Merging Adapters](#merging-adapters)
- [Best Practices](#best-practices)mlx
When your agent uses it
Running and fine-tuning LLMs on Apple Silicon with MLX. Use when working with models locally on Mac, converting Hugging Face models to MLX format, fine-tuning with LoRA/QLoRA on Apple Silicon, or serving models via HTTP API.
What it loads into context
Frontmatter
61 tokensAlways available to the agent
SKILL.md
2,492 tokensWhen the skill is activated
Reference documents
4,138 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Concepts
- Installation
- Text Generation
- Interactive Chat
- Model Conversion
- Quantization
- Fine-tuning with LoRA
- Serving Models
- Best Practices
- References
Files
SKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesSKILL.mdcore9,103 bytesreference/fine-tuning.mddoc8,928 bytesreference/quantization.mddoc5,995 bytes
Read SKILL.md
# Using MLX for LLMs on Apple Silicon
MLX-LM is a Python package for running large language models on Apple Silicon, leveraging the MLX framework for optimized performance with unified memory architecture.
## Table of Contents
- [Core Concepts](#core-concepts)
- [Installation](#installation)
- [Text Generation](#text-generation)
- [Interactive Chat](#interactive-chat)
- [Model Conversion](#model-conversion)
- [Quantization](#quantization)
- [Fine-tuning with LoRA](#fine-tuning-with-lora)
- [Serving Models](#serving-models)prompt-engineering
When your agent uses it
Crafting effective prompts for LLMs. Use when designing prompts, improving output quality, structuring complex instructions, or debugging poor model responses.
What it loads into context
Frontmatter
37 tokensAlways available to the agent
SKILL.md
2,124 tokensWhen the skill is activated
Reference documents
0 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Principles
- Be Clear and Direct
- Use Examples (Multishot)
- Reasoning Guidance
- XML Tags
- Role Prompting
- Long Context
- Output Control
- Self-Verification
- Best Practices
- References
Files
SKILL.mdcore9,824 bytesSKILL.mdcore9,824 bytesSKILL.mdcore9,824 bytesSKILL.mdcore9,824 bytes
Read SKILL.md
# Prompt Engineering
Prompt engineering is the practice of designing inputs that guide LLMs to produce desired outputs. Effective prompts reduce errors, improve consistency, and unlock model capabilities.
## Table of Contents
- [Core Principles](#core-principles)
- [Be Clear and Direct](#be-clear-and-direct)
- [Use Examples (Multishot)](#use-examples-multishot)
- [Reasoning Guidance](#reasoning-guidance)
- [XML Tags](#xml-tags)
- [Role Prompting](#role-prompting)
- [Long Context](#long-context)
- [Output Control](#output-control)pytorch
When your agent uses it
Building and training neural networks with PyTorch. Use when implementing deep learning models, training loops, data pipelines, model optimization with torch.compile, distributed training, or deploying PyTorch models.
What it loads into context
Frontmatter
44 tokensAlways available to the agent
SKILL.md
2,486 tokensWhen the skill is activated
Reference documents
3,820 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Concepts
- Model Architecture
- Training Loop
- Data Loading
- Performance Optimization
- Distributed Training
- Saving and Loading
- Best Practices
- References
Files
SKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesSKILL.mdcore10,473 bytesreference/debugging.mddoc7,610 bytesreference/training-patterns.mddoc7,574 bytes
Read SKILL.md
# Using PyTorch
PyTorch is a deep learning framework with dynamic computation graphs, strong GPU acceleration, and Pythonic design. This skill covers practical patterns for building production-quality neural networks.
## Table of Contents
- [Core Concepts](#core-concepts)
- [Model Architecture](#model-architecture)
- [Training Loop](#training-loop)
- [Data Loading](#data-loading)
- [Performance Optimization](#performance-optimization)
- [Distributed Training](#distributed-training)
- [Saving and Loading](#saving-and-loading)
- [Best Practices](#best-practices)qlora
When your agent uses it
Memory-efficient fine-tuning with 4-bit quantization and LoRA adapters. Use when fine-tuning large models (7B+) on consumer GPUs, when VRAM is limited, or when standard LoRA still exceeds memory. Builds on the lora skill.
What it loads into context
Frontmatter
61 tokensAlways available to the agent
SKILL.md
3,275 tokensWhen the skill is activated
Reference documents
0 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Innovations
- BitsAndBytesConfig Deep Dive
- Memory Requirements
- Complete Training Example
- Inference and Merging
- Troubleshooting
- Best Practices
- References
Files
SKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytesSKILL.mdcore12,705 bytes
Read SKILL.md
# QLoRA: Quantized Low-Rank Adaptation
QLoRA enables fine-tuning of large language models on consumer GPUs by combining 4-bit quantization with LoRA adapters. A 65B model can be fine-tuned on a single 48GB GPU while matching 16-bit fine-tuning performance.
> **Prerequisites**: This skill assumes familiarity with LoRA. See the `lora` skill for LoRA fundamentals (LoraConfig, target_modules, training patterns).
## Table of Contents
- [Core Innovations](#core-innovations)
- [BitsAndBytesConfig Deep Dive](#bitsandbytesconfig-deep-dive)
- [Memory Requirements](#memory-requirements)
- [Complete Training Example](#complete-training-example)
- [Inference and Merging](#inference-and-merging)
- [Troubleshooting](#troubleshooting)rlhf
When your agent uses it
Understanding Reinforcement Learning from Human Feedback (RLHF) for aligning language models. Use when learning about preference data, reward modeling, policy optimization, or direct alignment algorithms like DPO.
What it loads into context
Frontmatter
45 tokensAlways available to the agent
SKILL.md
2,957 tokensWhen the skill is activated
Reference documents
6,482 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Concepts
- The RLHF Pipeline
- Preference Data
- Instruction Tuning
- Reward Modeling
- Policy Optimization
- Direct Alignment Algorithms
- Challenges
- Best Practices
- References
Files
SKILL.mdcore14,189 bytesreference/direct-alignment.mddoc9,221 bytesreference/policy-optimization.mddoc10,470 bytesreference/reward-modeling.mddoc7,884 bytes
Read SKILL.md
# Understanding RLHF
Reinforcement Learning from Human Feedback (RLHF) is a technique for aligning language models with human preferences. Rather than relying solely on next-token prediction, RLHF uses human judgment to guide model behavior toward helpful, harmless, and honest outputs.
## Table of Contents
- [Core Concepts](#core-concepts)
- [The RLHF Pipeline](#the-rlhf-pipeline)
- [Preference Data](#preference-data)
- [Instruction Tuning](#instruction-tuning)
- [Reward Modeling](#reward-modeling)
- [Policy Optimization](#policy-optimization)
- [Direct Alignment Algorithms](#direct-alignment-algorithms)
- [Challenges](#challenges)transformers
When your agent uses it
Loading and using pretrained models with Hugging Face Transformers. Use when working with pretrained models from the Hub, running inference with Pipeline API, fine-tuning models with Trainer, or handling text, vision, audio, and multimodal tasks.
What it loads into context
Frontmatter
53 tokensAlways available to the agent
SKILL.md
3,327 tokensWhen the skill is activated
Reference documents
3,220 tokensLoaded on request
Token counts are estimates. Scripts run separately; they are not counted as reference context.
The playbook
- Table of Contents
- Core Concepts
- Pipeline API
- Model Loading
- Inference Patterns
- Fine-tuning with Trainer
- Working with Modalities
- Memory and Performance
- Best Practices
- References
Files
SKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesSKILL.mdcore13,654 bytesreference/fine-tuning.mddoc12,531 bytes
Read SKILL.md
# Using Hugging Face Transformers
Transformers is the model-definition framework for state-of-the-art machine learning across text, vision, audio, and multimodal domains. It provides unified APIs for loading pretrained models, running inference, and fine-tuning.
## Table of Contents
- [Core Concepts](#core-concepts)
- [Pipeline API](#pipeline-api)
- [Model Loading](#model-loading)
- [Inference Patterns](#inference-patterns)
- [Fine-tuning with Trainer](#fine-tuning-with-trainer)
- [Working with Modalities](#working-with-modalities)
- [Memory and Performance](#memory-and-performance)
- [Best Practices](#best-practices)Skills included
- agents
- context-engineering
- lora
- mlx
- prompt-engineering
- pytorch
- qlora
- rlhf
- transformers
Repository facts
- License
- MIT
- Created
- Jan 5, 2026
- Last pushed
- May 6, 2026
- Forks
- 2
- Open issues
- 0
- Project activity
- Slowing
Before you install
Network or broad shellAutomated signals, not a guarantee
Checked Nov 5, 2026
Not fully analyzed: some files were too long, unreadable or missing. This never lowers the level.
Show 102 notes
skills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/agents/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/context-engineering/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzedskills/lora/SKILL.md:1Analysis incomplete: Referenced executable target unavailable; not analyzed
- No broad tool permissions
- 6 dangerous shell commandsin 1 file: skills/rlhf/reference/policy-optimization.md:338
- 64 network referencesin 9 files: skills/mlx/SKILL.md:38 skills/rlhf/SKILL.md:360 skills/context-engineering/SKILL.md:387 and 6 more
- No hidden or obfuscated text
- No instruction overrides
- No executable scripts
- 2 references to secretsin 1 file: skills/mlx/SKILL.md:295
- No requests to vote in Hot
Show all 72 signals by file
skills/agents/SKILL.md
- Network accessline 364
- [Building Effective Agents](https://www.anthropic.com/engineering/building-effective-agents) - Anthropic's guide to agent patterns and best practices - Network accessline 365
- [LangGraph Workflows & Agents](https://docs.langchain.com/oss/javascript/langgraph/workflows-agents) - LangGraph documentation on agent architectures - Network accessline 366
- [Model Context Protocol](https://modelcontextprotocol.io/docs) - Standard for exposing tools and resources to agents - Network accessline 367
- [ReAct: Synergizing Reasoning and Acting](https://arxiv.org/abs/2210.03629) - Paper introducing the ReAct prompting pattern
skills/context-engineering/SKILL.md
- Network accessline 387
- [reference/evaluation-strategies.md](reference/evaluation-strategies.md) - Testing context management effectiveness - [reference/summarization-patterns.md](reference/summarization-patterns.md) - Detailed summarization implementations - [E - Network accessline 387
- [reference/evaluation-strategies.md](reference/evaluation-strategies.md) - Testing context management effectiveness - [reference/summarization-patterns.md](reference/summarization-patterns.md) - Detailed summarization implementations - [E - Network accessline 388
- [reference/summarization-patterns.md](reference/summarization-patterns.md) - Detailed summarization implementations - [Effective Context Engineering for AI Agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai - Network accessline 388
- [reference/summarization-patterns.md](reference/summarization-patterns.md) - Detailed summarization implementations - [Effective Context Engineering for AI Agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai - Network accessline 389
- [Effective Context Engineering for AI Agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) - Network accessline 389
- [Effective Context Engineering for AI Agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) - Network accessline 390
- [OpenAI Session Memory for Agents](https://cookbook.openai.com/examples/agents_sdk/session_memory) - Network accessline 390
- [OpenAI Session Memory for Agents](https://cookbook.openai.com/examples/agents_sdk/session_memory)
skills/lora/SKILL.md
- Network accessline 453
External documentation: - [PEFT LoRA developer guide](https://huggingface.co/docs/peft/developer_guides/lora) - Network accessline 453
External documentation: - [PEFT LoRA developer guide](https://huggingface.co/docs/peft/developer_guides/lora) - Network accessline 454
- [PEFT LoRA developer guide](https://huggingface.co/docs/peft/developer_guides/lora) - Network accessline 454
- [PEFT LoRA developer guide](https://huggingface.co/docs/peft/developer_guides/lora)
skills/mlx/SKILL.md
- Network accessline 38
MLX-LM supports most popular architectures: Llama, Mistral, Qwen, Phi, Gemma, Cohere, and many more. Check the [mlx-community](https://huggingface.co/mlx-community) on Hugging Face for pre-converted models. - Network accessline 38
MLX-LM supports most popular architectures: Llama, Mistral, Qwen, Phi, Gemma, Cohere, and many more. Check the [mlx-community](https://huggingface.co/mlx-community) on Hugging Face for pre-converted models. - Network accessline 280
# Use with OpenAI client curl http://localhost:8080/v1/chat/completions \ - Network accessline 280
# Use with OpenAI client curl http://localhost:8080/v1/chat/completions \ - Network accessline 281
curl http://localhost:8080/v1/chat/completions \ - Network accessline 281
curl http://localhost:8080/v1/chat/completions \ - Network accessline 295
[Credential-related source omitted] - Network accessline 295
[Credential-related source omitted] - Access to secretsline 295
[Credential-related source omitted] - Access to secretsline 295
[Credential-related source omitted] - Network accessline 335
External documentation: - [MLX-LM GitHub](https://github.com/ml-explore/mlx-lm) - Network accessline 335
External documentation: - [MLX-LM GitHub](https://github.com/ml-explore/mlx-lm) - Network accessline 336
- [MLX-LM GitHub](https://github.com/ml-explore/mlx-lm) - Network accessline 336
- [MLX-LM GitHub](https://github.com/ml-explore/mlx-lm) - Network accessline 337
- [MLX-LM LoRA guide](https://github.com/ml-explore/mlx-lm/blob/main/mlx_lm/LORA.md) - Network accessline 337
- [MLX-LM LoRA guide](https://github.com/ml-explore/mlx-lm/blob/main/mlx_lm/LORA.md)
skills/prompt-engineering/SKILL.md
- Network accessline 346
- [Claude Prompt Engineering Guide](https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview) - Network accessline 346
- [Claude Prompt Engineering Guide](https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview) - Network accessline 347
- [GPT-5.2 Prompting Guide](https://cookbook.openai.com/examples/gpt-5/gpt-5-2_prompting_guide) - Network accessline 347
- [GPT-5.2 Prompting Guide](https://cookbook.openai.com/examples/gpt-5/gpt-5-2_prompting_guide) - Network accessline 348
- [Anthropic Prompt Library](https://docs.anthropic.com/en/prompt-library) - Network accessline 348
- [Anthropic Prompt Library](https://docs.anthropic.com/en/prompt-library)
skills/pytorch/SKILL.md
- Network accessline 409
External documentation: - [PyTorch stable documentation](https://pytorch.org/docs/stable/) - Network accessline 409
External documentation: - [PyTorch stable documentation](https://pytorch.org/docs/stable/) - Network accessline 410
- [PyTorch stable documentation](https://pytorch.org/docs/stable/) - Network accessline 410
- [PyTorch stable documentation](https://pytorch.org/docs/stable/)
skills/qlora/SKILL.md
- Network accessline 409
- [Transformers bitsandbytes quantization](https://huggingface.co/docs/transformers/quantization/bitsandbytes) - Network accessline 409
- [Transformers bitsandbytes quantization](https://huggingface.co/docs/transformers/quantization/bitsandbytes) - Network accessline 410
- [TRL SFTTrainer](https://huggingface.co/docs/trl/sft_trainer) - Network accessline 410
- [TRL SFTTrainer](https://huggingface.co/docs/trl/sft_trainer) - Network accessline 411
- [PEFT LoRA developer guide](https://huggingface.co/docs/peft/developer_guides/lora) - Network accessline 411
- [PEFT LoRA developer guide](https://huggingface.co/docs/peft/developer_guides/lora)
skills/rlhf/SKILL.md
- Network accessline 360
- [RLHF Book by Nathan Lambert](https://rlhfbook.com/) - Comprehensive textbook on RLHF - Network accessline 360
- [RLHF Book by Nathan Lambert](https://rlhfbook.com/) - Comprehensive textbook on RLHF - Network accessline 361
- [Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155) - InstructGPT paper - Network accessline 361
- [Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155) - InstructGPT paper - Network accessline 362
- [Direct Preference Optimization](https://arxiv.org/abs/2305.18290) - DPO paper - Network accessline 362
- [Direct Preference Optimization](https://arxiv.org/abs/2305.18290) - DPO paper - Network accessline 363
- [TRL documentation](https://huggingface.co/docs/trl/) - Current SFT, DPO, GRPO, reward modeling, RLOO, and PPO trainers - Network accessline 363
- [TRL documentation](https://huggingface.co/docs/trl/) - Current SFT, DPO, GRPO, reward modeling, RLOO, and PPO trainers - Network accessline 364
- [DeepSeekMath](https://arxiv.org/abs/2402.03300) - Introduces GRPO - Network accessline 364
- [DeepSeekMath](https://arxiv.org/abs/2402.03300) - Introduces GRPO
skills/rlhf/reference/policy-optimization.md
- Dangerous shell operationline 338
Early training: RM score ↑, Human eval ↑ - Dangerous shell operationline 338
Early training: RM score ↑, Human eval ↑ - Dangerous shell operationline 339
Over-optimization: RM score ↑↑, Human eval ↓ - Dangerous shell operationline 339
Over-optimization: RM score ↑↑, Human eval ↓ - Dangerous shell operationline 371
5. **Periodic human evaluation**: Check actual quality, not just RM scores - Dangerous shell operationline 371
5. **Periodic human evaluation**: Check actual quality, not just RM scores
skills/transformers/SKILL.md
- Network accessline 473
External documentation: - [Transformers documentation](https://huggingface.co/docs/transformers/) - Network accessline 473
External documentation: - [Transformers documentation](https://huggingface.co/docs/transformers/) - Network accessline 474
- [Transformers documentation](https://huggingface.co/docs/transformers/) - Network accessline 474
- [Transformers documentation](https://huggingface.co/docs/transformers/) - Network accessline 475
- [Loading models](https://huggingface.co/docs/transformers/models) - Network accessline 475
- [Loading models](https://huggingface.co/docs/transformers/models) - Network accessline 476
- [Bitsandbytes quantization](https://huggingface.co/docs/transformers/quantization/bitsandbytes) - Network accessline 476
- [Bitsandbytes quantization](https://huggingface.co/docs/transformers/quantization/bitsandbytes)
How the signals fit together
Facts from a static read of the files at the pinned commit. They describe what the code does, not why.
Some files could not be read; the facts below still hold.
Signal confidence: 72 facts
- Network access in prose of the SKILL.mdagents/SKILL.md:364Low confidence
- Network access in prose of the SKILL.mdagents/SKILL.md:365Low confidence
- Network access in prose of the SKILL.mdagents/SKILL.md:366Low confidence
- Network access in prose of the SKILL.mdagents/SKILL.md:367Low confidence
- Network access in prose of the SKILL.mdskills/context-engineering/SKILL.md:387Low confidence
- Network access in prose of the SKILL.mdcontext-engineering/SKILL.md:387Low confidence
- Network access in prose of the SKILL.mdskills/context-engineering/SKILL.md:388Low confidence
- Network access in prose of the SKILL.mdcontext-engineering/SKILL.md:388Low confidence
- Network access in prose of the SKILL.mdskills/context-engineering/SKILL.md:389Low confidence
- Network access in prose of the SKILL.mdcontext-engineering/SKILL.md:389Low confidence
- Network access in prose of the SKILL.mdskills/context-engineering/SKILL.md:390Low confidence
- Network access in prose of the SKILL.mdcontext-engineering/SKILL.md:390Low confidence
- Network access in prose of the SKILL.mdskills/lora/SKILL.md:453Low confidence
- Network access in prose of the SKILL.mdlora/SKILL.md:453Low confidence
- Network access in prose of the SKILL.mdskills/lora/SKILL.md:454Low confidence
- Network access in prose of the SKILL.mdlora/SKILL.md:454Low confidence
- Network access in prose of the SKILL.mdskills/mlx/SKILL.md:38Low confidence
- Network access in prose of the SKILL.mdmlx/SKILL.md:38Low confidence
- Network access in code of the SKILL.mdskills/mlx/SKILL.md:280High confidence
- Network access in code of the SKILL.mdmlx/SKILL.md:280High confidence
- Network access in code of the SKILL.mdskills/mlx/SKILL.md:281High confidence
- Network access in code of the SKILL.mdmlx/SKILL.md:281High confidence
- Network access in code of the SKILL.mdskills/mlx/SKILL.md:295High confidence
- Network access in code of the SKILL.mdmlx/SKILL.md:295High confidence
- Access to secrets in code of the SKILL.md: a variable, placeholder or call, not a stored valueskills/mlx/SKILL.md:295Low confidence
- Access to secrets in code of the SKILL.md: a variable, placeholder or call, not a stored valuemlx/SKILL.md:295Low confidence
- Network access in prose of the SKILL.mdskills/mlx/SKILL.md:335Low confidence
- Network access in prose of the SKILL.mdmlx/SKILL.md:335Low confidence
- Network access in prose of the SKILL.mdskills/mlx/SKILL.md:336Low confidence
- Network access in prose of the SKILL.mdmlx/SKILL.md:336Low confidence
- Network access in prose of the SKILL.mdskills/mlx/SKILL.md:337Low confidence
- Network access in prose of the SKILL.mdmlx/SKILL.md:337Low confidence
- Network access in prose of the SKILL.mdskills/prompt-engineering/SKILL.md:346Low confidence
- Network access in prose of the SKILL.mdprompt-engineering/SKILL.md:346Low confidence
- Network access in prose of the SKILL.mdskills/prompt-engineering/SKILL.md:347Low confidence
- Network access in prose of the SKILL.mdprompt-engineering/SKILL.md:347Low confidence
- Network access in prose of the SKILL.mdskills/prompt-engineering/SKILL.md:348Low confidence
- Network access in prose of the SKILL.mdprompt-engineering/SKILL.md:348Low confidence
- Network access in prose of the SKILL.mdskills/pytorch/SKILL.md:409Low confidence
- Network access in prose of the SKILL.mdpytorch/SKILL.md:409Low confidence
- Network access in prose of the SKILL.mdskills/pytorch/SKILL.md:410Low confidence
- Network access in prose of the SKILL.mdpytorch/SKILL.md:410Low confidence
- Network access in prose of the SKILL.mdskills/qlora/SKILL.md:409Low confidence
- Network access in prose of the SKILL.mdqlora/SKILL.md:409Low confidence
- Network access in prose of the SKILL.mdskills/qlora/SKILL.md:410Low confidence
- Network access in prose of the SKILL.mdqlora/SKILL.md:410Low confidence
- Network access in prose of the SKILL.mdskills/qlora/SKILL.md:411Low confidence
- Network access in prose of the SKILL.mdqlora/SKILL.md:411Low confidence
- Network access in prose of the SKILL.mdskills/rlhf/SKILL.md:360Low confidence
- Network access in prose of the SKILL.mdrlhf/SKILL.md:360Low confidence
and 22 more
Works with
| Agent | Status | Evidence | Install to |
|---|---|---|---|
| Claude Code | Native | Manifest .claude-plugin/marketplace.json | .claude/skills/<skill-name>/ |
| Codex | Compatible | Standard SKILL.md skills/agents/SKILL.md | .agents/skills/<skill-name>/ |
| Cursor | Compatible | Standard SKILL.md skills/agents/SKILL.md | .cursor/skills/<skill-name>/ |
| Gemini CLI | Compatible | Standard SKILL.md skills/agents/SKILL.md | .gemini/skills/<skill-name>/ |
| GitHub Copilot | Compatible | Standard SKILL.md skills/agents/SKILL.md | .github/skills/<skill-name>/ |
| OpenCode | Compatible | Standard SKILL.md skills/agents/SKILL.md | .opencode/skills/<skill-name>/ |
- Native The repository ships a manifest, folder or frontmatter field for this agent.
- Compatible A standard SKILL.md that this agent can read.
- Unknown No evidence either way.
Derived automatically from paths, manifests and frontmatter. Not tested on each agent.
