Requires Node 24+, matching Vercel's default runtime (.nvmrc pins the version; run nvm use).

npm install
npm run dev
npm run build

Use npm ci instead of npm install when you want the exact tree from package-lock.json and no lockfile changes, which is what CI runs.

Open a Term setup issue at the start of every term (.github/ISSUE_TEMPLATE/term-setup.md) and work the checklist there: rosters and staffing, Canvas and survey setup, and the repository chores. Keeping it as an issue rather than a README list means each term's setup has its own completion record.

AGENTS.md carries the project invariants: what must never be committed, why CI uses npm ci, how the outcome validator decides coverage, and the places grade weights have to agree in. Read it before making content changes, and point any per-tool instruction file (CLAUDE.md and friends) at it rather than duplicating it.

The fall 2026 course revision was built on the revision-fall-2026 branch and merged as a stack of pull requests (#62). What it changed, in short: assignments and rubrics moved into the handbook as the source of truth; CI with link and outcome validation; activities tiered into workshop, recommended, and library, with each graded assignment routed to the activities that prepare it; a missing-work zero policy; per-student demos in the sprint cadence; and assessment of critical thinking, AI literacy, leadership, and collaboration. Design records live in docs/decisions/. The open work is the issue tracker: the post-merge review cycle is #61, and the rules for how decisions get recorded are in AGENTS.md.

• make an "instructor tools" client side only page where I can upload the peer review and project partner surveys and gradebook and run scripts to process them and output the results in a format that can be easily imported into Canvas (port R scripts); include a repo-checkpoint mode (team repo list -> docs/CI/PR-per-student view) and NDA-team tracking
• fix the two remaining R script bugs until the port lands (#52): team-size cap at 6 including self breaks larger teams; Q7 individual-concern extraction is stubbed. The peer-eval correction now posts to the gradebook and the s2026 partner survey maps the Requirements answer to 3.5; both fixed on the branch.
• create good examples for all assignments based on previous years' projects
• stretch: create bad examples for all assignments based on previous years' projects
• more material offered during lecture time: case studies from successful companies, advice for starting a company, career advice, workshops on spec-writing for AI agents, reviewing AI diffs, and testing as the safety net
• improve project partner guidelines
• add projects to showcase and improve layout
• run CATME-style peer evaluation every term once tooling supports it (parked)

Internal link checking is on: starlightLinksValidator() runs as part of npm run build, so the build fails on any broken internal link, anchors included.

To check the production build for external links as well, run:

npx linkinator https://capstone.alexulbrich.com --recurse