Week 12: Exercises & Checklist

Graded exercises

Easy: Run notebooks/01-harness-setup-and-comparison.ipynb to the final cell and record its harness-readiness score (0 to 3) in the Week 12 sheet of the Excel tracker. If it is below 2, install one more harness following its reference/skills/ guide and re-run.
Standard: Copy the SPEC.md template from the notebook into specs/eta-cli-SPEC.md, fill in every bracket (including the one refused tradeoff), then drive your first harness to build the CLI from it. Run the six-point test plan and paste the pass/fail line into the tracker.
Stretch: Run the same SPEC.md through a second harness and fill every cell of the comparison worksheet in the notebook (plan, tokens, cost, quality). Write the one-page comparison: which harness produced the better plan, at what cost, and which one you would pick for a team, and why.
Portfolio: Commit the working eta-cli (tests green), the filled SPEC.md, and the 1-page harness comparison to your fork. Add a README note stating which harness built it, how many tokens/dollars it cost, and the one tradeoff you refused. This is the first "I steered an agent from spec to shipped tool" artifact of the program.

Easy: The score is shutil.which over claude/opencode/dsh; a missing tool prints a graceful skip, so don't force an install just to make the number higher, install one you'll actually use this week.
Standard: Start from the refused-tradeoff line: state the one thing the CLI must never do (invent an ETA), then write the test-plan point that would catch it (unknown id → exit 2).
Stretch: Use the identical SPEC.md for harness B; the only thing you change is the harness. Fill all six worksheet columns, a comparison with blanks is a vibe.
Portfolio: The README note needs three facts: which harness, how many tokens/dollars, and the one tradeoff you refused. Those three are the whole "steered an agent" story.

[ ] Mon: Install and authenticate 2 harnesses (Claude Code + OpenCode or DSH); read the skill guides.
[ ] Tue: Write SPEC.md for the ETA CLI: user, constraints, one refused tradeoff, test plan.
[ ] Wed: Drive the agent to build the CLI; run the tests; iterate until green.
[ ] Thu: Repeat the same task in a second harness; compare plans, cost, and code quality.
[ ] Fri: Use case: ship the CLI; write the 1-page harness comparison for your team.
[ ] Sat: Take the Week 12 quiz (quiz.md), pass with 8/10; record the score in Notes.
[ ] Milestone: Update the Excel tracker; commit the CLI and comparison.