This repository is a starting point for evaluating skills. It shows a simple workflow: run a task without a skill, run it again with a skill, verify the output locally, and compare the results. From ...

run a task with no-skill run the same task with a skill verify the output locally compare pass/fail, runtime, and event count It is intentionally tutorial-sized, not a full benchmark harness. uv run ...

Once you’ve found information on your topic, you need to assess its reliability and decide whether to use it in your research. This page introduces why evaluation matters and provides some quick ways ...