wandbwandbOfficialยท1 Agent Skills Included

skills

Track experiments, analyze traces, and automate training research

Analyzes Weights & Biases projects, runs, artifacts, and Weave traces to answer experiment questions fast. Converts Table artifacts into verified EvalTable previews and runs bounded hyperparameter searches through W&B Launch. Eliminates manual dashboard digging, unsafe data overwrites, and uncontrolled compute spending during model training and evaluation.
npx skills add wandb/skills --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install wandb skills?โ–ผ

Run `npx skills add wandb/skills --all -g -y` in your terminal, then set your W&B API key with `export WANDB_API_KEY=<your-key>`.

How to analyze W&B experiment runs automatically?โ–ผ

The wandb-primary skill lets your coding agent query runs, sweeps, artifacts, and Weave traces directly to answer questions about training progress and failures.

Can AI run hyperparameter searches on W&B?โ–ผ

Yes. The wandb-autoresearch skill runs bounded, serial training trials through W&B Launch, smoke-tests jobs first, and keeps resumable research state.

Does this work with Claude Code and Codex?โ–ผ

Yes. All skills follow the standard SKILL.md format and are benchmarked against Claude Code and Codex, with pass rates published in the repository.

Is it safe to convert W&B Tables to EvalTables?โ–ผ

Yes. The wandb-eval-tables skill only creates non-destructive previews with scan-first checks, row caps, and verification, never overwriting your source data.

Related Repositories in Data & Analytics

View All in Data & Analyticsโ†’