v0.1.2

Fixes

• Reject unknown library and case names in run (#17)

v0.1.1

Fix

• Complete the PyPI page: README, links, keywords; silence the cache warning (#13)

v0.1.0

First release.

Feat

• Test whether coding agents load and follow your skills, in Docker Sandboxes, with
  every run traced in MLflow: skill-evals init, run, score, evaluate,
  validate and mlflow.
• Harnesses for Claude Code and Codex, with trace readers for both.
• Libraries installed by their own kits, fixtures that build the workspace, and YAML
  evals that combine libraries, harnesses, fixtures and scorers.
• Built-in scorers skill_invoked, skills_used and answer_matches, and support for
  your own scorers.

Fix

• Show messages instead of tracebacks for unknown eval and harness names, missing
  files, invalid YAML and scorer folders that are not packages (#11)