Use when errors occur deep in execution and you need to trace back to find the original trigger - systematically traces bugs backward through call stack, adding instrumentation when needed, to identify source of invalid data or incorrect behavior
Quick integration into your workflow with minimal setup
Active open-source community with continuous updates
MIT/Apache licensed for commercial and personal use
Customizable and extendable based on your needs
Download or copy the skill file from the source repository
Place the skill file in Claude's skills directory (usually ~/.claude/skills/）。
Restart Claude or run the reload command to load the skill
Tip: Read the documentation and code carefully before first use to understand functionality and permission requirements
All Skills from open-source community, preserving original authors' copyrights
obra__superpowers/skills/root-cause-tracing/skill.mdProven benefits and measurable impact
Reduce property prediction time from weeks to hours with pretrained GNNs.
Access 40+ curated datasets for instant model training and benchmarking.
Cut computational expenses with optimized PyTorch-based workflows.
Perfect for these scenarios
Predict drug-like properties using GNN models for early toxicity screening.
Analyze protein structures and interactions for target identification.
Design novel drug candidates with generative models for de novo discovery.
Optimize synthetic routes for drug candidates using ML-driven retrosynthesis.