liortesta

liortesta/ClawdAgent

Command Claude Code

Before writing ANY code, create a detailed plan with multi-agent input.

liortesta/ClawdAgent

Command Claude Code

Before considering this task done, provide concrete proof that everything works.

liortesta/ClawdAgent

Command Claude Code

Guide the user through setting up this Claude Code template in their project. Ask questions, configure, and verify — step by step.

liortesta/ClawdAgent

Command Claude Code

Spawn a NEW subagent with clean context to review the code changes. The reviewer MUST NOT see the original implementation discussion — only the final code.

liortesta/ClawdAgent

Command Claude Code

Run a 3-agent review on the current changes.

liortesta/ClawdAgent

Command Claude Code

Save the current session context to persistent memory, then tell the user to start a fresh conversation to save costs. The new session will have full access to everything saved here.

liortesta/ClawdAgent

Command Claude Code

Run the scout agent to find new tools, MCP servers, agent patterns, and Claude Code features.

liortesta/ClawdAgent

Command Claude Code

Run a complete diagnostic and automatic fix cycle.

liortesta/ClawdAgent

Command Claude Code

Analyze the project and set up the complete development environment.

liortesta/ClawdAgent

Command Claude Code

Before merging or deploying, run ALL of the following checks. Every gate must pass.

liortesta/ClawdAgent

Command Claude Code

Launch a coordinated multi-agent team with a shared task board and Leader→Workers pattern.

liortesta/ClawdAgent

Command Claude Code

Find high-leverage technical debt improvements in the codebase.

liortesta/ClawdAgent

Command Claude Code

Run a comprehensive verification of the entire MEGA setup. This is the FIRST thing to run after dropping this template into a new project.

liortesta/ClawdAgent

Command Claude Code

Use Playwright MCP to visually verify the UI that was just built or modified.

liortesta/ClawdAgent

Command Claude Code

Consolidate learnings from this session and update project memory.

liortesta/ClawdAgent

Skill Claude Code needs its repo

Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA/QLoRA. Single-file implementations, no abstraction layers.

liortesta/ClawdAgent

Skill Claude Code needs its repo

State-space model with O(n) complexity vs Transformers' O(n²). 5× faster inference, million-token sequences, no KV cache. Selective SSM with hardware-aware design. Mamba-1 (dstate=16) and Mamba-2 (dstate=128, multi-head). Models 130M-2.8B on HuggingFace.

liortesta/ClawdAgent

Skill Claude Code needs its repo

Educational GPT implementation in 300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).

liortesta/ClawdAgent

Skill Claude Code

RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.

liortesta/ClawdAgent

Skill Claude Code needs its repo

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

liortesta/ClawdAgent

Skill Claude Code

Fast tokenizers optimized for research and production. Rust-based implementation tokenizes 1GB in <20 seconds. Supports BPE, WordPiece, and Unigram algorithms. Train custom vocabularies, track alignments, handle padding/truncation. Integrates seamlessly with transformers. Use when you need high-performance…

liortesta/ClawdAgent

Skill Claude Code

Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory), deterministic vocabulary. Used by T5, ALBERT, XLNet, mBART. Train on raw text without pre-tokenization. Use when you need multilingual support, CJK languages, or…

liortesta/ClawdAgent

Skill Claude Code needs its repo

Expert guidance for fine-tuning LLMs with Axolotl - YAML configs, 100+ models, LoRA/QLoRA, DPO/KTO/ORPO/GRPO, multimodal support.

liortesta/ClawdAgent

Skill Claude Code

Expert guidance for fine-tuning LLMs with LLaMA-Factory - WebUI no-code, 100+ models, 2/3/4/5/6/8-bit QLoRA, multimodal support.