Query, analyze, and launch W&B projects with run histories and artifacts.
npx skills add https://github.com/wandb/skills --skill wandb-primary
Please help me install this Agent Skill. Skill: wandb-primary Source: https://github.com/wandb/skills/tree/main/skills/wandb-primary Command: npx skills add https://github.com/wandb/skills --skill wandb-primary
This Skill provides a cohesive set of utilities to query W&B project data, inspect training runs, Weave traces, and orchestrate launches, so you can diagnose performance, compare configurations, and reproduce or modify experiments with confidence.
Install the wandb-primary skill and run a quick diagnostic to list queues and fetch the latest run summary.
High-intent search queries and answers about installing and using this skill.
To analyze W&B runs, you can query project data to inspect training runs, read histories, and compare config diffs across experiments to diagnose performance and reproduce modified training launches.
Querying Weave traces and artifacts is done through fast access utilities that read and summarize run data, enabling you to inspect traces and report on machine learning experiment performance.
You can launch new runs from code artifacts or queued job artifacts by orchestrating W&B launches, which coordinate execution while providing precise resource controls for the experiments.
Yes, analyzing large W&B project histories requires the WandB Python SDK, the Weave SDK, and helper utilities to perform fast queries, read extensive run histories, and inspect artifacts.
Yes, you can inspect configuration diffs across W&B runs to diagnose performance issues, compare experiment settings directly, and reproduce or modify training runs with confidence.
When querying large W&B project histories and traces, processing relies entirely on the WandB Python SDK, the Weave SDK, and helper utilities, requiring these dependencies to execute fast queries and launches.
gil00pitaDefine Claude agents programmatically with TypeScript or Python using the AgentDefinition schema.
doodledoodAnalyze code diffs to identify and quantify test coverage gaps.
GenerativeAgentsManage Weights & Biases projects by analyzing runs, traces, and reports.
canvas-orgDiagnose execution traces and apply minimal fixes to Claude harness configs.
