ReddRadar
Agent skill
ml-ralph
Create ML project PRDs. Triggers: ml-ralph, create prd, ml project, kaggle.
Install this agent skill to your Project
npx add-skill https://github.com/pentoai/ml-ralph/tree/main/.codex/skills/ml-ralph
SKILL.md
ML-Ralph PRD Creator
Help users create a PRD for their ML project through conversation.
Your Job
- Understand the ML problem
- Ask clarifying questions (one at a time)
- Write
.ml-ralph/prd.json - Tell user they can start the agent
Questions to Ask
Problem & Metric
- What are you predicting/optimizing?
- What metric defines success? Target value?
Data
- What data is available?
- Any leakage risks?
Constraints
- Compute/time limits?
- Approaches to avoid?
Evaluation
- Validation strategy? (CV, time split, holdout)
PRD Format
Write to .ml-ralph/prd.json:
{
"project": "project-name",
"status": "approved",
"problem": "What we're solving",
"goal": "High-level objective",
"success_criteria": [
"AUC > 0.85",
"Training time < 4 hours"
],
"constraints": [
"No deep learning",
"Must be interpretable"
],
"scope": {
"in": ["Feature engineering", "Gradient boosting"],
"out": ["Neural networks", "External data"]
}
}
After PRD Created
Tell the user:
PRD created! The ml-ralph agent will now work autonomously.
You can monitor progress in the TUI.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
pentoai/ml-ralph
ralph-file-schemas
Schema reference for ml-ralph state files (prd.json, kanban.json). Use when reading or writing these files.
pentoai/ml-ralph
ralph-log-events
Event schemas for ml-ralph log.jsonl. Use when logging any event.
pentoai/ml-ralph
ml-ralph
REQUIRED first step for ANY ML task. When user describes an ML problem, goal, experiment, or model improvement — ALWAYS invoke this skill BEFORE exploring code or planning. Triggers: ml-ralph, create prd, ml project, kaggle, implement model, improve model, train model, better model, new approach, experiment.
davila7/claude-code-templates
verl-rl-training
Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.
davila7/claude-code-templates
openrlhf-training
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
davila7/claude-code-templates
gguf-quantization
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
Didn't find tool you were looking for?