mirror of https://github.com/wandb/skills.git synced 2026-09-19 09:00:04 +08:00

Files

T

Zubin Aysola bafb5f39c8 Fix eval workflow: target main, add Claude Code matrix (#3 )

* Fix eval workflow: target main, add Claude Code agent matrix

- ref: wbaf-refact → ref: main
- Matrix strategy: benchmark both codex-mcp and claude-code
- Add ANTHROPIC_API_KEY for Claude Code agent
- Use all.yaml instead of codex-all.yaml (which no longer exists)
- Unique temp agent per matrix cell to avoid config collisions
- Separate README update job that aggregates results across agents

* Add install.sh, replace wandb skill with wandb-primary, update CI

- Add install.sh for easy skill installation (local or global Claude Code)
- Replace skills/wandb with skills/wandb-primary (comprehensive W&B + Weave skill
  with helper scripts and reference docs)
- Update README with installation instructions (npx skills + install script)
- Fix CI matrix: use codex and claude-code as base agents (not codex-mcp)
- Fix all wandb-primary-v2 path references to wandb-primary
- Remove stale .badges directory

* Overhaul CI: per-agent badges, wba-all eval source, no task name leaks

- Switch eval source from all.yaml to wba-all.yaml (W&B-specific tasks)
- Run update_all_eval.py before each eval to pick up new tasks
- Replace raw eval-log artifact with results-summary.json (counts only)
- Generate per-agent badge JSON files (.badges/codex.json, .badges/claude-code.json)
- Add Codex and Claude Code shield badges to README
- Keep eval_tasks input override for manual dispatch
- No task names, rubric content, or eval details leak to public artifacts

-- AI was used to create large parts of this PR

2026-03-05 13:18:06 -08:00

3.6 KiB

Raw Permalink Blame History

agent-skills

Agent skills for working with Weights & Biases. Explore projects, analyze runs, query metrics, and create reports — all from your coding agent.

Supported Coding Agents

These skills can be installed for any agent supported by skills.sh, including Claude Code, Cursor, Windsurf, Goose, and many more.

Prerequisites

A W&B API key

Installation

Quick Install

Using npx skills:

Local (current project):

npx skills add wandb/agent-skills --skill '*' --yes

Global (all projects):

npx skills add wandb/agent-skills --skill '*' --yes --global

To link skills to a specific agent (e.g. Claude Code):

npx skills add wandb/agent-skills --agent claude-code --skill '*' --yes --global

Install Script (Claude Code only)

Alternatively, clone the repo and use the install script:

git clone https://github.com/wandb/agent-skills.git
cd agent-skills

# Install for Claude Code in current directory (default)
./install.sh

# Install for Claude Code globally
./install.sh --global

# Force reinstall without prompts
./install.sh --force --yes

Flag	Description
`--global`, `-g`	Install globally instead of current directory
`--force`, `-f`	Overwrite skills with same names as this package
`--yes`, `-y`	Skip confirmation prompts

Usage

After installation, set your API key:

export WANDB_API_KEY=<your-key>

Then run your coding agent from the directory where you installed (for local installs) or from anywhere (for global installs).

Available Skills

Skill	Description	Status
`wandb-primary`	Comprehensive primary skill for W&B SDK (runs, metrics, artifacts, sweeps) and Weave SDK (traces, evaluations, scorers)	evaluated

Skill Format

Each skill is a directory under skills/ with a SKILL.md file:

skills/<name>/
└── SKILL.md          # YAML frontmatter + instructions

SKILL.md structure

---
name: my-skill
description: One-line description of when to use this skill.
---

# Skill Title

Instructions, code patterns, and best practices for the agent...

The frontmatter requires name and description. The body contains instructions that get injected into the agent's context.

Evaluation

Skills are automatically evaluated on merge to main using the WandBAgentFactory eval framework, benchmarked against both Codex and Claude Code agents. Results are tracked in Weave.

Contributing

Create skills/<your-skill>/SKILL.md with frontmatter and instructions
Open a PR — CI will run the skill through the eval suite
Merge to main to publish

To update an existing installation:

./install.sh --force

3.6 KiB Raw Permalink Blame History