Agent skill · research science · aiskillstore
ml-training-debugger
Diagnose machine learning training failures including loss divergence, mode collapse, gradient issues, architecture problems, and optimization failures. This skill spawns a specialist ML debugging ...
Why this skill is useful
Adds a specialized ML debugging agent that systematically analyzes training failures and provides evidence-based fixes, which the AI cannot generate on its own.
What it needs
Requires claude_agent_sdk installed locally. About 4k tokens when loaded. Last updated 2026-08-07. 405 stars on the source repository.
What this skill does
ML Training Debugger Version: 1.0.0 Type: Agent-based skill with SDK implementation Domain: Machine learning training diagnostics Description Diagnose machine learning training failures including loss divergence, mode collapse, gradient issues, architecture problems, and optimization failures. This skill spawns a specialist ML debugging agent that systematically analyzes training artifacts to identify root causes and propose evidence-based fixes. Use this skill when encountering training failures, when loss curves exhibit pathological behavior, when models produce degenerate outputs, when experiencing GPU memory issues, or when hyperparameter tuning produces inconsistent results. Triggers This skill activates when users request: "Debug my training run" "Why is my loss diverging?" "Model outputs are all the same token" "Training failed at epoch X" "Help diagnose mode collapse" "Why are gradients exploding/vanishing?" "Model not learning anything" Skill Architecture Skill Layer (Lightweight) The skill handles: 1. Detection: Identify ML training debugging requests 2. Context Gathering: Collect training logs, loss curves, model code 3. Agent Spawning: Invoke ML debugging specialist with context 4. Result Processing: Format diagnosis and fixes for user Agent Layer (Specialist) The ML debugging agent handles: 1. Systematic Analysis: Apply debugging methodology to artifacts 2. Root Cause Identification: Diagnose underlying issues 3. Fix Prioritization: Rank solutions by impact 4. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aiskillstore/ml-training-debugger