Materials for Unit 4 of AI301 (test and submit). This repo holds the week's runnable artifacts: the pr-precheck tool shell and its eval harness.
Overview
Materials for Unit 4 of AI301 (test and submit). This repo holds the week's runnable artifacts: the pr-precheck tool shell and its eval harness. All instructions live on the course portal (Overview, Activity, and Assignment tabs for this unit); this repo is the package those pages tell you to install and build out. - skill/: the pr-precheck tool as empty structured templates plus a fixed interface: read skill/CONTRACT.md first, top to bottom. The layout and the contract are given; the content of every file is yours: SKILL.md, the rubric, the evidence guide, the procedure (your voice guide pastes forward from Unit 2, and the sandbox scope.md comes written except its Repo: line, which you fill). Building the whole tool is this unit's deliverable. - eval/: the eval harness, the gold labels, and 24 frozen packages (20 scored plus the 4 calibration packages from the in-class activity). See eval/README.md for the full run and output guide. Copy the whole skill/ folder to ~/.
README
AI301 Unit 4 starter: pr-precheck
Materials for Unit 4 of AI301 (test and submit). This repo holds the week’s runnable artifacts: the pr-precheck tool shell and its eval harness. All instructions live on the course portal (Overview, Activity, and Assignment tabs for this unit); this repo is the package those pages tell you to install and build out.
What’s here
skill/: the pr-precheck tool as empty structured templates plus a fixed interface: readskill/CONTRACT.mdfirst, top to bottom. The layout and the contract are given; the content of every file is yours:SKILL.md, the rubric, the evidence guide, the procedure (your voice guide pastes forward from Unit 2, and the sandboxscope.mdcomes written except itsRepo:line, which you fill). Building the whole tool is this unit’s deliverable.eval/: the eval harness, the gold labels, and 24 frozen packages (20 scored plus the 4 calibration packages from the in-class activity). Seeeval/README.mdfor the full run and output guide.
Install the tool shell
Copy the whole skill/ folder to ~/.claude/skills/pr-precheck/
(create the folders if they do not exist). Build your tool inside that
installed copy and point the harness at it, so eval runs and live runs
share one canonical tool.
Run the eval
From eval/, with the Claude Code CLI installed:
python3 run_eval.py --skill path/to/your/tool-directory
--limit 3 gives a smoke run. Full docs: eval/README.md.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add codepath/ai301-unit4-starter